xref: /freebsd/sys/vm/vm_object.h (revision f39bffc62c1395bde25d152c7f68fdf7cbaab414)
1 /*-
2  * SPDX-License-Identifier: (BSD-3-Clause AND MIT-CMU)
3  *
4  * Copyright (c) 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * The Mach Operating System project at Carnegie-Mellon University.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	from: @(#)vm_object.h	8.3 (Berkeley) 1/12/94
35  *
36  *
37  * Copyright (c) 1987, 1990 Carnegie-Mellon University.
38  * All rights reserved.
39  *
40  * Authors: Avadis Tevanian, Jr., Michael Wayne Young
41  *
42  * Permission to use, copy, modify and distribute this software and
43  * its documentation is hereby granted, provided that both the copyright
44  * notice and this permission notice appear in all copies of the
45  * software, derivative works or modified versions, and any portions
46  * thereof, and that both notices appear in supporting documentation.
47  *
48  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
49  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
50  * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
51  *
52  * Carnegie Mellon requests users of this software to return to
53  *
54  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
55  *  School of Computer Science
56  *  Carnegie Mellon University
57  *  Pittsburgh PA 15213-3890
58  *
59  * any improvements or extensions that they make and grant Carnegie the
60  * rights to redistribute these changes.
61  *
62  * $FreeBSD$
63  */
64 
65 /*
66  *	Virtual memory object module definitions.
67  */
68 
69 #ifndef	_VM_OBJECT_
70 #define	_VM_OBJECT_
71 
72 #include <sys/queue.h>
73 #include <sys/_lock.h>
74 #include <sys/_mutex.h>
75 #include <sys/_pctrie.h>
76 #include <sys/_rwlock.h>
77 #include <sys/_domainset.h>
78 
79 #include <vm/_vm_radix.h>
80 
81 /*
82  *	Types defined:
83  *
84  *	vm_object_t		Virtual memory object.
85  *
86  * List of locks
87  *	(c)	const until freed
88  *	(o)	per-object lock
89  *	(f)	free pages queue mutex
90  *
91  */
92 
93 #ifndef VM_PAGE_HAVE_PGLIST
94 TAILQ_HEAD(pglist, vm_page);
95 #define VM_PAGE_HAVE_PGLIST
96 #endif
97 
98 struct vm_object {
99 	struct rwlock lock;
100 	TAILQ_ENTRY(vm_object) object_list; /* list of all objects */
101 	LIST_HEAD(, vm_object) shadow_head; /* objects that this is a shadow for */
102 	LIST_ENTRY(vm_object) shadow_list; /* chain of shadow objects */
103 	struct pglist memq;		/* list of resident pages */
104 	struct vm_radix rtree;		/* root of the resident page radix trie*/
105 	vm_pindex_t size;		/* Object size */
106 	struct domainset_ref domain;	/* NUMA policy. */
107 	int generation;			/* generation ID */
108 	int ref_count;			/* How many refs?? */
109 	int shadow_count;		/* how many objects that this is a shadow for */
110 	vm_memattr_t memattr;		/* default memory attribute for pages */
111 	objtype_t type;			/* type of pager */
112 	u_short flags;			/* see below */
113 	u_short pg_color;		/* (c) color of first page in obj */
114 	u_int paging_in_progress;	/* Paging (in or out) so don't collapse or destroy */
115 	int resident_page_count;	/* number of resident pages */
116 	struct vm_object *backing_object; /* object that I'm a shadow of */
117 	vm_ooffset_t backing_object_offset;/* Offset in backing object */
118 	TAILQ_ENTRY(vm_object) pager_object_list; /* list of all objects of this pager type */
119 	LIST_HEAD(, vm_reserv) rvq;	/* list of reservations */
120 	void *handle;
121 	union {
122 		/*
123 		 * VNode pager
124 		 *
125 		 *	vnp_size - current size of file
126 		 */
127 		struct {
128 			off_t vnp_size;
129 			vm_ooffset_t writemappings;
130 		} vnp;
131 
132 		/*
133 		 * Device pager
134 		 *
135 		 *	devp_pglist - list of allocated pages
136 		 */
137 		struct {
138 			TAILQ_HEAD(, vm_page) devp_pglist;
139 			struct cdev_pager_ops *ops;
140 			struct cdev *dev;
141 		} devp;
142 
143 		/*
144 		 * SG pager
145 		 *
146 		 *	sgp_pglist - list of allocated pages
147 		 */
148 		struct {
149 			TAILQ_HEAD(, vm_page) sgp_pglist;
150 		} sgp;
151 
152 		/*
153 		 * Swap pager
154 		 *
155 		 *	swp_tmpfs - back-pointer to the tmpfs vnode,
156 		 *		     if any, which uses the vm object
157 		 *		     as backing store.  The handle
158 		 *		     cannot be reused for linking,
159 		 *		     because the vnode can be
160 		 *		     reclaimed and recreated, making
161 		 *		     the handle changed and hash-chain
162 		 *		     invalid.
