vm_page.h revision 127868
1/*
2 * Copyright (c) 1991, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * The Mach Operating System project at Carnegie-Mellon University.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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. All advertising materials mentioning features or use of this software
17 *    must display the following acknowledgement:
18 *	This product includes software developed by the University of
19 *	California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 *    may be used to endorse or promote products derived from this software
22 *    without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 *	from: @(#)vm_page.h	8.2 (Berkeley) 12/13/93
37 *
38 *
39 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
40 * All rights reserved.
41 *
42 * Authors: Avadis Tevanian, Jr., Michael Wayne Young
43 *
44 * Permission to use, copy, modify and distribute this software and
45 * its documentation is hereby granted, provided that both the copyright
46 * notice and this permission notice appear in all copies of the
47 * software, derivative works or modified versions, and any portions
48 * thereof, and that both notices appear in supporting documentation.
49 *
50 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
51 * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
52 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
53 *
54 * Carnegie Mellon requests users of this software to return to
55 *
56 *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
57 *  School of Computer Science
58 *  Carnegie Mellon University
59 *  Pittsburgh PA 15213-3890
60 *
61 * any improvements or extensions that they make and grant Carnegie the
62 * rights to redistribute these changes.
63 *
64 * $FreeBSD: head/sys/vm/vm_page.h 127868 2004-04-04 23:33:36Z alc $
65 */
66
67/*
68 *	Resident memory system definitions.
69 */
70
71#ifndef	_VM_PAGE_
72#define	_VM_PAGE_
73
74#if !defined(KLD_MODULE)
75#include "opt_vmpage.h"
76#endif
77
78#include <vm/pmap.h>
79
80/*
81 *	Management of resident (logical) pages.
82 *
83 *	A small structure is kept for each resident
84 *	page, indexed by page number.  Each structure
85 *	is an element of several lists:
86 *
87 *		A hash table bucket used to quickly
88 *		perform object/offset lookups
89 *
90 *		A list of all pages for a given object,
91 *		so they can be quickly deactivated at
92 *		time of deallocation.
93 *
94 *		An ordered list of pages due for pageout.
95 *
96 *	In addition, the structure contains the object
97 *	and offset to which this page belongs (for pageout),
98 *	and sundry status bits.
99 *
100 *	Fields in this structure are locked either by the lock on the
101 *	object that the page belongs to (O) or by the lock on the page
102 *	queues (P).
103 *
104 *	The 'valid' and 'dirty' fields are distinct.  A page may have dirty
105 *	bits set without having associated valid bits set.  This is used by
106 *	NFS to implement piecemeal writes.
107 */
108
109TAILQ_HEAD(pglist, vm_page);
110
111struct vm_page {
112	TAILQ_ENTRY(vm_page) pageq;	/* queue info for FIFO queue or free list (P) */
113	TAILQ_ENTRY(vm_page) listq;	/* pages in same object (O) 	*/
114	struct vm_page *left;		/* splay tree link (O)		*/
115	struct vm_page *right;		/* splay tree link (O)		*/
116
117	vm_object_t object;		/* which object am I in (O,P)*/
118	vm_pindex_t pindex;		/* offset into object (O,P) */
119	vm_paddr_t phys_addr;		/* physical address of page */
120	struct md_page md;		/* machine dependant stuff */
121	u_short	queue;			/* page queue index */
122	u_short	flags,			/* see below */
123		pc;			/* page color */
124	u_short wire_count;		/* wired down maps refs (P) */
125	short hold_count;		/* page hold count */
126	u_char	act_count;		/* page usage count */
127	u_char	busy;			/* page busy count */
128	/* NOTE that these must support one bit per DEV_BSIZE in a page!!! */
129	/* so, on normal X86 kernels, they must be at least 8 bits wide */
130#if PAGE_SIZE == 4096
131	u_char	valid;			/* map of valid DEV_BSIZE chunks (O) */
132	u_char	dirty;			/* map of dirty DEV_BSIZE chunks */
133#elif PAGE_SIZE == 8192
134	u_short	valid;			/* map of valid DEV_BSIZE chunks (O) */
135	u_short	dirty;			/* map of dirty DEV_BSIZE chunks */
136#elif PAGE_SIZE == 16384
137	u_int valid;			/* map of valid DEV_BSIZE chunks (O) */
138	u_int dirty;			/* map of dirty DEV_BSIZE chunks */
139#elif PAGE_SIZE == 32768
140	u_long valid;			/* map of valid DEV_BSIZE chunks (O) */
141	u_long dirty;			/* map of dirty DEV_BSIZE chunks */
142#endif
143	u_int cow;			/* page cow mapping count */
144};
145
146/* Make sure that u_long is at least 64 bits when PAGE_SIZE is 32K. */
147#if PAGE_SIZE == 32768
148#ifdef CTASSERT
149CTASSERT(sizeof(u_long) >= 8);
150#endif
151#endif
152
153/*
154 * note: currently use SWAPBLK_NONE as an absolute value rather then
155 * a flag bit.
