Deleted Added
full compact
vm_page.h (227103) vm_page.h (227568)
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 *
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: head/sys/vm/vm_page.h 227103 2011-11-05 09:03:18Z kib $
60 * $FreeBSD: head/sys/vm/vm_page.h 227568 2011-11-16 16:46:09Z alc $
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 * In general, operations on this structure's mutable fields are
93 * synchronized using either one of or a combination of the lock on the
94 * object that the page belongs to (O), the pool lock for the page (P),
95 * or the lock for either the free or paging queues (Q). If a field is
96 * annotated below with two of these locks, then holding either lock is
97 * sufficient for read access, but both locks are required for write
98 * access.
99 *
100 * In contrast, the synchronization of accesses to the page's
101 * dirty field is machine dependent (M). In the
102 * machine-independent layer, the lock on the object that the
103 * page belongs to must be held in order to operate on the field.
104 * However, the pmap layer is permitted to set all bits within
105 * the field without holding that lock. If the underlying
106 * architecture does not support atomic read-modify-write
107 * operations on the field's type, then the machine-independent
108 * layer uses a 32-bit atomic on the aligned 32-bit word that
109 * contains the dirty field. In the machine-independent layer,
110 * the implementation of read-modify-write operations on the
111 * field is encapsulated in vm_page_clear_dirty_mask().
112 */
113
114TAILQ_HEAD(pglist, vm_page);
115
116#if PAGE_SIZE == 4096
117#define VM_PAGE_BITS_ALL 0xffu
118typedef uint8_t vm_page_bits_t;
119#elif PAGE_SIZE == 8192
120#define VM_PAGE_BITS_ALL 0xffffu
121typedef uint16_t vm_page_bits_t;
122#elif PAGE_SIZE == 16384
123#define VM_PAGE_BITS_ALL 0xffffffffu
124typedef uint32_t vm_page_bits_t;
125#elif PAGE_SIZE == 32768
126#define VM_PAGE_BITS_ALL 0xfffffffffffffffflu
127typedef uint64_t vm_page_bits_t;
128#endif
129
130struct vm_page {
131 TAILQ_ENTRY(vm_page) pageq; /* queue info for FIFO queue or free list (Q) */
132 TAILQ_ENTRY(vm_page) listq; /* pages in same object (O) */
133 struct vm_page *left; /* splay tree link (O) */
134 struct vm_page *right; /* splay tree link (O) */
135
136 vm_object_t object; /* which object am I in (O,P)*/
137 vm_pindex_t pindex; /* offset into object (O,P) */
138 vm_paddr_t phys_addr; /* physical address of page */
139 struct md_page md; /* machine dependant stuff */
140 uint8_t queue; /* page queue index (P,Q) */
141 int8_t segind;
142 short hold_count; /* page hold count (P) */
143 uint8_t order; /* index of the buddy queue */
144 uint8_t pool;
145 u_short cow; /* page cow mapping count (P) */
146 u_int wire_count; /* wired down maps refs (P) */
147 uint8_t aflags; /* access is atomic */
148 uint8_t flags; /* see below, often immutable after alloc */
149 u_short oflags; /* page flags (O) */
150 u_char act_count; /* page usage count (O) */
151 u_char busy; /* page busy count (O) */
152 /* NOTE that these must support one bit per DEV_BSIZE in a page!!! */
153 /* so, on normal X86 kernels, they must be at least 8 bits wide */
154 vm_page_bits_t valid; /* map of valid DEV_BSIZE chunks (O) */
155 vm_page_bits_t dirty; /* map of dirty DEV_BSIZE chunks (M) */
156};
157
158/*
159 * Page flags stored in oflags:
160 *
161 * Access to these page flags is synchronized by the lock on the object
162 * containing the page (O).
163 *
164 * Note: VPO_UNMANAGED (used by OBJT_DEVICE, OBJT_PHYS and OBJT_SG)
165 * indicates that the page is not under PV management but
166 * otherwise should be treated as a normal page. Pages not
167 * under PV management cannot be paged out via the
168 * object/vm_page_t because there is no knowledge of their pte
169 * mappings, and such pages are also not on any PQ queue.
