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 *
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
61 /*
62 * Virtual memory object module definitions.
63 */
64
65 #ifndef _VM_OBJECT_
66 #define _VM_OBJECT_
67
68 #include <sys/queue.h>
69 #include <sys/_blockcount.h>
70 #include <sys/_lock.h>
71 #include <sys/_mutex.h>
72 #include <sys/_pctrie.h>
73 #include <sys/_rwlock.h>
74 #include <sys/_domainset.h>
75
76 #include <vm/_vm_radix.h>
77
78 /*
79 * Types defined:
80 *
81 * vm_object_t Virtual memory object.
82 *
83 * List of locks
84 * (a) atomic
85 * (c) const until freed
86 * (o) per-object lock
87 * (f) free pages queue mutex
88 *
89 */
90
91 #ifndef VM_PAGE_HAVE_PGLIST
92 TAILQ_HEAD(pglist, vm_page);
93 #define VM_PAGE_HAVE_PGLIST
94 #endif
95
96 struct vm_object {
97 struct rwlock lock;
98 TAILQ_ENTRY(vm_object) object_list; /* list of all objects */
99 LIST_HEAD(, vm_object) shadow_head; /* objects that this is a shadow for */
100 LIST_ENTRY(vm_object) shadow_list; /* chain of shadow objects */
101 struct vm_radix rtree; /* root of the resident page radix trie*/
102 vm_pindex_t size; /* Object size */
103 struct domainset_ref domain; /* NUMA policy. */
104 volatile int generation; /* generation ID */
105 int cleangeneration; /* Generation at clean time */
106 volatile u_int ref_count; /* How many refs?? */
107 int shadow_count; /* how many objects that this is a shadow for */
108 vm_memattr_t memattr; /* default memory attribute for pages */
109 objtype_t type; /* type of pager */
110 u_short pg_color; /* (c) color of first page in obj */
111 u_int flags; /* see below */
112 blockcount_t paging_in_progress; /* (a) Paging (in or out) so don't collapse or destroy */
113 blockcount_t busy; /* (a) object is busy, disallow page busy. */
114 int resident_page_count; /* number of resident pages */
115 struct vm_object *backing_object; /* object that I'm a shadow of */
116 vm_ooffset_t backing_object_offset;/* Offset in backing object */
117 TAILQ_ENTRY(vm_object) pager_object_list; /* list of all objects of this pager type */
118 LIST_HEAD(, vm_reserv) rvq; /* list of reservations */
119 void *handle;
120 union {
121 /*
122 * VNode pager
123 *
124 * vnp_size - current size of file
125 */
126 struct {
127 off_t vnp_size;
128 vm_ooffset_t writemappings;
129 } vnp;
130
131 /*
132 * Device pager
133 *
134 * devp_pglist - list of allocated pages
135 */
136 struct {
137 TAILQ_HEAD(, vm_page) devp_pglist;
138 const struct cdev_pager_ops *ops;
139 void *handle;
140 } devp;
141
142 /*
143 * SG pager
144 *
145 * sgp_pglist - list of allocated pages
146 */
147 struct {
148 TAILQ_HEAD(, vm_page) sgp_pglist;
149 } sgp;
150
151 /*
152 * Swap pager
153 *
154 * swp_priv - pager-private.
155 * swp_blks - pc-trie of the allocated swap blocks.
156 * writemappings - count of bytes mapped for write
157 *
158 */
159 struct {
160 void *swp_priv;
161 struct pctrie swp_blks;
162 vm_ooffset_t writemappings;
163 } swp;
164
165 /*
166 * Phys pager
167 */
168 struct {
169 const struct phys_pager_ops *ops;
170 union {
171 void *data_ptr;
172 uintptr_t data_val;
173 };
174 void *phys_priv;
175 } phys;
176 } un_pager;
177 struct ucred *cred;
178 void *umtx_data;
179 };
180
181 /*
182 * Flags
183 */
184 #define OBJ_FICTITIOUS 0x00000001 /* (c) contains fictitious pages */
185 #define OBJ_UNMANAGED 0x00000002 /* (c) contains unmanaged pages */
186 #define OBJ_POPULATE 0x00000004 /* pager implements populate() */
187 #define OBJ_DEAD 0x00000008 /* dead objects (during rundown) */
188 #define OBJ_ANON 0x00000010 /* (c) contains anonymous memory */
189 #define OBJ_UMTXDEAD 0x00000020 /* umtx pshared was terminated */
190 #define OBJ_SIZEVNLOCK 0x00000040 /* lock vnode to check obj size */
191 #define OBJ_PG_DTOR 0x00000080 /* do not reset object, leave that
192 for dtor */
193 #define OBJ_SHADOWLIST 0x00000100 /* Object is on the shadow list. */
194 #define OBJ_SWAP 0x00000200 /* object swaps, type will be OBJT_SWAP
195 or dynamically registered */
196 #define OBJ_SPLIT 0x00000400 /* object is being split */
197 #define OBJ_COLLAPSING 0x00000800 /* Parent of collapse. */
198 #define OBJ_COLORED 0x00001000 /* pg_color is defined */
199 #define OBJ_ONEMAPPING 0x00002000 /* Each page has at most one managed
200 mapping, all in the same vm_map */
201 #define OBJ_PAGERPRIV1 0x00004000 /* Pager private */
202 #define OBJ_PAGERPRIV2 0x00008000 /* Pager private */
203 #define OBJ_SYSVSHM 0x00010000 /* SysV SHM */
204 #define OBJ_POSIXSHM 0x00020000 /* Posix SHM */
205
206 /*
207 * Helpers to perform conversion between vm_object page indexes and offsets.
