xref: /freebsd/sys/vm/vm_page.h (revision 2ad872c5794e4c26fdf6ed219ad3f09ca0d5304a)
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  * $Id: vm_page.h,v 1.48 1998/10/28 13:37:02 dg Exp $
65  */
66 
67 /*
68  *	Resident memory system definitions.
69  */
70 
71 #ifndef	_VM_PAGE_
72 #define	_VM_PAGE_
73 
74 #include "opt_vmpage.h"
75 
76 #include <vm/pmap.h>
77 #include <machine/atomic.h>
78 
79 /*
80  *	Management of resident (logical) pages.
81  *
82  *	A small structure is kept for each resident
83  *	page, indexed by page number.  Each structure
84  *	is an element of several lists:
85  *
86  *		A hash table bucket used to quickly
87  *		perform object/offset lookups
88  *
89  *		A list of all pages for a given object,
90  *		so they can be quickly deactivated at
91  *		time of deallocation.
92  *
93  *		An ordered list of pages due for pageout.
94  *
95  *	In addition, the structure contains the object
96  *	and offset to which this page belongs (for pageout),
97  *	and sundry status bits.
98  *
99  *	Fields in this structure are locked either by the lock on the
100  *	object that the page belongs to (O) or by the lock on the page
101  *	queues (P).
102  */
103 
104 TAILQ_HEAD(pglist, vm_page);
105 
106 struct vm_page {
107 	TAILQ_ENTRY(vm_page) pageq;	/* queue info for FIFO queue or free list (P) */
108 	TAILQ_ENTRY(vm_page) hashq;	/* hash table links (O) */
109 	TAILQ_ENTRY(vm_page) listq;	/* pages in same object (O) */
110 
111 	vm_object_t object;		/* which object am I in (O,P) */
112 	vm_pindex_t pindex;		/* offset into object (O,P) */
113 	vm_offset_t phys_addr;		/* physical address of page */
114 	u_short	queue;			/* page queue index */
115 	u_short	flags,			/* see below */
116 		pc;			/* page color */
117 	u_short wire_count;		/* wired down maps refs (P) */
118 	short hold_count;		/* page hold count */
119 	u_char	act_count;		/* page usage count */
120 	u_char	busy;			/* page busy count */
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 */
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 */
128 	u_short	dirty;			/* map of dirty DEV_BSIZE chunks */
129 #endif
130 };
131 
132 /*
133  * Page coloring parameters
134  */
135 /* Each of PQ_FREE, PQ_ZERO and PQ_CACHE have PQ_HASH_SIZE entries */
136 
137 /* Define one of the following */
138 #if defined(PQ_HUGECACHE)
139 #define PQ_PRIME1 31	/* Prime number somewhat less than PQ_HASH_SIZE */
140 #define PQ_PRIME2 23	/* Prime number somewhat less than PQ_HASH_SIZE */
141 #define PQ_PRIME3 17	/* Prime number somewhat less than PQ_HASH_SIZE */
142 #define PQ_L2_SIZE 256	/* A number of colors opt for 1M cache */
143 #define PQ_L1_SIZE 4	/* Four page L1 cache */
144 #endif
145 
146 /* Define one of the following */
147 #if defined(PQ_LARGECACHE)
148 #define PQ_PRIME1 31	/* Prime number somewhat less than PQ_HASH_SIZE */
149 #define PQ_PRIME2 23	/* Prime number somewhat less than PQ_HASH_SIZE */
150 #define PQ_PRIME3 17	/* Prime number somewhat less than PQ_HASH_SIZE */
151 #define PQ_L2_SIZE 128	/* A number of colors opt for 512K cache */