170 *
171 */
172#define VPO_BUSY 0x0001 /* page is in transit */
173#define VPO_WANTED 0x0002 /* someone is waiting for page */
174#define VPO_UNMANAGED 0x0004 /* No PV management for page */
175#define VPO_SWAPINPROG 0x0200 /* swap I/O in progress on page */
176#define VPO_NOSYNC 0x0400 /* do not collect for syncer */
177
178#define PQ_NONE 255
179#define PQ_INACTIVE 0
180#define PQ_ACTIVE 1
181#define PQ_HOLD 2
182#define PQ_COUNT 3
183
184struct vpgqueues {
185 struct pglist pl;
186 int *cnt;
187};
188
189extern struct vpgqueues vm_page_queues[PQ_COUNT];
190
191struct vpglocks {
192 struct mtx data;
193 char pad[CACHE_LINE_SIZE - sizeof(struct mtx)];
194} __aligned(CACHE_LINE_SIZE);
195
196extern struct vpglocks vm_page_queue_free_lock;
197extern struct vpglocks pa_lock[];
198
199#if defined(__arm__)
200#define PDRSHIFT PDR_SHIFT
201#elif !defined(PDRSHIFT)
202#define PDRSHIFT 21
203#endif
204
205#define pa_index(pa) ((pa) >> PDRSHIFT)
206#define PA_LOCKPTR(pa) &pa_lock[pa_index((pa)) % PA_LOCK_COUNT].data
207#define PA_LOCKOBJPTR(pa) ((struct lock_object *)PA_LOCKPTR((pa)))
208#define PA_LOCK(pa) mtx_lock(PA_LOCKPTR(pa))
209#define PA_TRYLOCK(pa) mtx_trylock(PA_LOCKPTR(pa))
210#define PA_UNLOCK(pa) mtx_unlock(PA_LOCKPTR(pa))
211#define PA_UNLOCK_COND(pa) \
212 do { \
213 if ((pa) != 0) { \
214 PA_UNLOCK((pa)); \
215 (pa) = 0; \
216 } \
217 } while (0)
218
219#define PA_LOCK_ASSERT(pa, a) mtx_assert(PA_LOCKPTR(pa), (a))
220
221#define vm_page_lockptr(m) (PA_LOCKPTR(VM_PAGE_TO_PHYS((m))))
222#define vm_page_lock(m) mtx_lock(vm_page_lockptr((m)))
223#define vm_page_unlock(m) mtx_unlock(vm_page_lockptr((m)))
224#define vm_page_trylock(m) mtx_trylock(vm_page_lockptr((m)))
225#define vm_page_lock_assert(m, a) mtx_assert(vm_page_lockptr((m)), (a))
226
227#define vm_page_queue_free_mtx vm_page_queue_free_lock.data
228/*
229 * These are the flags defined for vm_page.
230 *
231 * aflags are updated by atomic accesses. Use the vm_page_aflag_set()
232 * and vm_page_aflag_clear() functions to set and clear the flags.
233 *
234 * PGA_REFERENCED may be cleared only if the object containing the page is
235 * locked.
236 *
237 * PGA_WRITEABLE is set exclusively on managed pages by pmap_enter(). When it
238 * does so, the page must be VPO_BUSY.
239 */
240#define PGA_WRITEABLE 0x01 /* page may be mapped writeable */
241#define PGA_REFERENCED 0x02 /* page has been referenced */
242
243/*
244 * Page flags. If changed at any other time than page allocation or
245 * freeing, the modification must be protected by the vm_page lock.
246 */
247#define PG_CACHED 0x01 /* page is cached */
248#define PG_FREE 0x02 /* page is free */
249#define PG_FICTITIOUS 0x04 /* physical page doesn't exist (O) */
250#define PG_ZERO 0x08 /* page is zeroed */
251#define PG_MARKER 0x10 /* special queue marker page */
252#define PG_SLAB 0x20 /* object pointer is actually a slab */
253#define PG_WINATCFLS 0x40 /* flush dirty page on inactive q */
254
255/*
256 * Misc constants.
