xref: /freebsd/sys/vm/vm_page.h (revision d056fa046c6a91b90cd98165face0e42a33a5173)
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  * 4. 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.
19  *
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
30  * SUCH DAMAGE.
31  *
32  *	from: @(#)vm_page.h	8.2 (Berkeley) 12/13/93
33  *
34  *
35  * Copyright (c) 1987, 1990 Carnegie-Mellon University.
36  * All rights reserved.
37  *
38  * Authors: Avadis Tevanian, Jr., Michael Wayne Young
39  *
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.
45  *
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.
49  *
50  * Carnegie Mellon requests users of this software to return to
51  *
52  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
53  *  School of Computer Science
54  *  Carnegie Mellon University
55  *  Pittsburgh PA 15213-3890
56  *
57  * any improvements or extensions that they make and grant Carnegie the
58  * rights to redistribute these changes.
59  *
60  * $FreeBSD$
61  */
62 
63 /*
64  *	Resident memory system definitions.
65  */
66 
67 #ifndef	_VM_PAGE_
68 #define	_VM_PAGE_
69 
70 #include <vm/pmap.h>
71 
72 /*
73  *	Management of resident (logical) pages.
74  *
75  *	A small structure is kept for each resident
76  *	page, indexed by page number.  Each structure
77  *	is an element of several lists:
78  *
79  *		A hash table bucket used to quickly
80  *		perform object/offset lookups
81  *
82  *		A list of all pages for a given object,
83  *		so they can be quickly deactivated at
84  *		time of deallocation.
85  *
86  *		An ordered list of pages due for pageout.
87  *
88  *	In addition, the structure contains the object
89  *	and offset to which this page belongs (for pageout),
90  *	and sundry status bits.
91  *
92  *	Fields in this structure are locked either by the lock on the
93  *	object that the page belongs to (O) or by the lock on the page
94  *	queues (P).
95  *
96  *	The 'valid' and 'dirty' fields are distinct.  A page may have dirty
97  *	bits set without having associated valid bits set.  This is used by
98  *	NFS to implement piecemeal writes.
99  */
100 
101 TAILQ_HEAD(pglist, vm_page);
102 
103 struct vm_page {
104 	TAILQ_ENTRY(vm_page) pageq;	/* queue info for FIFO queue or free list (P) */
105 	TAILQ_ENTRY(vm_page) listq;	/* pages in same object (O) 	*/
106 	struct vm_page *left;		/* splay tree link (O)		*/
107 	struct vm_page *right;		/* splay tree link (O)		*/
108 
109 	vm_object_t object;		/* which object am I in (O,P)*/
110 	vm_pindex_t pindex;		/* offset into object (O,P) */
111 	vm_paddr_t phys_addr;		/* physical address of page */
112 	struct md_page md;		/* machine dependant stuff */
113 	u_short	queue;			/* page queue index */
114 	u_short	flags,			/* see below */
115 		pc;			/* page color */
116 	u_short wire_count;		/* wired down maps refs (P) */
117 	u_int cow;			/* page cow mapping count */
118 	short hold_count;		/* page hold count */
119 	u_char	act_count;		/* page usage count */
120 	u_char	busy;			/* page busy count (O) */
121 	/* NOTE that these must support one bit per DEV_BSIZE in a page!!! */
122 	/* so, on normal X86 kernels, they must be at least 8 bits wide */
123 #if PAGE_SIZE == 4096