152 #define PQ_L1_SIZE 4	/* Four page L1 cache (for PII) */
153 #endif
154 
155 
156 /*
157  * Use 'options PQ_NOOPT' to disable page coloring
158  */
159 #if defined(PQ_NOOPT)
160 #define PQ_PRIME1 1
161 #define PQ_PRIME2 1
162 #define PQ_PRIME3 1
163 #define PQ_L2_SIZE 1
164 #define PQ_L1_SIZE 1
165 #endif
166 
167 #if defined(PQ_NORMALCACHE)
168 #define PQ_PRIME1 5	/* Prime number somewhat less than PQ_HASH_SIZE */
169 #define PQ_PRIME2 3	/* Prime number somewhat less than PQ_HASH_SIZE */
170 #define PQ_PRIME3 11	/* Prime number somewhat less than PQ_HASH_SIZE */
171 #define PQ_L2_SIZE 16	/* A reasonable number of colors (opt for 64K cache) */
172 #define PQ_L1_SIZE 2	/* Two page L1 cache */
173 #endif
174 
175 #if defined(PQ_MEDIUMCACHE) || !defined(PQ_L2_SIZE)
176 #define PQ_PRIME1 13	/* Prime number somewhat less than PQ_HASH_SIZE */
177 #define PQ_PRIME2 7	/* Prime number somewhat less than PQ_HASH_SIZE */
178 #define PQ_PRIME3 5	/* Prime number somewhat less than PQ_HASH_SIZE */
179 #define PQ_L2_SIZE 64	/* A number of colors opt for 256K cache */
180 #define PQ_L1_SIZE 2	/* Two page L1 cache */
181 #endif
182 
183 #define PQ_L2_MASK (PQ_L2_SIZE - 1)
184 
185 #define PQ_NONE 0
186 #define PQ_FREE	1
187 #define PQ_ZERO (1 + PQ_L2_SIZE)
188 #define PQ_INACTIVE (1 + 2*PQ_L2_SIZE)
189 #define PQ_ACTIVE (2 + 2*PQ_L2_SIZE)
190 #define PQ_CACHE (3 + 2*PQ_L2_SIZE)
191 #define PQ_COUNT (3 + 3*PQ_L2_SIZE)
192 
193 extern struct vpgqueues {
194 	struct pglist *pl;
195 	int	*cnt;
196 	int	*lcnt;
197 } vm_page_queues[PQ_COUNT];
198 
199 /*
200  * These are the flags defined for vm_page.
201  *
202  * Note: PG_FILLED and PG_DIRTY are added for the filesystems.
203  */
204 #define	PG_BUSY		0x01		/* page is in transit (O) */
205 #define	PG_WANTED	0x02		/* someone is waiting for page (O) */
206 #define	PG_FICTITIOUS	0x08		/* physical page doesn't exist (O) */
207 #define	PG_WRITEABLE	0x10		/* page is mapped writeable */
208 #define PG_MAPPED	0x20		/* page is mapped */
209 #define	PG_ZERO		0x40		/* page is zeroed */
210 #define PG_REFERENCED	0x80		/* page has been referenced */
211 #define PG_CLEANCHK	0x100		/* page will be checked for cleaning */
212 
213 /*
214  * Misc constants.
215  */
216 
217 #define ACT_DECLINE		1
218 #define ACT_ADVANCE		3
219 #define ACT_INIT		5
220 #define ACT_MAX			64
221 #define PFCLUSTER_BEHIND	3
222 #define PFCLUSTER_AHEAD		3
223 
224 #ifdef KERNEL
225 /*
226  * Each pageable resident page falls into one of four lists:
227  *
228  *	free
229  *		Available for allocation now.
230  *
231  * The following are all LRU sorted:
232  *
233  *	cache
234  *		Almost available for allocation. Still in an
235  *		object, but clean and immediately freeable at
236  *		non-interrupt times.
237  *
238  *	inactive
239  *		Low activity, candidates for reclamation.
240  *		This is the list of pages that should be
241  *		paged out next.
242  *
243  *	active
244  *		Pages that are "active" i.e. they have been
245  *		recently referenced.