156 */
157
158#define SWAPBLK_MASK	((daddr_t)((u_daddr_t)-1 >> 1))		/* mask */
159#define SWAPBLK_NONE	((daddr_t)((u_daddr_t)SWAPBLK_MASK + 1))/* flag */
160
161#if !defined(KLD_MODULE)
162/*
163 * Page coloring parameters
164 */
165/* Each of PQ_FREE, and PQ_CACHE have PQ_HASH_SIZE entries */
166
167/* Backward compatibility for existing PQ_*CACHE config options. */
168#if !defined(PQ_CACHESIZE)
169#if defined(PQ_HUGECACHE)
170#define PQ_CACHESIZE 1024
171#elif defined(PQ_LARGECACHE)
172#define PQ_CACHESIZE 512
173#elif defined(PQ_MEDIUMCACHE)
174#define PQ_CACHESIZE 256
175#elif defined(PQ_NORMALCACHE)
176#define PQ_CACHESIZE 64
177#elif defined(PQ_NOOPT)
178#define PQ_CACHESIZE 0
179#else
180#define PQ_CACHESIZE 128
181#endif
182#endif			/* !defined(PQ_CACHESIZE) */
183
184#if PQ_CACHESIZE >= 1024
185#define PQ_PRIME1 31	/* Prime number somewhat less than PQ_HASH_SIZE */
186#define PQ_PRIME2 23	/* Prime number somewhat less than PQ_HASH_SIZE */
187#define PQ_L2_SIZE 256	/* A number of colors opt for 1M cache */
188
189#elif PQ_CACHESIZE >= 512
190#define PQ_PRIME1 31	/* Prime number somewhat less than PQ_HASH_SIZE */
191#define PQ_PRIME2 23	/* Prime number somewhat less than PQ_HASH_SIZE */
192#define PQ_L2_SIZE 128	/* A number of colors opt for 512K cache */
193
194#elif PQ_CACHESIZE >= 256
195#define PQ_PRIME1 13	/* Prime number somewhat less than PQ_HASH_SIZE */
196#define PQ_PRIME2 7	/* Prime number somewhat less than PQ_HASH_SIZE */
197#define PQ_L2_SIZE 64	/* A number of colors opt for 256K cache */
198
199#elif PQ_CACHESIZE >= 128
200#define PQ_PRIME1 9	/* Produces a good PQ_L2_SIZE/3 + PQ_PRIME1 */
201#define PQ_PRIME2 5	/* Prime number somewhat less than PQ_HASH_SIZE */
202#define PQ_L2_SIZE 32	/* A number of colors opt for 128k cache */
203
204#elif PQ_CACHESIZE >= 64
205#define PQ_PRIME1 5	/* Prime number somewhat less than PQ_HASH_SIZE */
206#define PQ_PRIME2 3	/* Prime number somewhat less than PQ_HASH_SIZE */
207#define PQ_L2_SIZE 16	/* A reasonable number of colors (opt for 64K cache) */
208
209#else
210#define PQ_PRIME1 1	/* Disable page coloring. */
211#define PQ_PRIME2 1
212#define PQ_L2_SIZE 1
213
214#endif
215
216#define PQ_L2_MASK (PQ_L2_SIZE - 1)
217
218#define PQ_NONE 0
219#define PQ_FREE	1
220#define PQ_INACTIVE (1 + 1*PQ_L2_SIZE)
221#define PQ_ACTIVE (2 + 1*PQ_L2_SIZE)
222#define PQ_CACHE (3 + 1*PQ_L2_SIZE)
223#define PQ_HOLD  (3 + 2*PQ_L2_SIZE)
224#define PQ_COUNT (4 + 2*PQ_L2_SIZE)
225
226struct vpgqueues {
227	struct pglist pl;
228	int	*cnt;
229	int	lcnt;
230};
231
232extern struct vpgqueues vm_page_queues[PQ_COUNT];
233extern struct mtx vm_page_queue_free_mtx;
234
235#endif			/* !defined(KLD_MODULE) */
236
237/*
238 * These are the flags defined for vm_page.