257 */
258#define ACT_DECLINE 1
259#define ACT_ADVANCE 3
260#define ACT_INIT 5
261#define ACT_MAX 64
262
263#ifdef _KERNEL
264
265#include <vm/vm_param.h>
266
267/*
268 * Each pageable resident page falls into one of five lists:
269 *
270 * free
271 * Available for allocation now.
272 *
273 * cache
274 * Almost available for allocation. Still associated with
275 * an object, but clean and immediately freeable.
276 *
277 * hold
278 * Will become free after a pending I/O operation
279 * completes.
280 *
281 * The following lists are LRU sorted:
282 *
283 * inactive
284 * Low activity, candidates for reclamation.
285 * This is the list of pages that should be
286 * paged out next.
287 *
288 * active
289 * Pages that are "active" i.e. they have been
290 * recently referenced.
291 *
292 */
293
294struct vnode;
295extern int vm_page_zero_count;
296
297extern vm_page_t vm_page_array; /* First resident page in table */
298extern int vm_page_array_size; /* number of vm_page_t's */
299extern long first_page; /* first physical page number */
300
301#define VM_PAGE_IS_FREE(m) (((m)->flags & PG_FREE) != 0)
302
303#define VM_PAGE_TO_PHYS(entry) ((entry)->phys_addr)
304
305vm_page_t vm_phys_paddr_to_vm_page(vm_paddr_t pa);
306
307static __inline vm_page_t PHYS_TO_VM_PAGE(vm_paddr_t pa);
308
309static __inline vm_page_t
310PHYS_TO_VM_PAGE(vm_paddr_t pa)
311{
312#ifdef VM_PHYSSEG_SPARSE
313 return (vm_phys_paddr_to_vm_page(pa));
314#elif defined(VM_PHYSSEG_DENSE)
315 return (&vm_page_array[atop(pa) - first_page]);
316#else
317#error "Either VM_PHYSSEG_DENSE or VM_PHYSSEG_SPARSE must be defined."
318#endif
319}
320
321extern struct vpglocks vm_page_queue_lock;
322
323#define vm_page_queue_mtx vm_page_queue_lock.data
324#define vm_page_lock_queues() mtx_lock(&vm_page_queue_mtx)
325#define vm_page_unlock_queues() mtx_unlock(&vm_page_queue_mtx)
326
327/* page allocation classes: */
328#define VM_ALLOC_NORMAL 0
329#define VM_ALLOC_INTERRUPT 1
330#define VM_ALLOC_SYSTEM 2
331#define VM_ALLOC_CLASS_MASK 3
332/* page allocation flags: */
333#define VM_ALLOC_WIRED 0x0020 /* non pageable */
334#define VM_ALLOC_ZERO 0x0040 /* Try to obtain a zeroed page */
335#define VM_ALLOC_RETRY 0x0080 /* Mandatory with vm_page_grab() */
336#define VM_ALLOC_NOOBJ 0x0100 /* No associated object */
337#define VM_ALLOC_NOBUSY 0x0200 /* Do not busy the page */
338#define VM_ALLOC_IFCACHED 0x0400 /* Fail if the page is not cached */
339#define VM_ALLOC_IFNOTCACHED 0x0800 /* Fail if the page is cached */
340#define VM_ALLOC_IGN_SBUSY 0x1000 /* vm_page_grab() only */
341
342#define VM_ALLOC_COUNT_SHIFT 16
343#define VM_ALLOC_COUNT(count) ((count) << VM_ALLOC_COUNT_SHIFT)
344
345void vm_page_aflag_set(vm_page_t m, uint8_t bits);
346void vm_page_aflag_clear(vm_page_t m, uint8_t bits);
347void vm_page_busy(vm_page_t m);
348void vm_page_flash(vm_page_t m);
349void vm_page_io_start(vm_page_t m);
350void vm_page_io_finish(vm_page_t m);
351void vm_page_hold(vm_page_t mem);
352void vm_page_unhold(vm_page_t mem);
353void vm_page_free(vm_page_t m);
354void vm_page_free_zero(vm_page_t m);
355void vm_page_dirty(vm_page_t m);
356void vm_page_wakeup(vm_page_t m);
357
358void vm_pageq_remove(vm_page_t m);
359
360void vm_page_activate (vm_page_t);
361vm_page_t vm_page_alloc (vm_object_t, vm_pindex_t, int);
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 * In general, operations on this structure's mutable fields are
93 * synchronized using either one of or a combination of the lock on the
94 * object that the page belongs to (O), the pool lock for the page (P),
95 * or the lock for either the free or paging queues (Q). If a field is
96 * annotated below with two of these locks, then holding either lock is
97 * sufficient for read access, but both locks are required for write
98 * access.