246  *
247  *	zero
248  *		Pages that are really free and have been pre-zeroed
249  *
250  */
251 
252 extern struct pglist vm_page_queue_free[PQ_L2_SIZE];/* memory free queue */
253 extern struct pglist vm_page_queue_zero[PQ_L2_SIZE];/* zeroed memory free queue */
254 extern struct pglist vm_page_queue_active;	/* active memory queue */
255 extern struct pglist vm_page_queue_inactive;	/* inactive memory queue */
256 extern struct pglist vm_page_queue_cache[PQ_L2_SIZE];/* cache memory queue */
257 
258 extern int vm_page_zero_count;
259 
260 extern vm_page_t vm_page_array;		/* First resident page in table */
261 extern long first_page;			/* first physical page number */
262 
263  /* ... represented in vm_page_array */
264 extern long last_page;			/* last physical page number */
265 
266  /* ... represented in vm_page_array */
267  /* [INCLUSIVE] */
268 extern vm_offset_t first_phys_addr;	/* physical address for first_page */
269 extern vm_offset_t last_phys_addr;	/* physical address for last_page */
270 
271 #define VM_PAGE_TO_PHYS(entry)	((entry)->phys_addr)
272 
273 #define IS_VM_PHYSADDR(pa) \
274 		((pa) >= first_phys_addr && (pa) <= last_phys_addr)
275 
276 #define PHYS_TO_VM_PAGE(pa) \
277 		(&vm_page_array[atop(pa) - first_page ])
278 
279 /*
280  *	Functions implemented as macros
281  */
282 
283 static __inline void
284 vm_page_flag_set(vm_page_t m, unsigned int bits)
285 {
286 	atomic_set_short(&(m)->flags, bits);
287 }
288 
289 static __inline void
290 vm_page_flag_clear(vm_page_t m, unsigned int bits)
291 {
292 	atomic_clear_short(&(m)->flags, bits);
293 }
294 
295 #if 0
296 static __inline void
297 vm_page_assert_wait(vm_page_t m, int interruptible)
298 {
299 	vm_page_flag_set(m, PG_WANTED);
300 	assert_wait((int) m, interruptible);
301 }
302 #endif
303 
304 static __inline void
305 vm_page_busy(vm_page_t m)
306 {
307 	vm_page_flag_set(m, PG_BUSY);
308 }
309 
310 static __inline void
311 vm_page_wakeup(vm_page_t m)
312 {
313 	vm_page_flag_clear(m, PG_BUSY);
314 	if (m->flags & PG_WANTED) {
315 		vm_page_flag_clear(m, PG_WANTED);
316 		wakeup(m);
317 	}
318 }
319 
320 static __inline void
321 vm_page_io_start(vm_page_t m)
322 {
323 	atomic_add_char(&(m)->busy, 1);
324 }
325 
326 static __inline void
327 vm_page_io_finish(vm_page_t m)
328 {
329 	atomic_subtract_char(&m->busy, 1);
330 	if ((m->flags & PG_WANTED) && m->busy == 0) {
331 		vm_page_flag_clear(m, PG_WANTED);
332 		wakeup(m);
333 	}
334 }
335 
336 
337 #if PAGE_SIZE == 4096
338 #define VM_PAGE_BITS_ALL 0xff
339 #endif
340 
341 #if PAGE_SIZE == 8192
342 #define VM_PAGE_BITS_ALL 0xffff
343 #endif
344 
345 #define VM_ALLOC_NORMAL		0
346 #define VM_ALLOC_INTERRUPT	1
347 #define VM_ALLOC_SYSTEM		2
348 #define	VM_ALLOC_ZERO		3
349 #define	VM_ALLOC_RETRY		0x80
350 
351 void vm_page_activate __P((vm_page_t));
352 vm_page_t vm_page_alloc __P((vm_object_t, vm_pindex_t, int));
353 vm_page_t vm_page_grab __P((vm_object_t, vm_pindex_t, int));