124 	u_char	valid;			/* map of valid DEV_BSIZE chunks (O) */
125 	u_char	dirty;			/* map of dirty DEV_BSIZE chunks */
126 #elif PAGE_SIZE == 8192
127 	u_short	valid;			/* map of valid DEV_BSIZE chunks (O) */
128 	u_short	dirty;			/* map of dirty DEV_BSIZE chunks */
129 #elif PAGE_SIZE == 16384
130 	u_int valid;			/* map of valid DEV_BSIZE chunks (O) */
131 	u_int dirty;			/* map of dirty DEV_BSIZE chunks */
132 #elif PAGE_SIZE == 32768
133 	u_long valid;			/* map of valid DEV_BSIZE chunks (O) */
134 	u_long dirty;			/* map of dirty DEV_BSIZE chunks */
135 #endif
136 };
137 
138 /* Make sure that u_long is at least 64 bits when PAGE_SIZE is 32K. */
139 #if PAGE_SIZE == 32768
140 #ifdef CTASSERT
141 CTASSERT(sizeof(u_long) >= 8);
142 #endif
143 #endif
144 
145 /* PQ_CACHE and PQ_FREE represents a PQ_NUMCOLORS consecutive queue. */
146 #define PQ_NONE		0
147 #define PQ_FREE		1
148 #define PQ_INACTIVE	(page_queue_coloring.inactive)
149 #define PQ_ACTIVE	(page_queue_coloring.active)
150 #define PQ_CACHE	(page_queue_coloring.cache)
151 #define PQ_HOLD		(page_queue_coloring.hold)
152 #define PQ_COUNT	(page_queue_coloring.count)
153 #define PQ_MAXCOLORS	1024
154 #define PQ_MAXCOUNT	(4 + 2 * PQ_MAXCOLORS)
155 #define PQ_NUMCOLORS	(page_queue_coloring.numcolors)
156 #define PQ_PRIME1	(page_queue_coloring.prime1)
157 #define PQ_PRIME2	(page_queue_coloring.prime2)
158 #define PQ_COLORMASK	(page_queue_coloring.colormask)
159 #define PQ_MAXLENGTH	(page_queue_coloring.maxlength)
160 
161 /* Returns the real queue a page is on. */
162 #define VM_PAGE_GETQUEUE(m)	((m)->queue)
163 
164 /* Returns the well known queue a page is on. */
165 #define VM_PAGE_GETKNOWNQUEUE1(m)	((m)->queue - (m)->pc)
166 #define VM_PAGE_GETKNOWNQUEUE2(m)	VM_PAGE_GETQUEUE(m)
167 
168 /* Given the real queue number and a page color return the well know queue. */
169 #define VM_PAGE_RESOLVEQUEUE(m, q)	((q) - (m)->pc)
170 
171 /* Returns true if the page is in the named well known queue. */
172 #define VM_PAGE_INQUEUE1(m, q)	(VM_PAGE_GETKNOWNQUEUE1(m) == (q))
173 #define VM_PAGE_INQUEUE2(m, q)	(VM_PAGE_GETKNOWNQUEUE2(m) == (q))
174 
175 /* Sets the queue a page is on. */
176 #define VM_PAGE_SETQUEUE1(m, q)	(VM_PAGE_GETQUEUE(m) = (q) + (m)->pc)
177 #define VM_PAGE_SETQUEUE2(m, q)	(VM_PAGE_GETQUEUE(m) = (q))
178 
179 struct vpgqueues {
180 	struct pglist pl;
181 	int	*cnt;
182 	int	lcnt;
183 };
184 
185 struct pq_coloring {
186 	int numcolors;
187 	int colormask;
188 	int prime1;
189 	int prime2;
190 	int inactive;
191 	int active;
192 	int cache;
193 	int hold;
194 	int count;
195 	int maxlength;
196 };
197 
198 extern struct vpgqueues vm_page_queues[PQ_MAXCOUNT];
199 extern struct mtx vm_page_queue_free_mtx;
200 extern struct pq_coloring page_queue_coloring;
201 
202 /*
203  * These are the flags defined for vm_page.
204  *
205  * Note: PG_UNMANAGED (used by OBJT_PHYS) indicates that the page is
206  * 	 not under PV management but otherwise should be treated as a
207  *	 normal page.  Pages not under PV management cannot be paged out
208  *	 via the object/vm_page_t because there is no knowledge of their
209  *	 pte mappings, nor can they be removed from their objects via
210  *	 the object, and such pages are also not on any PQ queue.