163 		 *
164 		 *	swp_blks -   pc-trie of the allocated swap blocks.
165 		 *
166 		 */
167 		struct {
168 			void *swp_tmpfs;
169 			struct pctrie swp_blks;
170 		} swp;
171 	} un_pager;
172 	struct ucred *cred;
173 	vm_ooffset_t charge;
174 	void *umtx_data;
175 };
176 
177 /*
178  * Flags
179  */
180 #define	OBJ_FICTITIOUS	0x0001		/* (c) contains fictitious pages */
181 #define	OBJ_UNMANAGED	0x0002		/* (c) contains unmanaged pages */
182 #define	OBJ_POPULATE	0x0004		/* pager implements populate() */
183 #define	OBJ_DEAD	0x0008		/* dead objects (during rundown) */
184 #define	OBJ_NOSPLIT	0x0010		/* dont split this object */
185 #define	OBJ_UMTXDEAD	0x0020		/* umtx pshared was terminated */
186 #define	OBJ_PIPWNT	0x0040		/* paging in progress wanted */
187 #define	OBJ_PG_DTOR	0x0080		/* dont reset object, leave that for dtor */
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 /*
197  * Helpers to perform conversion between vm_object page indexes and offsets.
198  * IDX_TO_OFF() converts an index into an offset.
199  * OFF_TO_IDX() converts an offset into an index.  Since offsets are signed
200  *   by default, the sign propagation in OFF_TO_IDX(), when applied to
201  *   negative offsets, is intentional and returns a vm_object page index
202  *   that cannot be created by a userspace mapping.
203  * UOFF_TO_IDX() treats the offset as an unsigned value and converts it
204  *   into an index accordingly.  Use it only when the full range of offset
205  *   values are allowed.  Currently, this only applies to device mappings.
206  * OBJ_MAX_SIZE specifies the maximum page index corresponding to the
207  *   maximum unsigned offset.
208  */
209 #define	IDX_TO_OFF(idx) (((vm_ooffset_t)(idx)) << PAGE_SHIFT)
210 #define	OFF_TO_IDX(off) ((vm_pindex_t)(((vm_ooffset_t)(off)) >> PAGE_SHIFT))
211 #define	UOFF_TO_IDX(off) (((vm_pindex_t)(off)) >> PAGE_SHIFT)
212 #define	OBJ_MAX_SIZE	(UOFF_TO_IDX(UINT64_MAX) + 1)
213 
214 #ifdef	_KERNEL
215 
216 #define OBJPC_SYNC	0x1			/* sync I/O */
217 #define OBJPC_INVAL	0x2			/* invalidate */
218 #define OBJPC_NOSYNC	0x4			/* skip if VPO_NOSYNC */
219 
220 /*
221  * The following options are supported by vm_object_page_remove().
222  */
223 #define	OBJPR_CLEANONLY	0x1		/* Don't remove dirty pages. */
224 #define	OBJPR_NOTMAPPED	0x2		/* Don't unmap pages. */
225 
226 TAILQ_HEAD(object_q, vm_object);
227 
228 extern struct object_q vm_object_list;	/* list of allocated objects */
229 extern struct mtx vm_object_list_mtx;	/* lock for object list and count */
230 
231 extern struct vm_object kernel_object_store;
232 
233 /* kernel and kmem are aliased for backwards KPI compat. */
234 #define	kernel_object	(&kernel_object_store)
235 #define	kmem_object	(&kernel_object_store)
236 
237 #define	VM_OBJECT_ASSERT_LOCKED(object)					\
238 	rw_assert(&(object)->lock, RA_LOCKED)
239 #define	VM_OBJECT_ASSERT_RLOCKED(object)				\
240 	rw_assert(&(object)->lock, RA_RLOCKED)
241 #define	VM_OBJECT_ASSERT_WLOCKED(object)				\
242 	rw_assert(&(object)->lock, RA_WLOCKED)
243 #define	VM_OBJECT_ASSERT_UNLOCKED(object)				\
244 	rw_assert(&(object)->lock, RA_UNLOCKED)
245 #define	VM_OBJECT_LOCK_DOWNGRADE(object)				\
246 	rw_downgrade(&(object)->lock)
247 #define	VM_OBJECT_RLOCK(object)						\
248 	rw_rlock(&(object)->lock)
249 #define	VM_OBJECT_RUNLOCK(object)					\
250 	rw_runlock(&(object)->lock)
251 #define	VM_OBJECT_SLEEP(object, wchan, pri, wmesg, timo)		\
252 	rw_sleep((wchan), &(object)->lock, (pri), (wmesg), (timo))
253 #define	VM_OBJECT_TRYRLOCK(object)					\
254 	rw_try_rlock(&(object)->lock)
255 #define	VM_OBJECT_TRYWLOCK(object)					\
256 	rw_try_wlock(&(object)->lock)
257 #define	VM_OBJECT_TRYUPGRADE(object)					\
258 	rw_try_upgrade(&(object)->lock)
259 #define	VM_OBJECT_WLOCK(object)						\
260 	rw_wlock(&(object)->lock)
261 #define	VM_OBJECT_WOWNED(object)					\
262 	rw_wowned(&(object)->lock)
263 #define	VM_OBJECT_WUNLOCK(object)					\
264 	rw_wunlock(&(object)->lock)
265 
266 /*
267  *	The object must be locked or thread private.