99 *
100 * In contrast, the synchronization of accesses to the page's
101 * dirty field is machine dependent (M). In the
102 * machine-independent layer, the lock on the object that the
103 * page belongs to must be held in order to operate on the field.
104 * However, the pmap layer is permitted to set all bits within
105 * the field without holding that lock. If the underlying
106 * architecture does not support atomic read-modify-write
107 * operations on the field's type, then the machine-independent
108 * layer uses a 32-bit atomic on the aligned 32-bit word that
109 * contains the dirty field. In the machine-independent layer,
110 * the implementation of read-modify-write operations on the
111 * field is encapsulated in vm_page_clear_dirty_mask().
112 */
113
114TAILQ_HEAD(pglist, vm_page);
115
116#if PAGE_SIZE == 4096
117#define VM_PAGE_BITS_ALL 0xffu
118typedef uint8_t vm_page_bits_t;
119#elif PAGE_SIZE == 8192
120#define VM_PAGE_BITS_ALL 0xffffu
121typedef uint16_t vm_page_bits_t;
122#elif PAGE_SIZE == 16384
123#define VM_PAGE_BITS_ALL 0xffffffffu
124typedef uint32_t vm_page_bits_t;
125#elif PAGE_SIZE == 32768
126#define VM_PAGE_BITS_ALL 0xfffffffffffffffflu
127typedef uint64_t vm_page_bits_t;
128#endif
129
130struct vm_page {
131 TAILQ_ENTRY(vm_page) pageq; /* queue info for FIFO queue or free list (Q) */
132 TAILQ_ENTRY(vm_page) listq; /* pages in same object (O) */
133 struct vm_page *left; /* splay tree link (O) */
134 struct vm_page *right; /* splay tree link (O) */
135
136 vm_object_t object; /* which object am I in (O,P)*/
137 vm_pindex_t pindex; /* offset into object (O,P) */
138 vm_paddr_t phys_addr; /* physical address of page */
139 struct md_page md; /* machine dependant stuff */
140 uint8_t queue; /* page queue index (P,Q) */
141 int8_t segind;
142 short hold_count; /* page hold count (P) */
143 uint8_t order; /* index of the buddy queue */
144 uint8_t pool;
145 u_short cow; /* page cow mapping count (P) */
146 u_int wire_count; /* wired down maps refs (P) */
147 uint8_t aflags; /* access is atomic */
148 uint8_t flags; /* see below, often immutable after alloc */
149 u_short oflags; /* page flags (O) */
150 u_char act_count; /* page usage count (O) */
151 u_char busy; /* page busy count (O) */
152 /* NOTE that these must support one bit per DEV_BSIZE in a page!!! */
153 /* so, on normal X86 kernels, they must be at least 8 bits wide */
154 vm_page_bits_t valid; /* map of valid DEV_BSIZE chunks (O) */
155 vm_page_bits_t dirty; /* map of dirty DEV_BSIZE chunks (M) */
156};
157
158/*
159 * Page flags stored in oflags:
160 *
161 * Access to these page flags is synchronized by the lock on the object
162 * containing the page (O).
163 *
164 * Note: VPO_UNMANAGED (used by OBJT_DEVICE, OBJT_PHYS and OBJT_SG)
165 * indicates that the page is not under PV management but
166 * otherwise should be treated as a normal page. Pages not
167 * under PV management cannot be paged out via the
168 * object/vm_page_t because there is no knowledge of their pte
169 * mappings, and such pages are also not on any PQ queue.
