xref: /freebsd/sys/vm/vm_object.h (revision b740c88bfb6453416926271c089262e7164dace3)
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_object.h	8.3 (Berkeley) 1/12/94
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  *	Virtual memory object module definitions.
65  */
66 
67 #ifndef	_VM_OBJECT_
68 #define	_VM_OBJECT_
69 
70 #include <sys/queue.h>
71 #include <sys/_lock.h>
72 #include <sys/_mutex.h>
73 #include <sys/_rwlock.h>
74 
75 #include <vm/_vm_radix.h>
76 
77 /*
78  *	Types defined:
79  *
80  *	vm_object_t		Virtual memory object.
81  *
82  *	The root of cached pages pool is protected by both the per-object lock
83  *	and the free pages queue mutex.
84  *	On insert in the cache radix trie, the per-object lock is expected
85  *	to be already held and the free pages queue mutex will be
86  *	acquired during the operation too.
87  *	On remove and lookup from the cache radix trie, only the free
88  *	pages queue mutex is expected to be locked.
89  *	These rules allow for reliably checking for the presence of cached
90  *	pages with only the per-object lock held, thereby reducing contention
91  *	for the free pages queue mutex.
92  *
93  * List of locks
94  *	(c)	const until freed
95  *	(o)	per-object lock
96  *	(f)	free pages queue mutex
97  *
98  */
99 
100 struct vm_object {
101 	struct rwlock lock;
102 	TAILQ_ENTRY(vm_object) object_list; /* list of all objects */
103 	LIST_HEAD(, vm_object) shadow_head; /* objects that this is a shadow for */
104 	LIST_ENTRY(vm_object) shadow_list; /* chain of shadow objects */
105 	TAILQ_HEAD(respgs, vm_page) memq; /* list of resident pages */
106 	struct vm_radix rtree;		/* root of the resident page radix trie*/
107 	vm_pindex_t size;		/* Object size */
108 	int generation;			/* generation ID */
109 	int ref_count;			/* How many refs?? */
110 	int shadow_count;		/* how many objects that this is a shadow for */
111 	vm_memattr_t memattr;		/* default memory attribute for pages */
112 	objtype_t type;			/* type of pager */
113 	u_short flags;			/* see below */
114 	u_short pg_color;		/* (c) color of first page in obj */
115 	u_int paging_in_progress;	/* Paging (in or out) so don't collapse or destroy */
116 	int resident_page_count;	/* number of resident pages */
117 	struct vm_object *backing_object; /* object that I'm a shadow of */
118 	vm_ooffset_t backing_object_offset;/* Offset in backing object */
119 	TAILQ_ENTRY(vm_object) pager_object_list; /* list of all objects of this pager type */
120 	LIST_HEAD(, vm_reserv) rvq;	/* list of reservations */
121 	struct vm_radix cache;		/* (o + f) root of the cache page radix trie */
122 	void *handle;
123 	union {
124 		/*
125 		 * VNode pager
126 		 *
127 		 *	vnp_size - current size of file
128 		 */
129 		struct {
130 			off_t vnp_size;
131 			vm_ooffset_t writemappings;
132 		} vnp;
133 
134 		/*
135 		 * Device pager
136 		 *
137 		 *	devp_pglist - list of allocated pages
138 		 */
139 		struct {
140 			TAILQ_HEAD(, vm_page) devp_pglist;
141 			struct cdev_pager_ops *ops;
142 			struct cdev *dev;
143 		} devp;
144 
145 		/*
146 		 * SG pager
147 		 *
148 		 *	sgp_pglist - list of allocated pages
149 		 */
150 		struct {
151 			TAILQ_HEAD(, vm_page) sgp_pglist;
152 		} sgp;
153 
154 		/*
155 		 * Swap pager
156 		 *
157 		 *	swp_tmpfs - back-pointer to the tmpfs vnode,
158 		 *		     if any, which uses the vm object
159 		 *		     as backing store.  The handle
160 		 *		     cannot be reused for linking,
161 		 *		     because the vnode can be
162 		 *		     reclaimed and recreated, making
163 		 *		     the handle changed and hash-chain
164 		 *		     invalid.
