xref: /freebsd/sys/vm/vm_object.h (revision fa4d25f5b4573a54eebeb7f254b52153b8d3811e)
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/_blockcount.h>
74 #include <sys/_lock.h>
75 #include <sys/_mutex.h>
76 #include <sys/_pctrie.h>
77 #include <sys/_rwlock.h>
78 #include <sys/_domainset.h>
79 
80 #include <vm/_vm_radix.h>
81 
82 /*
83  *	Types defined:
84  *
85  *	vm_object_t		Virtual memory object.
86  *
87  * List of locks
88  *	(a)	atomic
89  *	(c)	const until freed
90  *	(o)	per-object lock
91  *	(f)	free pages queue mutex
92  *
93  */
94 
95 #ifndef VM_PAGE_HAVE_PGLIST
96 TAILQ_HEAD(pglist, vm_page);
97 #define VM_PAGE_HAVE_PGLIST
98 #endif
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 	struct pglist 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 	struct domainset_ref domain;	/* NUMA policy. */
109 	volatile int generation;	/* generation ID */
110 	int cleangeneration;		/* Generation at clean time */
111 	volatile u_int ref_count;	/* How many refs?? */
112 	int shadow_count;		/* how many objects that this is a shadow for */
113 	vm_memattr_t memattr;		/* default memory attribute for pages */
114 	objtype_t type;			/* type of pager */
115 	u_short flags;			/* see below */
116 	u_short pg_color;		/* (c) color of first page in obj */
117 	blockcount_t paging_in_progress; /* (a) Paging (in or out) so don't collapse or destroy */
118 	blockcount_t busy;		/* (a) object is busy, disallow page busy. */
119 	int resident_page_count;	/* number of resident pages */
120 	struct vm_object *backing_object; /* object that I'm a shadow of */
121 	vm_ooffset_t backing_object_offset;/* Offset in backing object */
122 	TAILQ_ENTRY(vm_object) pager_object_list; /* list of all objects of this pager type */
123 	LIST_HEAD(, vm_reserv) rvq;	/* list of reservations */
124 	void *handle;
125 	union {
126 		/*
127 		 * VNode pager
128 		 *
129 		 *	vnp_size - current size of file
130 		 */
131 		struct {
132 			off_t vnp_size;
133 			vm_ooffset_t writemappings;
134 		} vnp;
135 
136 		/*
137 		 * Device pager
138 		 *
139 		 *	devp_pglist - list of allocated pages
140 		 */
141 		struct {
142 			TAILQ_HEAD(, vm_page) devp_pglist;
143 			const struct cdev_pager_ops *ops;
144 			struct cdev *dev;
145 		} devp;
146 
147 		/*
148 		 * SG pager
149 		 *
150 		 *	sgp_pglist - list of allocated pages
151 		 */
152 		struct {
153 			TAILQ_HEAD(, vm_page) sgp_pglist;
154 		} sgp;
155 
156 		/*
157 		 * Swap pager
158 		 *
159 		 *	swp_priv - pager-private.
160 		 *	swp_blks - pc-trie of the allocated swap blocks.
161 		 *	writemappings - count of bytes mapped for write
162 		 *
163 		 */
164 		struct {
165 			void *swp_priv;
166 			struct pctrie swp_blks;
167 			vm_ooffset_t writemappings;
168 		} swp;
169 
170 		/*
171 		 * Phys pager
172 		 */
173 		struct {
174 			const struct phys_pager_ops *ops;
175 			union {
176 				void *data_ptr;
177 				uintptr_t data_val;
178 			};
179 		} phys;
180 	} un_pager;
181 	struct ucred *cred;
182 	vm_ooffset_t charge;
183 	void *umtx_data;
184 };
185 
186 /*
187  * Flags
188  */
189 #define	OBJ_FICTITIOUS	0x0001		/* (c) contains fictitious pages */
190 #define	OBJ_UNMANAGED	0x0002		/* (c) contains unmanaged pages */
191 #define	OBJ_POPULATE	0x0004		/* pager implements populate() */
192 #define	OBJ_DEAD	0x0008		/* dead objects (during rundown) */
193 #define	OBJ_ANON	0x0010		/* (c) contains anonymous memory */
194 #define	OBJ_UMTXDEAD	0x0020		/* umtx pshared was terminated */
195 #define	OBJ_SIZEVNLOCK	0x0040		/* lock vnode to check obj size */
196 #define	OBJ_PG_DTOR	0x0080		/* dont reset object, leave that for dtor */
197 #define	OBJ_SHADOWLIST	0x0100		/* Object is on the shadow list. */
198 #define	OBJ_SWAP	0x0200		/* object swaps, type will be OBJT_SWAP
199 					   or dynamically registered */
200 #define	OBJ_SPLIT	0x0400		/* object is being split */
201 #define	OBJ_COLLAPSING	0x0800		/* Parent of collapse. */
202 #define	OBJ_COLORED	0x1000		/* pg_color is defined */
203 #define	OBJ_ONEMAPPING	0x2000		/* One USE (a single, non-forked) mapping flag */
204 #define	OBJ_PAGERPRIV1	0x4000		/* Pager private */
205 #define	OBJ_PAGERPRIV2	0x8000		/* Pager private */
206 
207 /*
208  * Helpers to perform conversion between vm_object page indexes and offsets.