170 *
171 */
172#define VPO_BUSY 0x0001 /* page is in transit */
173#define VPO_WANTED 0x0002 /* someone is waiting for page */
174#define VPO_UNMANAGED 0x0004 /* No PV management for page */
175#define VPO_SWAPINPROG 0x0200 /* swap I/O in progress on page */
176#define VPO_NOSYNC 0x0400 /* do not collect for syncer */
177
178#define PQ_NONE 255
179#define PQ_INACTIVE 0
180#define PQ_ACTIVE 1
181#define PQ_HOLD 2
182#define PQ_COUNT 3
183
184struct vpgqueues {
185 struct pglist pl;
186 int *cnt;
187};
188
189extern struct vpgqueues vm_page_queues[PQ_COUNT];
190
191struct vpglocks {
192 struct mtx data;
193 char pad[CACHE_LINE_SIZE - sizeof(struct mtx)];
194} __aligned(CACHE_LINE_SIZE);
195
196extern struct vpglocks vm_page_queue_free_lock;
197extern struct vpglocks pa_lock[];
198
199#if defined(__arm__)
200#define PDRSHIFT PDR_SHIFT
201#elif !defined(PDRSHIFT)
202#define PDRSHIFT 21
203#endif
204
205#define pa_index(pa) ((pa) >> PDRSHIFT)
206#define PA_LOCKPTR(pa) &pa_lock[pa_index((pa)) % PA_LOCK_COUNT].data
207#define PA_LOCKOBJPTR(pa) ((struct lock_object *)PA_LOCKPTR((pa)))
208#define PA_LOCK(pa) mtx_lock(PA_LOCKPTR(pa))
209#define PA_TRYLOCK(pa) mtx_trylock(PA_LOCKPTR(pa))
210#define PA_UNLOCK(pa) mtx_unlock(PA_LOCKPTR(pa))
211#define PA_UNLOCK_COND(pa) \
212 do { \
213 if ((pa) != 0) { \
214 PA_UNLOCK((pa)); \
215 (pa) = 0; \
216 } \
217 } while (0)
218
219#define PA_LOCK_ASSERT(pa, a) mtx_assert(PA_LOCKPTR(pa), (a))
220
221#define vm_page_lockptr(m) (PA_LOCKPTR(VM_PAGE_TO_PHYS((m))))
222#define vm_page_lock(m) mtx_lock(vm_page_lockptr((m)))
223#define vm_page_unlock(m) mtx_unlock(vm_page_lockptr((m)))
224#define vm_page_trylock(m) mtx_trylock(vm_page_lockptr((m)))
225#define vm_page_lock_assert(m, a) mtx_assert(vm_page_lockptr((m)), (a))
226
227#define vm_page_queue_free_mtx vm_page_queue_free_lock.data
228/*
229 * These are the flags defined for vm_page.
230 *
231 * aflags are updated by atomic accesses. Use the vm_page_aflag_set()
232 * and vm_page_aflag_clear() functions to set and clear the flags.
233 *
234 * PGA_REFERENCED may be cleared only if the object containing the page is
235 * locked.
236 *
237 * PGA_WRITEABLE is set exclusively on managed pages by pmap_enter(). When it
238 * does so, the page must be VPO_BUSY.
239 */
240#define PGA_WRITEABLE 0x01 /* page may be mapped writeable */
241#define PGA_REFERENCED 0x02 /* page has been referenced */
242
243/*
244 * Page flags. If changed at any other time than page allocation or
245 * freeing, the modification must be protected by the vm_page lock.
246 */
247#define PG_CACHED 0x01 /* page is cached */
248#define PG_FREE 0x02 /* page is free */
249#define PG_FICTITIOUS 0x04 /* physical page doesn't exist (O) */
250#define PG_ZERO 0x08 /* page is zeroed */
251#define PG_MARKER 0x10 /* special queue marker page */
252#define PG_SLAB 0x20 /* object pointer is actually a slab */
253#define PG_WINATCFLS 0x40 /* flush dirty page on inactive q */
254
255/*
256 * Misc constants.
257 */
258#define ACT_DECLINE 1
259#define ACT_ADVANCE 3
260#define ACT_INIT 5
261#define ACT_MAX 64
262
263#ifdef _KERNEL
264
265#include <vm/vm_param.h>
266
267/*
268 * Each pageable resident page falls into one of five lists:
269 *
270 * free
271 * Available for allocation now.
272 *
273 * cache
274 * Almost available for allocation. Still associated with
275 * an object, but clean and immediately freeable.