165 		 *
166 		 *	swp_bcount - number of swap 'swblock' metablocks, each
167 		 *		     contains up to 16 swapblk assignments.
168 		 *		     see vm/swap_pager.h
169 		 */
170 		struct {
171 			void *swp_tmpfs;
172 			int swp_bcount;
173 		} swp;
174 	} un_pager;
175 	struct ucred *cred;
176 	vm_ooffset_t charge;
177 };
178 
179 /*
180  * Flags
181  */
182 #define	OBJ_FICTITIOUS	0x0001		/* (c) contains fictitious pages */
183 #define	OBJ_UNMANAGED	0x0002		/* (c) contains unmanaged pages */
184 #define OBJ_ACTIVE	0x0004		/* active objects */
185 #define OBJ_DEAD	0x0008		/* dead objects (during rundown) */
186 #define	OBJ_NOSPLIT	0x0010		/* dont split this object */
187 #define OBJ_PIPWNT	0x0040		/* paging in progress wanted */
188 #define OBJ_MIGHTBEDIRTY 0x0100		/* object might be dirty, only for vnode */
189 #define	OBJ_TMPFS_NODE	0x0200		/* object belongs to tmpfs VREG node */
190 #define	OBJ_TMPFS_DIRTY	0x0400		/* dirty tmpfs obj */
191 #define	OBJ_COLORED	0x1000		/* pg_color is defined */
192 #define	OBJ_ONEMAPPING	0x2000		/* One USE (a single, non-forked) mapping flag */
193 #define	OBJ_DISCONNECTWNT 0x4000	/* disconnect from vnode wanted */
194 #define	OBJ_TMPFS	0x8000		/* has tmpfs vnode allocated */
195 
196 #define IDX_TO_OFF(idx) (((vm_ooffset_t)(idx)) << PAGE_SHIFT)
197 #define OFF_TO_IDX(off) ((vm_pindex_t)(((vm_ooffset_t)(off)) >> PAGE_SHIFT))
198 
199 #ifdef	_KERNEL
200 
201 #define OBJPC_SYNC	0x1			/* sync I/O */
202 #define OBJPC_INVAL	0x2			/* invalidate */
203 #define OBJPC_NOSYNC	0x4			/* skip if VPO_NOSYNC */
204 
205 /*
206  * The following options are supported by vm_object_page_remove().
207  */
208 #define	OBJPR_CLEANONLY	0x1		/* Don't remove dirty pages. */
209 #define	OBJPR_NOTMAPPED	0x2		/* Don't unmap pages. */
210 #define	OBJPR_NOTWIRED	0x4		/* Don't remove wired pages. */
211 
212 TAILQ_HEAD(object_q, vm_object);
213 
214 extern struct object_q vm_object_list;	/* list of allocated objects */
215 extern struct mtx vm_object_list_mtx;	/* lock for object list and count */
216 
217 extern struct vm_object kernel_object_store;
218 extern struct vm_object kmem_object_store;
219 
220 #define	kernel_object	(&kernel_object_store)
221 #define	kmem_object	(&kmem_object_store)
222 
223 #define	VM_OBJECT_ASSERT_LOCKED(object)					\
224 	rw_assert(&(object)->lock, RA_LOCKED)
225 #define	VM_OBJECT_ASSERT_RLOCKED(object)				\
226 	rw_assert(&(object)->lock, RA_RLOCKED)
227 #define	VM_OBJECT_ASSERT_WLOCKED(object)				\
228 	rw_assert(&(object)->lock, RA_WLOCKED)
229 #define	VM_OBJECT_ASSERT_UNLOCKED(object)				\
230 	rw_assert(&(object)->lock, RA_UNLOCKED)
231 #define	VM_OBJECT_LOCK_DOWNGRADE(object)				\
232 	rw_downgrade(&(object)->lock)
233 #define	VM_OBJECT_RLOCK(object)						\
234 	rw_rlock(&(object)->lock)
235 #define	VM_OBJECT_RUNLOCK(object)					\
236 	rw_runlock(&(object)->lock)
237 #define	VM_OBJECT_SLEEP(object, wchan, pri, wmesg, timo)		\
238 	rw_sleep((wchan), &(object)->lock, (pri), (wmesg), (timo))
239 #define	VM_OBJECT_TRYRLOCK(object)					\
240 	rw_try_rlock(&(object)->lock)
241 #define	VM_OBJECT_TRYWLOCK(object)					\
242 	rw_try_wlock(&(object)->lock)
243 #define	VM_OBJECT_TRYUPGRADE(object)					\
244 	rw_try_upgrade(&(object)->lock)
245 #define	VM_OBJECT_WLOCK(object)						\
246 	rw_wlock(&(object)->lock)
247 #define	VM_OBJECT_WUNLOCK(object)					\
248 	rw_wunlock(&(object)->lock)
249 
250 /*
251  *	The object must be locked or thread private.