354 void vm_page_cache __P((register vm_page_t));
355 static __inline void vm_page_copy __P((vm_page_t, vm_page_t));
356 void vm_page_deactivate __P((vm_page_t));
357 void vm_page_free __P((vm_page_t));
358 void vm_page_free_zero __P((vm_page_t));
359 void vm_page_insert __P((vm_page_t, vm_object_t, vm_pindex_t));
360 vm_page_t vm_page_lookup __P((vm_object_t, vm_pindex_t));
361 void vm_page_remove __P((vm_page_t));
362 void vm_page_rename __P((vm_page_t, vm_object_t, vm_pindex_t));
363 vm_offset_t vm_page_startup __P((vm_offset_t, vm_offset_t, vm_offset_t));
364 void vm_page_unwire __P((vm_page_t, int));
365 void vm_page_wire __P((vm_page_t));
366 void vm_page_unqueue __P((vm_page_t));
367 void vm_page_unqueue_nowakeup __P((vm_page_t));
368 void vm_page_set_validclean __P((vm_page_t, int, int));
369 void vm_page_set_invalid __P((vm_page_t, int, int));
370 static __inline boolean_t vm_page_zero_fill __P((vm_page_t));
371 int vm_page_is_valid __P((vm_page_t, int, int));
372 void vm_page_test_dirty __P((vm_page_t));
373 int vm_page_bits __P((int, int));
374 vm_page_t vm_page_list_find __P((int, int));
375 int vm_page_queue_index __P((vm_offset_t, int));
376 vm_page_t vm_page_select __P((vm_object_t, vm_pindex_t, int));
377 int vm_page_sleep(vm_page_t m, char *msg, char *busy);
378 
379 /*
380  * Keep page from being freed by the page daemon
381  * much of the same effect as wiring, except much lower
382  * overhead and should be used only for *very* temporary
383  * holding ("wiring").
384  */
385 static __inline void
386 vm_page_hold(vm_page_t mem)
387 {
388 	mem->hold_count++;
389 }
390 
391 static __inline void
392 vm_page_unhold(vm_page_t mem)
393 {
394 	--mem->hold_count;
395 	KASSERT(mem->hold_count >= 0, ("vm_page_unhold: hold count < 0!!!"));
396 }
397 
398 static __inline void
399 vm_page_protect(vm_page_t mem, int prot)
400 {
401 	if (prot == VM_PROT_NONE) {
402 		if (mem->flags & (PG_WRITEABLE|PG_MAPPED)) {
403 			pmap_page_protect(VM_PAGE_TO_PHYS(mem), VM_PROT_NONE);
404 			vm_page_flag_clear(mem, PG_WRITEABLE|PG_MAPPED);
405 		}
406 	} else if ((prot == VM_PROT_READ) && (mem->flags & PG_WRITEABLE)) {
407 		pmap_page_protect(VM_PAGE_TO_PHYS(mem), VM_PROT_READ);
408 		vm_page_flag_clear(mem, PG_WRITEABLE);
409 	}
410 }
411 
412 /*
413  *	vm_page_zero_fill:
414  *
415  *	Zero-fill the specified page.
416  *	Written as a standard pagein routine, to
417  *	be used by the zero-fill object.
418  */
419 static __inline boolean_t
420 vm_page_zero_fill(m)
421 	vm_page_t m;
422 {
423 	pmap_zero_page(VM_PAGE_TO_PHYS(m));
424 	return (TRUE);
425 }
426 
427 /*
428  *	vm_page_copy:
429  *
430  *	Copy one page to another
431  */
432 static __inline void
433 vm_page_copy(src_m, dest_m)
434 	vm_page_t src_m;
435 	vm_page_t dest_m;
436 {
437 	pmap_copy_page(VM_PAGE_TO_PHYS(src_m), VM_PAGE_TO_PHYS(dest_m));
438 	dest_m->valid = VM_PAGE_BITS_ALL;
439 }
440 
441 #endif				/* KERNEL */
442 #endif				/* !_VM_PAGE_ */
443