276 *
277 * hold
278 * Will become free after a pending I/O operation
279 * completes.
280 *
281 * The following lists are LRU sorted:
282 *
283 * inactive
284 * Low activity, candidates for reclamation.
285 * This is the list of pages that should be
286 * paged out next.
287 *
288 * active
289 * Pages that are "active" i.e. they have been
290 * recently referenced.
291 *
292 */
293
294struct vnode;
295extern int vm_page_zero_count;
296
297extern vm_page_t vm_page_array; /* First resident page in table */
298extern int vm_page_array_size; /* number of vm_page_t's */
299extern long first_page; /* first physical page number */
300
301#define VM_PAGE_IS_FREE(m) (((m)->flags & PG_FREE) != 0)
302
303#define VM_PAGE_TO_PHYS(entry) ((entry)->phys_addr)
304
305vm_page_t vm_phys_paddr_to_vm_page(vm_paddr_t pa);
306
307static __inline vm_page_t PHYS_TO_VM_PAGE(vm_paddr_t pa);
308
309static __inline vm_page_t
310PHYS_TO_VM_PAGE(vm_paddr_t pa)
311{
312#ifdef VM_PHYSSEG_SPARSE
313 return (vm_phys_paddr_to_vm_page(pa));
314#elif defined(VM_PHYSSEG_DENSE)
315 return (&vm_page_array[atop(pa) - first_page]);
316#else
317#error "Either VM_PHYSSEG_DENSE or VM_PHYSSEG_SPARSE must be defined."
318#endif
319}
320
321extern struct vpglocks vm_page_queue_lock;
322
323#define vm_page_queue_mtx vm_page_queue_lock.data
324#define vm_page_lock_queues() mtx_lock(&vm_page_queue_mtx)
325#define vm_page_unlock_queues() mtx_unlock(&vm_page_queue_mtx)
326
327/* page allocation classes: */
328#define VM_ALLOC_NORMAL 0
329#define VM_ALLOC_INTERRUPT 1
330#define VM_ALLOC_SYSTEM 2
331#define VM_ALLOC_CLASS_MASK 3
332/* page allocation flags: */
333#define VM_ALLOC_WIRED 0x0020 /* non pageable */
334#define VM_ALLOC_ZERO 0x0040 /* Try to obtain a zeroed page */
335#define VM_ALLOC_RETRY 0x0080 /* Mandatory with vm_page_grab() */
336#define VM_ALLOC_NOOBJ 0x0100 /* No associated object */
337#define VM_ALLOC_NOBUSY 0x0200 /* Do not busy the page */
338#define VM_ALLOC_IFCACHED 0x0400 /* Fail if the page is not cached */
339#define VM_ALLOC_IFNOTCACHED 0x0800 /* Fail if the page is cached */
340#define VM_ALLOC_IGN_SBUSY 0x1000 /* vm_page_grab() only */
341
342#define VM_ALLOC_COUNT_SHIFT 16
343#define VM_ALLOC_COUNT(count) ((count) << VM_ALLOC_COUNT_SHIFT)
344
345void vm_page_aflag_set(vm_page_t m, uint8_t bits);
346void vm_page_aflag_clear(vm_page_t m, uint8_t bits);
347void vm_page_busy(vm_page_t m);
348void vm_page_flash(vm_page_t m);
349void vm_page_io_start(vm_page_t m);
350void vm_page_io_finish(vm_page_t m);
351void vm_page_hold(vm_page_t mem);
352void vm_page_unhold(vm_page_t mem);
353void vm_page_free(vm_page_t m);
354void vm_page_free_zero(vm_page_t m);
355void vm_page_dirty(vm_page_t m);
356void vm_page_wakeup(vm_page_t m);
357
358void vm_pageq_remove(vm_page_t m);
359
360void vm_page_activate (vm_page_t);
361vm_page_t vm_page_alloc (vm_object_t, vm_pindex_t, int);