252  */
253 static __inline void
254 vm_object_set_flag(vm_object_t object, u_short bits)
255 {
256 
257 	object->flags |= bits;
258 }
259 
260 /*
261  *	Conditionally set the object's color, which (1) enables the allocation
262  *	of physical memory reservations for anonymous objects and larger-than-
263  *	superpage-sized named objects and (2) determines the first page offset
264  *	within the object at which a reservation may be allocated.  In other
265  *	words, the color determines the alignment of the object with respect
266  *	to the largest superpage boundary.  When mapping named objects, like
267  *	files or POSIX shared memory objects, the color should be set to zero
268  *	before a virtual address is selected for the mapping.  In contrast,
269  *	for anonymous objects, the color may be set after the virtual address
270  *	is selected.
271  *
272  *	The object must be locked.
273  */
274 static __inline void
275 vm_object_color(vm_object_t object, u_short color)
276 {
277 
278 	if ((object->flags & OBJ_COLORED) == 0) {
279 		object->pg_color = color;
280 		object->flags |= OBJ_COLORED;
281 	}
282 }
283 
284 void vm_object_clear_flag(vm_object_t object, u_short bits);
285 void vm_object_pip_add(vm_object_t object, short i);
286 void vm_object_pip_subtract(vm_object_t object, short i);
287 void vm_object_pip_wakeup(vm_object_t object);
288 void vm_object_pip_wakeupn(vm_object_t object, short i);
289 void vm_object_pip_wait(vm_object_t object, char *waitid);
290 
291 static __inline boolean_t
292 vm_object_cache_is_empty(vm_object_t object)
293 {
294 
295 	return (vm_radix_is_empty(&object->cache));
296 }
297 
298 vm_object_t vm_object_allocate (objtype_t, vm_pindex_t);
299 boolean_t vm_object_coalesce(vm_object_t, vm_ooffset_t, vm_size_t, vm_size_t,
300    boolean_t);
301 void vm_object_collapse (vm_object_t);
302 void vm_object_deallocate (vm_object_t);
303 void vm_object_destroy (vm_object_t);
304 void vm_object_terminate (vm_object_t);
305 void vm_object_set_writeable_dirty (vm_object_t);
306 void vm_object_init (void);
307 void vm_object_madvise(vm_object_t, vm_pindex_t, vm_pindex_t, int);
308 void vm_object_page_cache(vm_object_t object, vm_pindex_t start,
309     vm_pindex_t end);
310 boolean_t vm_object_page_clean(vm_object_t object, vm_ooffset_t start,
311     vm_ooffset_t end, int flags);
312 void vm_object_page_remove(vm_object_t object, vm_pindex_t start,
313     vm_pindex_t end, int options);
314 boolean_t vm_object_populate(vm_object_t, vm_pindex_t, vm_pindex_t);
315 void vm_object_print(long addr, boolean_t have_addr, long count, char *modif);
316 void vm_object_reference (vm_object_t);
317 void vm_object_reference_locked(vm_object_t);
318 int  vm_object_set_memattr(vm_object_t object, vm_memattr_t memattr);
319 void vm_object_shadow (vm_object_t *, vm_ooffset_t *, vm_size_t);
320 void vm_object_split(vm_map_entry_t);
321 boolean_t vm_object_sync(vm_object_t, vm_ooffset_t, vm_size_t, boolean_t,
322     boolean_t);
323 void vm_object_unwire(vm_object_t object, vm_ooffset_t offset,
324     vm_size_t length, uint8_t queue);
325 #endif				/* _KERNEL */
326 
327 #endif				/* _VM_OBJECT_ */
328