239 *
240 * Note: PG_FILLED and PG_DIRTY are added for the filesystems.
241 *
242 * Note: PG_UNMANAGED (used by OBJT_PHYS) indicates that the page is
243 * 	 not under PV management but otherwise should be treated as a
244 *	 normal page.  Pages not under PV management cannot be paged out
245 *	 via the object/vm_page_t because there is no knowledge of their
246 *	 pte mappings, nor can they be removed from their objects via
247 *	 the object, and such pages are also not on any PQ queue.
248 */
249#define	PG_BUSY		0x0001		/* page is in transit (O) */
250#define	PG_WANTED	0x0002		/* someone is waiting for page (O) */
251#define PG_WINATCFLS	0x0004		/* flush dirty page on inactive q */
252#define	PG_FICTITIOUS	0x0008		/* physical page doesn't exist (O) */
253#define	PG_WRITEABLE	0x0010		/* page is mapped writeable */
254#define	PG_ZERO		0x0040		/* page is zeroed */
255#define PG_REFERENCED	0x0080		/* page has been referenced */
256#define PG_CLEANCHK	0x0100		/* page will be checked for cleaning */
257#define PG_SWAPINPROG	0x0200		/* swap I/O in progress on page	     */
258#define PG_NOSYNC	0x0400		/* do not collect for syncer */
259#define PG_UNMANAGED	0x0800		/* No PV management for page */
260#define PG_MARKER	0x1000		/* special queue marker page */
261#define	PG_SLAB		0x2000		/* object pointer is actually a slab */
262
263/*
264 * Misc constants.
265 */
266#define ACT_DECLINE		1
267#define ACT_ADVANCE		3
268#define ACT_INIT		5
269#define ACT_MAX			64
270
271#ifdef _KERNEL
272/*
273 * Each pageable resident page falls into one of four lists:
274 *
275 *	free
276 *		Available for allocation now.
277 *
278 * The following are all LRU sorted:
279 *
280 *	cache
281 *		Almost available for allocation. Still in an
282 *		object, but clean and immediately freeable at
283 *		non-interrupt times.
284 *
285 *	inactive
286 *		Low activity, candidates for reclamation.
287 *		This is the list of pages that should be
288 *		paged out next.
289 *
290 *	active
291 *		Pages that are "active" i.e. they have been
292 *		recently referenced.