362vm_page_t vm_page_alloc_contig(vm_object_t object, vm_pindex_t pindex, int req,
363 u_long npages, vm_paddr_t low, vm_paddr_t high, u_long alignment,
364 vm_paddr_t boundary, vm_memattr_t memattr);
362vm_page_t vm_page_alloc_freelist(int, int);
365vm_page_t vm_page_alloc_freelist(int, int);
363struct vnode *vm_page_alloc_init(vm_page_t);
364vm_page_t vm_page_grab (vm_object_t, vm_pindex_t, int);
365void vm_page_cache(vm_page_t);
366void vm_page_cache_free(vm_object_t, vm_pindex_t, vm_pindex_t);
367void vm_page_cache_remove(vm_page_t);
368void vm_page_cache_transfer(vm_object_t, vm_pindex_t, vm_object_t);
369int vm_page_try_to_cache (vm_page_t);
370int vm_page_try_to_free (vm_page_t);
371void vm_page_dontneed(vm_page_t);
372void vm_page_deactivate (vm_page_t);
373vm_page_t vm_page_find_least(vm_object_t, vm_pindex_t);
374vm_page_t vm_page_getfake(vm_paddr_t paddr, vm_memattr_t memattr);
375void vm_page_insert (vm_page_t, vm_object_t, vm_pindex_t);
376vm_page_t vm_page_lookup (vm_object_t, vm_pindex_t);
377vm_page_t vm_page_next(vm_page_t m);
378int vm_page_pa_tryrelock(pmap_t, vm_paddr_t, vm_paddr_t *);
379vm_page_t vm_page_prev(vm_page_t m);
380void vm_page_putfake(vm_page_t m);
381void vm_page_reference(vm_page_t m);
382void vm_page_remove (vm_page_t);
383void vm_page_rename (vm_page_t, vm_object_t, vm_pindex_t);
384void vm_page_requeue(vm_page_t m);
385void vm_page_set_valid(vm_page_t m, int base, int size);
386void vm_page_sleep(vm_page_t m, const char *msg);
387vm_page_t vm_page_splay(vm_pindex_t, vm_page_t);
388vm_offset_t vm_page_startup(vm_offset_t vaddr);
389void vm_page_unhold_pages(vm_page_t *ma, int count);
390void vm_page_unwire (vm_page_t, int);
391void vm_page_updatefake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr);
392void vm_page_wire (vm_page_t);
393void vm_page_set_validclean (vm_page_t, int, int);
394void vm_page_clear_dirty (vm_page_t, int, int);
395void vm_page_set_invalid (vm_page_t, int, int);
396int vm_page_is_valid (vm_page_t, int, int);
397void vm_page_test_dirty (vm_page_t);
398vm_page_bits_t vm_page_bits(int base, int size);
399void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid);
400void vm_page_free_toq(vm_page_t m);
401void vm_page_zero_idle_wakeup(void);
402void vm_page_cowfault (vm_page_t);
403int vm_page_cowsetup(vm_page_t);
404void vm_page_cowclear (vm_page_t);
405
406#ifdef INVARIANTS
407void vm_page_object_lock_assert(vm_page_t m);
408#define VM_PAGE_OBJECT_LOCK_ASSERT(m) vm_page_object_lock_assert(m)
409#else
410#define VM_PAGE_OBJECT_LOCK_ASSERT(m) (void)0
411#endif
412
413/*
414 * vm_page_sleep_if_busy:
415 *
416 * Sleep and release the page queues lock if VPO_BUSY is set or,
417 * if also_m_busy is TRUE, busy is non-zero. Returns TRUE if the
418 * thread slept and the page queues lock was released.
419 * Otherwise, retains the page queues lock and returns FALSE.
420 *
421 * The object containing the given page must be locked.