209  * IDX_TO_OFF() converts an index into an offset.
210  * OFF_TO_IDX() converts an offset into an index.
211  * OBJ_MAX_SIZE specifies the maximum page index corresponding to the
212  *   maximum unsigned offset.
213  */
214 #define	IDX_TO_OFF(idx) (((vm_ooffset_t)(idx)) << PAGE_SHIFT)
215 #define	OFF_TO_IDX(off) ((vm_pindex_t)(((vm_ooffset_t)(off)) >> PAGE_SHIFT))
216 #define	OBJ_MAX_SIZE	(OFF_TO_IDX(UINT64_MAX) + 1)
217 
218 #ifdef	_KERNEL
219 
220 #define OBJPC_SYNC	0x1			/* sync I/O */
221 #define OBJPC_INVAL	0x2			/* invalidate */
222 #define OBJPC_NOSYNC	0x4			/* skip if PGA_NOSYNC */
223 
224 /*
225  * The following options are supported by vm_object_page_remove().
226  */
227 #define	OBJPR_CLEANONLY	0x1		/* Don't remove dirty pages. */
228 #define	OBJPR_NOTMAPPED	0x2		/* Don't unmap pages. */
229 #define	OBJPR_VALIDONLY	0x4		/* Ignore invalid pages. */
230 
231 TAILQ_HEAD(object_q, vm_object);
232 
233 extern struct object_q vm_object_list;	/* list of allocated objects */
234 extern struct mtx vm_object_list_mtx;	/* lock for object list and count */
235 
236 extern struct vm_object kernel_object_store;
237 
238 /* kernel and kmem are aliased for backwards KPI compat. */
239 #define	kernel_object	(&kernel_object_store)
240 #define	kmem_object	(&kernel_object_store)
241 
242 #define	VM_OBJECT_ASSERT_LOCKED(object)					\
243 	rw_assert(&(object)->lock, RA_LOCKED)
244 #define	VM_OBJECT_ASSERT_RLOCKED(object)				\
245 	rw_assert(&(object)->lock, RA_RLOCKED)
246 #define	VM_OBJECT_ASSERT_WLOCKED(object)				\
247 	rw_assert(&(object)->lock, RA_WLOCKED)
248 #define	VM_OBJECT_ASSERT_UNLOCKED(object)				\
249 	rw_assert(&(object)->lock, RA_UNLOCKED)
250 #define	VM_OBJECT_LOCK_DOWNGRADE(object)				\
251 	rw_downgrade(&(object)->lock)
252 #define	VM_OBJECT_RLOCK(object)						\
253 	rw_rlock(&(object)->lock)
254 #define	VM_OBJECT_RUNLOCK(object)					\
255 	rw_runlock(&(object)->lock)
256 #define	VM_OBJECT_SLEEP(object, wchan, pri, wmesg, timo)		\
257 	rw_sleep((wchan), &(object)->lock, (pri), (wmesg), (timo))
258 #define	VM_OBJECT_TRYRLOCK(object)					\
259 	rw_try_rlock(&(object)->lock)
260 #define	VM_OBJECT_TRYWLOCK(object)					\
261 	rw_try_wlock(&(object)->lock)
262 #define	VM_OBJECT_TRYUPGRADE(object)					\
263 	rw_try_upgrade(&(object)->lock)
264 #define	VM_OBJECT_WLOCK(object)						\
265 	rw_wlock(&(object)->lock)
266 #define	VM_OBJECT_WOWNED(object)					\
267 	rw_wowned(&(object)->lock)
268 #define	VM_OBJECT_WUNLOCK(object)					\
269 	rw_wunlock(&(object)->lock)
270 #define	VM_OBJECT_DROP(object)						\
271 	lock_class_rw.lc_unlock(&(object)->lock.lock_object)
272 #define	VM_OBJECT_PICKUP(object, state)					\
273 	lock_class_rw.lc_lock(&(object)->lock.lock_object, (state))
274 
275 #define	VM_OBJECT_ASSERT_PAGING(object)					\
276 	KASSERT(blockcount_read(&(object)->paging_in_progress) != 0,	\
277 	    ("vm_object %p is not paging", object))
278 #define	VM_OBJECT_ASSERT_REFERENCE(object)				\
279 	KASSERT((object)->reference_count != 0,				\
280 	    ("vm_object %p is not referenced", object))
281 
282 struct vnode;
283 
284 /*
285  *	The object must be locked or thread private.