211  */
212 #define	PG_BUSY		0x0001		/* page is in transit (O) */
213 #define	PG_WANTED	0x0002		/* someone is waiting for page (O) */
214 #define PG_WINATCFLS	0x0004		/* flush dirty page on inactive q */
215 #define	PG_FICTITIOUS	0x0008		/* physical page doesn't exist (O) */
216 #define	PG_WRITEABLE	0x0010		/* page is mapped writeable */
217 #define	PG_ZERO		0x0040		/* page is zeroed */
218 #define PG_REFERENCED	0x0080		/* page has been referenced */
219 #define PG_CLEANCHK	0x0100		/* page will be checked for cleaning */
220 #define PG_SWAPINPROG	0x0200		/* swap I/O in progress on page	     */
221 #define PG_NOSYNC	0x0400		/* do not collect for syncer */
222 #define PG_UNMANAGED	0x0800		/* No PV management for page */
223 #define PG_MARKER	0x1000		/* special queue marker page */
224 #define	PG_SLAB		0x2000		/* object pointer is actually a slab */
225 
226 /*
227  * Misc constants.
228  */
229 #define ACT_DECLINE		1
230 #define ACT_ADVANCE		3
231 #define ACT_INIT		5
232 #define ACT_MAX			64
233 
234 #ifdef _KERNEL
235 /*
236  * Each pageable resident page falls into one of four lists:
237  *
238  *	free
239  *		Available for allocation now.
240  *
241  * The following are all LRU sorted:
242  *
243  *	cache
244  *		Almost available for allocation. Still in an
245  *		object, but clean and immediately freeable at
246  *		non-interrupt times.
247  *
248  *	inactive
249  *		Low activity, candidates for reclamation.
250  *		This is the list of pages that should be
251  *		paged out next.
252  *
253  *	active
254  *		Pages that are "active" i.e. they have been
255  *		recently referenced.
256  *
257  *	zero
258  *		Pages that are really free and have been pre-zeroed
259  *
260  */
261 
262 extern int vm_page_zero_count;
263 
264 extern vm_page_t vm_page_array;		/* First resident page in table */
265 extern int vm_page_array_size;		/* number of vm_page_t's */
266 extern long first_page;			/* first physical page number */
267 
268 #define VM_PAGE_TO_PHYS(entry)	((entry)->phys_addr)
269 
270 #define PHYS_TO_VM_PAGE(pa) \
271 		(&vm_page_array[atop(pa) - first_page ])
272 
273 extern struct mtx vm_page_queue_mtx;
274 #define vm_page_lock_queues()   mtx_lock(&vm_page_queue_mtx)
275 #define vm_page_unlock_queues() mtx_unlock(&vm_page_queue_mtx)
276 
277 #if PAGE_SIZE == 4096
278 #define VM_PAGE_BITS_ALL 0xffu
279 #elif PAGE_SIZE == 8192
280 #define VM_PAGE_BITS_ALL 0xffffu
281 #elif PAGE_SIZE == 16384
282 #define VM_PAGE_BITS_ALL 0xffffffffu
283 #elif PAGE_SIZE == 32768
284 #define VM_PAGE_BITS_ALL 0xfffffffffffffffflu
285 #endif
286 
287 /* page allocation classes: */
288 #define VM_ALLOC_NORMAL		0
289 #define VM_ALLOC_INTERRUPT	1
290 #define VM_ALLOC_SYSTEM		2
291 #define	VM_ALLOC_CLASS_MASK	3
292 /* page allocation flags: */
293 #define	VM_ALLOC_WIRED		0x0020	/* non pageable */
294 #define	VM_ALLOC_ZERO		0x0040	/* Try to obtain a zeroed page */
295 #define	VM_ALLOC_RETRY		0x0080	/* vm_page_grab() only */