422 */
423static __inline int
424vm_page_sleep_if_busy(vm_page_t m, int also_m_busy, const char *msg)
425{
426
427 if ((m->oflags & VPO_BUSY) || (also_m_busy && m->busy)) {
428 vm_page_sleep(m, msg);
429 return (TRUE);
430 }
431 return (FALSE);
432}
433
434/*
435 * vm_page_undirty:
436 *
437 * Set page to not be dirty. Note: does not clear pmap modify bits
438 */
439static __inline void
440vm_page_undirty(vm_page_t m)
441{
442
443 VM_PAGE_OBJECT_LOCK_ASSERT(m);
444 m->dirty = 0;
445}
446
447#endif /* _KERNEL */
448#endif /* !_VM_PAGE_ */
366vm_page_t vm_page_grab (vm_object_t, vm_pindex_t, int);
367void vm_page_cache(vm_page_t);
368void vm_page_cache_free(vm_object_t, vm_pindex_t, vm_pindex_t);
369void vm_page_cache_remove(vm_page_t);
370void vm_page_cache_transfer(vm_object_t, vm_pindex_t, vm_object_t);
371int vm_page_try_to_cache (vm_page_t);
372int vm_page_try_to_free (vm_page_t);
373void vm_page_dontneed(vm_page_t);
374void vm_page_deactivate (vm_page_t);
375vm_page_t vm_page_find_least(vm_object_t, vm_pindex_t);
376vm_page_t vm_page_getfake(vm_paddr_t paddr, vm_memattr_t memattr);
377void vm_page_insert (vm_page_t, vm_object_t, vm_pindex_t);
378vm_page_t vm_page_lookup (vm_object_t, vm_pindex_t);
379vm_page_t vm_page_next(vm_page_t m);
380int vm_page_pa_tryrelock(pmap_t, vm_paddr_t, vm_paddr_t *);
381vm_page_t vm_page_prev(vm_page_t m);
382void vm_page_putfake(vm_page_t m);
383void vm_page_reference(vm_page_t m);
384void vm_page_remove (vm_page_t);
385void vm_page_rename (vm_page_t, vm_object_t, vm_pindex_t);
386void vm_page_requeue(vm_page_t m);
387void vm_page_set_valid(vm_page_t m, int base, int size);
388void vm_page_sleep(vm_page_t m, const char *msg);
389vm_page_t vm_page_splay(vm_pindex_t, vm_page_t);
390vm_offset_t vm_page_startup(vm_offset_t vaddr);
391void vm_page_unhold_pages(vm_page_t *ma, int count);
392void vm_page_unwire (vm_page_t, int);
393void vm_page_updatefake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr);
394void vm_page_wire (vm_page_t);
395void vm_page_set_validclean (vm_page_t, int, int);
396void vm_page_clear_dirty (vm_page_t, int, int);
397void vm_page_set_invalid (vm_page_t, int, int);
398int vm_page_is_valid (vm_page_t, int, int);
399void vm_page_test_dirty (vm_page_t);
400vm_page_bits_t vm_page_bits(int base, int size);
401void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid);
402void vm_page_free_toq(vm_page_t m);
403void vm_page_zero_idle_wakeup(void);
404void vm_page_cowfault (vm_page_t);
405int vm_page_cowsetup(vm_page_t);
406void vm_page_cowclear (vm_page_t);
407
408#ifdef INVARIANTS
409void vm_page_object_lock_assert(vm_page_t m);
410#define VM_PAGE_OBJECT_LOCK_ASSERT(m) vm_page_object_lock_assert(m)
411#else
412#define VM_PAGE_OBJECT_LOCK_ASSERT(m) (void)0
413#endif
414
415/*
416 * vm_page_sleep_if_busy:
417 *
418 * Sleep and release the page queues lock if VPO_BUSY is set or,
419 * if also_m_busy is TRUE, busy is non-zero. Returns TRUE if the
420 * thread slept and the page queues lock was released.
421 * Otherwise, retains the page queues lock and returns FALSE.
422 *
423 * The object containing the given page must be locked.
424 */
425static __inline int
426vm_page_sleep_if_busy(vm_page_t m, int also_m_busy, const char *msg)
427{
428
429 if ((m->oflags & VPO_BUSY) || (also_m_busy && m->busy)) {
430 vm_page_sleep(m, msg);
431 return (TRUE);
432 }
433 return (FALSE);
434}
435
436/*
437 * vm_page_undirty:
438 *
439 * Set page to not be dirty. Note: does not clear pmap modify bits
440 */
441static __inline void
442vm_page_undirty(vm_page_t m)
443{
444
445 VM_PAGE_OBJECT_LOCK_ASSERT(m);
446 m->dirty = 0;
447}
448
449#endif /* _KERNEL */
450#endif /* !_VM_PAGE_ */