286  */
287 static __inline void
288 vm_object_set_flag(vm_object_t object, u_short bits)
289 {
290 
291 	object->flags |= bits;
292 }
293 
294 /*
295  *	Conditionally set the object's color, which (1) enables the allocation
296  *	of physical memory reservations for anonymous objects and larger-than-
297  *	superpage-sized named objects and (2) determines the first page offset
298  *	within the object at which a reservation may be allocated.  In other
299  *	words, the color determines the alignment of the object with respect
300  *	to the largest superpage boundary.  When mapping named objects, like
301  *	files or POSIX shared memory objects, the color should be set to zero
302  *	before a virtual address is selected for the mapping.  In contrast,
303  *	for anonymous objects, the color may be set after the virtual address
304  *	is selected.
305  *
306  *	The object must be locked.
307  */
308 static __inline void
309 vm_object_color(vm_object_t object, u_short color)
310 {
311 
312 	if ((object->flags & OBJ_COLORED) == 0) {
313 		object->pg_color = color;
314 		vm_object_set_flag(object, OBJ_COLORED);
315 	}
316 }
317 
318 static __inline bool
319 vm_object_reserv(vm_object_t object)
320 {
321 
322 	if (object != NULL &&
323 	    (object->flags & (OBJ_COLORED | OBJ_FICTITIOUS)) == OBJ_COLORED) {
324 		return (true);
325 	}
326 	return (false);
327 }
328 
329 void vm_object_clear_flag(vm_object_t object, u_short bits);
330 void vm_object_pip_add(vm_object_t object, short i);
331 void vm_object_pip_wakeup(vm_object_t object);
332 void vm_object_pip_wakeupn(vm_object_t object, short i);
333 void vm_object_pip_wait(vm_object_t object, const char *waitid);
334 void vm_object_pip_wait_unlocked(vm_object_t object, const char *waitid);
335 
336 void vm_object_busy(vm_object_t object);
337 void vm_object_unbusy(vm_object_t object);
338 void vm_object_busy_wait(vm_object_t object, const char *wmesg);
339 
340 static inline bool
341 vm_object_busied(vm_object_t object)
342 {
343 
344 	return (blockcount_read(&object->busy) != 0);
345 }
346 #define	VM_OBJECT_ASSERT_BUSY(object)	MPASS(vm_object_busied((object)))
347 
348 void umtx_shm_object_init(vm_object_t object);
349 void umtx_shm_object_terminated(vm_object_t object);
350 extern int umtx_shm_vnobj_persistent;
351 
352 vm_object_t vm_object_allocate (objtype_t, vm_pindex_t);
353 vm_object_t vm_object_allocate_anon(vm_pindex_t, vm_object_t, struct ucred *,
354    vm_size_t);
355 vm_object_t vm_object_allocate_dyn(objtype_t, vm_pindex_t, u_short);
356 boolean_t vm_object_coalesce(vm_object_t, vm_ooffset_t, vm_size_t, vm_size_t,
357    boolean_t);
358 void vm_object_collapse (vm_object_t);
359 void vm_object_deallocate (vm_object_t);
360 void vm_object_destroy (vm_object_t);
361 void vm_object_terminate (vm_object_t);
362 void vm_object_set_writeable_dirty (vm_object_t);
363 void vm_object_set_writeable_dirty_(vm_object_t object);
364 bool vm_object_mightbedirty(vm_object_t object);
365 bool vm_object_mightbedirty_(vm_object_t object);
366 void vm_object_init (void);
367 int  vm_object_kvme_type(vm_object_t object, struct vnode **vpp);
368 void vm_object_madvise(vm_object_t, vm_pindex_t, vm_pindex_t, int);
369 boolean_t vm_object_page_clean(vm_object_t object, vm_ooffset_t start,
370     vm_ooffset_t end, int flags);
371 void vm_object_page_noreuse(vm_object_t object, vm_pindex_t start,
372     vm_pindex_t end);
373 void vm_object_page_remove(vm_object_t object, vm_pindex_t start,
374     vm_pindex_t end, int options);
375 boolean_t vm_object_populate(vm_object_t, vm_pindex_t, vm_pindex_t);
376 void vm_object_print(long addr, boolean_t have_addr, long count, char *modif);
377 void vm_object_reference (vm_object_t);
378 void vm_object_reference_locked(vm_object_t);
379 int  vm_object_set_memattr(vm_object_t object, vm_memattr_t memattr);
380 void vm_object_shadow(vm_object_t *, vm_ooffset_t *, vm_size_t, struct ucred *,
381     bool);
382 void vm_object_split(vm_map_entry_t);
383 boolean_t vm_object_sync(vm_object_t, vm_ooffset_t, vm_size_t, boolean_t,
384     boolean_t);
385 void vm_object_unwire(vm_object_t object, vm_ooffset_t offset,
386     vm_size_t length, uint8_t queue);
387 struct vnode *vm_object_vnode(vm_object_t object);
388 bool vm_object_is_active(vm_object_t obj);
389 #endif				/* _KERNEL */
390 
391 #endif				/* _VM_OBJECT_ */
392