268  */
269 static __inline void
270 vm_object_set_flag(vm_object_t object, u_short bits)
271 {
272 
273 	object->flags |= bits;
274 }
275 
276 /*
277  *	Conditionally set the object's color, which (1) enables the allocation
278  *	of physical memory reservations for anonymous objects and larger-than-
279  *	superpage-sized named objects and (2) determines the first page offset
280  *	within the object at which a reservation may be allocated.  In other
281  *	words, the color determines the alignment of the object with respect
282  *	to the largest superpage boundary.  When mapping named objects, like
283  *	files or POSIX shared memory objects, the color should be set to zero
284  *	before a virtual address is selected for the mapping.  In contrast,
285  *	for anonymous objects, the color may be set after the virtual address
286  *	is selected.
287  *
288  *	The object must be locked.
289  */
290 static __inline void
291 vm_object_color(vm_object_t object, u_short color)
292 {
293 
294 	if ((object->flags & OBJ_COLORED) == 0) {
295 		object->pg_color = color;
296 		object->flags |= OBJ_COLORED;
297 	}
298 }
299 
300 static __inline bool
301 vm_object_reserv(vm_object_t object)
302 {
303 
304 	if (object != NULL &&
305 	    (object->flags & (OBJ_COLORED | OBJ_FICTITIOUS)) == OBJ_COLORED) {
306 		return (true);
307 	}
308 	return (false);
309 }
310 
311 void vm_object_clear_flag(vm_object_t object, u_short bits);
312 void vm_object_pip_add(vm_object_t object, short i);
313 void vm_object_pip_subtract(vm_object_t object, short i);
314 void vm_object_pip_wakeup(vm_object_t object);
315 void vm_object_pip_wakeupn(vm_object_t object, short i);
316 void vm_object_pip_wait(vm_object_t object, char *waitid);
317 
318 void umtx_shm_object_init(vm_object_t object);
319 void umtx_shm_object_terminated(vm_object_t object);
320 extern int umtx_shm_vnobj_persistent;
321 
322 vm_object_t vm_object_allocate (objtype_t, vm_pindex_t);
323 boolean_t vm_object_coalesce(vm_object_t, vm_ooffset_t, vm_size_t, vm_size_t,
324    boolean_t);
325 void vm_object_collapse (vm_object_t);
326 void vm_object_deallocate (vm_object_t);
327 void vm_object_destroy (vm_object_t);
328 void vm_object_terminate (vm_object_t);
329 void vm_object_set_writeable_dirty (vm_object_t);
330 void vm_object_init (void);
331 void vm_object_madvise(vm_object_t, vm_pindex_t, vm_pindex_t, int);
332 boolean_t vm_object_page_clean(vm_object_t object, vm_ooffset_t start,
333     vm_ooffset_t end, int flags);
334 void vm_object_page_noreuse(vm_object_t object, vm_pindex_t start,
335     vm_pindex_t end);
336 void vm_object_page_remove(vm_object_t object, vm_pindex_t start,
337     vm_pindex_t end, int options);
338 boolean_t vm_object_populate(vm_object_t, vm_pindex_t, vm_pindex_t);
339 void vm_object_print(long addr, boolean_t have_addr, long count, char *modif);
340 void vm_object_reference (vm_object_t);
341 void vm_object_reference_locked(vm_object_t);
342 int  vm_object_set_memattr(vm_object_t object, vm_memattr_t memattr);
343 void vm_object_shadow (vm_object_t *, vm_ooffset_t *, vm_size_t);
344 void vm_object_split(vm_map_entry_t);
345 boolean_t vm_object_sync(vm_object_t, vm_ooffset_t, vm_size_t, boolean_t,
346     boolean_t);
347 void vm_object_unwire(vm_object_t object, vm_ooffset_t offset,
348     vm_size_t length, uint8_t queue);
349 struct vnode *vm_object_vnode(vm_object_t object);
350 #endif				/* _KERNEL */
351 
352 #endif				/* _VM_OBJECT_ */
353