293 *
294 *	zero
295 *		Pages that are really free and have been pre-zeroed
296 *
297 */
298
299extern int vm_page_zero_count;
300
301extern vm_page_t vm_page_array;		/* First resident page in table */
302extern int vm_page_array_size;		/* number of vm_page_t's */
303extern long first_page;			/* first physical page number */
304
305#define VM_PAGE_TO_PHYS(entry)	((entry)->phys_addr)
306
307#define PHYS_TO_VM_PAGE(pa) \
308		(&vm_page_array[atop(pa) - first_page ])
309
310extern struct mtx vm_page_queue_mtx;
311#define vm_page_lock_queues()   mtx_lock(&vm_page_queue_mtx)
312#define vm_page_unlock_queues() mtx_unlock(&vm_page_queue_mtx)
313
314#if PAGE_SIZE == 4096
315#define VM_PAGE_BITS_ALL 0xffu
316#elif PAGE_SIZE == 8192
317#define VM_PAGE_BITS_ALL 0xffffu
318#elif PAGE_SIZE == 16384
319#define VM_PAGE_BITS_ALL 0xffffffffu
320#elif PAGE_SIZE == 32768
321#define VM_PAGE_BITS_ALL 0xfffffffffffffffflu
322#endif
323
324/* page allocation classes: */
325#define VM_ALLOC_NORMAL		0
326#define VM_ALLOC_INTERRUPT	1
327#define VM_ALLOC_SYSTEM		2
328#define	VM_ALLOC_CLASS_MASK	3
329/* page allocation flags: */
330#define	VM_ALLOC_WIRED		0x0020	/* non pageable */
331#define	VM_ALLOC_ZERO		0x0040	/* Try to obtain a zeroed page */
332#define	VM_ALLOC_RETRY		0x0080	/* vm_page_grab() only */
333#define	VM_ALLOC_NOOBJ		0x0100	/* No associated object */
334
335void vm_page_flag_set(vm_page_t m, unsigned short bits);
336void vm_page_flag_clear(vm_page_t m, unsigned short bits);
337void vm_page_busy(vm_page_t m);
338void vm_page_flash(vm_page_t m);
339void vm_page_io_start(vm_page_t m);
340void vm_page_io_finish(vm_page_t m);
341void vm_page_hold(vm_page_t mem);
342void vm_page_unhold(vm_page_t mem);
343void vm_page_free(vm_page_t m);
344void vm_page_free_zero(vm_page_t m);
345int vm_page_sleep_if_busy(vm_page_t m, int also_m_busy, const char *msg);
346void vm_page_dirty(vm_page_t m);
347void vm_page_wakeup(vm_page_t m);
348
349void vm_pageq_init(void);
350vm_page_t vm_pageq_add_new_page(vm_paddr_t pa);
351void vm_pageq_enqueue(int queue, vm_page_t m);
352void vm_pageq_remove_nowakeup(vm_page_t m);
353void vm_pageq_remove(vm_page_t m);
354vm_page_t vm_pageq_find(int basequeue, int index, boolean_t prefer_zero);
355void vm_pageq_requeue(vm_page_t m);
356
357void vm_page_activate (vm_page_t);
358vm_page_t vm_page_alloc (vm_object_t, vm_pindex_t, int);
359vm_page_t vm_page_grab (vm_object_t, vm_pindex_t, int);
360void vm_page_cache (register vm_page_t);
361int vm_page_try_to_cache (vm_page_t);
362int vm_page_try_to_free (vm_page_t);
363void vm_page_dontneed (register vm_page_t);
364void vm_page_deactivate (vm_page_t);
365void vm_page_insert (vm_page_t, vm_object_t, vm_pindex_t);
366vm_page_t vm_page_lookup (vm_object_t, vm_pindex_t);
367void vm_page_remove (vm_page_t);
368void vm_page_rename (vm_page_t, vm_object_t, vm_pindex_t);
369vm_page_t vm_page_select_cache(int);
370vm_page_t vm_page_splay(vm_pindex_t, vm_page_t);
371vm_offset_t vm_page_startup(vm_offset_t vaddr);
372void vm_page_unmanage (vm_page_t);
373void vm_page_unwire (vm_page_t, int);
374void vm_page_wire (vm_page_t);
375void vm_page_set_validclean (vm_page_t, int, int);
376void vm_page_clear_dirty (vm_page_t, int, int);
377void vm_page_set_invalid (vm_page_t, int, int);
378int vm_page_is_valid (vm_page_t, int, int);
379void vm_page_test_dirty (vm_page_t);
380int vm_page_bits (int, int);
381void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid);
382void vm_page_free_toq(vm_page_t m);
383void vm_page_zero_idle_wakeup(void);
384void vm_page_cowfault (vm_page_t);
385void vm_page_cowsetup (vm_page_t);
386void vm_page_cowclear (vm_page_t);
387
388/*
389 *	vm_page_undirty:
390 *
391 *	Set page to not be dirty.  Note: does not clear pmap modify bits
392 */
393static __inline void
394vm_page_undirty(vm_page_t m)
395{
396	m->dirty = 0;
397}
398
399#endif				/* _KERNEL */
400#endif				/* !_VM_PAGE_ */
401