296 #define	VM_ALLOC_NOOBJ		0x0100	/* No associated object */
297 #define	VM_ALLOC_NOBUSY		0x0200	/* Do not busy the page */
298 
299 void vm_page_flag_set(vm_page_t m, unsigned short bits);
300 void vm_page_flag_clear(vm_page_t m, unsigned short bits);
301 void vm_page_busy(vm_page_t m);
302 void vm_page_flash(vm_page_t m);
303 void vm_page_io_start(vm_page_t m);
304 void vm_page_io_finish(vm_page_t m);
305 void vm_page_hold(vm_page_t mem);
306 void vm_page_unhold(vm_page_t mem);
307 void vm_page_free(vm_page_t m);
308 void vm_page_free_zero(vm_page_t m);
309 int vm_page_sleep_if_busy(vm_page_t m, int also_m_busy, const char *msg);
310 void vm_page_dirty(vm_page_t m);
311 void vm_page_wakeup(vm_page_t m);
312 
313 void vm_pageq_init(void);
314 vm_page_t vm_pageq_add_new_page(vm_paddr_t pa);
315 void vm_pageq_enqueue(int queue, vm_page_t m);
316 void vm_pageq_remove_nowakeup(vm_page_t m);
317 void vm_pageq_remove(vm_page_t m);
318 vm_page_t vm_pageq_find(int basequeue, int index, boolean_t prefer_zero);
319 void vm_pageq_requeue(vm_page_t m);
320 
321 void vm_page_activate (vm_page_t);
322 vm_page_t vm_page_alloc (vm_object_t, vm_pindex_t, int);
323 vm_page_t vm_page_alloc_contig (vm_pindex_t, vm_paddr_t, vm_paddr_t,
324 	    vm_offset_t, vm_offset_t);
325 void vm_page_release_contig (vm_page_t, vm_pindex_t);
326 vm_page_t vm_page_grab (vm_object_t, vm_pindex_t, int);
327 void vm_page_cache (register vm_page_t);
328 int vm_page_try_to_cache (vm_page_t);
329 int vm_page_try_to_free (vm_page_t);
330 void vm_page_dontneed (register vm_page_t);
331 void vm_page_deactivate (vm_page_t);
332 void vm_page_insert (vm_page_t, vm_object_t, vm_pindex_t);
333 vm_page_t vm_page_lookup (vm_object_t, vm_pindex_t);
334 void vm_page_remove (vm_page_t);
335 void vm_page_rename (vm_page_t, vm_object_t, vm_pindex_t);
336 vm_page_t vm_page_select_cache(int);
337 vm_page_t vm_page_splay(vm_pindex_t, vm_page_t);
338 vm_offset_t vm_page_startup(vm_offset_t vaddr);
339 void vm_page_unmanage (vm_page_t);
340 void vm_page_unwire (vm_page_t, int);
341 void vm_page_wire (vm_page_t);
342 void vm_page_set_validclean (vm_page_t, int, int);
343 void vm_page_clear_dirty (vm_page_t, int, int);
344 void vm_page_set_invalid (vm_page_t, int, int);
345 int vm_page_is_valid (vm_page_t, int, int);
346 void vm_page_test_dirty (vm_page_t);
347 int vm_page_bits (int, int);
348 void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid);
349 void vm_page_free_toq(vm_page_t m);
350 void vm_page_zero_idle_wakeup(void);
351 void vm_page_cowfault (vm_page_t);
352 void vm_page_cowsetup (vm_page_t);
353 void vm_page_cowclear (vm_page_t);
354 
355 /*
356  *	vm_page_undirty:
357  *
358  *	Set page to not be dirty.  Note: does not clear pmap modify bits
359  */
360 static __inline void
361 vm_page_undirty(vm_page_t m)
362 {
363 	m->dirty = 0;
364 }
365 
366 #endif				/* _KERNEL */
367 #endif				/* !_VM_PAGE_ */
368