208 * IDX_TO_OFF() converts an index into an offset.
209 * OFF_TO_IDX() converts an offset into an index.
210 * OBJ_MAX_SIZE specifies the maximum page index corresponding to the
211 * maximum unsigned offset.
212 */
213 #define IDX_TO_OFF(idx) (((vm_ooffset_t)(idx)) << PAGE_SHIFT)
214 #define OFF_TO_IDX(off) ((vm_pindex_t)(((vm_ooffset_t)(off)) >> PAGE_SHIFT))
215 #define OBJ_MAX_SIZE (OFF_TO_IDX(UINT64_MAX) + 1)
216
217 #ifdef _KERNEL
218
219 #define OBJPC_SYNC 0x1 /* sync I/O */
220 #define OBJPC_INVAL 0x2 /* invalidate */
221 #define OBJPC_NOSYNC 0x4 /* skip if PGA_NOSYNC */
222
223 /*
224 * The following options are supported by vm_object_page_remove().
225 */
226 #define OBJPR_CLEANONLY 0x1 /* Don't remove dirty pages. */
227 #define OBJPR_NOTMAPPED 0x2 /* Don't unmap pages. */
228 #define OBJPR_VALIDONLY 0x4 /* Ignore invalid pages. */
229
230 /*
231 * Options for vm_object_coalesce().
232 */
233 #define OBJCO_CHARGED 0x1 /* The next_size was charged already */
234 #define OBJCO_NO_CHARGE 0x2 /* Do not do swap accounting at all */
235
236 TAILQ_HEAD(object_q, vm_object);
237
238 extern struct object_q vm_object_list; /* list of allocated objects */
239 extern struct mtx vm_object_list_mtx; /* lock for object list and count */
240
241 extern struct vm_object kernel_object_store;
242
243 #define kernel_object (&kernel_object_store)
244
245 #define VM_OBJECT_ASSERT_LOCKED(object) \
246 rw_assert(&(object)->lock, RA_LOCKED)
247 #define VM_OBJECT_ASSERT_RLOCKED(object) \
248 rw_assert(&(object)->lock, RA_RLOCKED)
249 #define VM_OBJECT_ASSERT_WLOCKED(object) \
250 rw_assert(&(object)->lock, RA_WLOCKED)
251 #define VM_OBJECT_ASSERT_UNLOCKED(object) \
252 rw_assert(&(object)->lock, RA_UNLOCKED)
253 #define VM_OBJECT_LOCK_DOWNGRADE(object) \
254 rw_downgrade(&(object)->lock)
255 #define VM_OBJECT_RLOCK(object) \
256 rw_rlock(&(object)->lock)
257 #define VM_OBJECT_RUNLOCK(object) \
258 rw_runlock(&(object)->lock)
259 #define VM_OBJECT_SLEEP(object, wchan, pri, wmesg, timo) \
260 rw_sleep((wchan), &(object)->lock, (pri), (wmesg), (timo))
261 #define VM_OBJECT_TRYRLOCK(object) \
262 rw_try_rlock(&(object)->lock)
263 #define VM_OBJECT_TRYWLOCK(object) \
264 rw_try_wlock(&(object)->lock)
265 #define VM_OBJECT_TRYUPGRADE(object) \
266 rw_try_upgrade(&(object)->lock)
267 #define VM_OBJECT_WLOCK(object) \
268 rw_wlock(&(object)->lock)
269 #define VM_OBJECT_WOWNED(object) \
270 rw_wowned(&(object)->lock)
271 #define VM_OBJECT_WUNLOCK(object) \
272 rw_wunlock(&(object)->lock)
273 #define VM_OBJECT_UNLOCK(object) \
274 rw_unlock(&(object)->lock)
275 #define VM_OBJECT_DROP(object) \
276 lock_class_rw.lc_unlock(&(object)->lock.lock_object)
277 #define VM_OBJECT_PICKUP(object, state) \
278 lock_class_rw.lc_lock(&(object)->lock.lock_object, (state))
279
280 #define VM_OBJECT_ASSERT_PAGING(object) \
281 KASSERT(blockcount_read(&(object)->paging_in_progress) != 0, \
282 ("vm_object %p is not paging", object))
283 #define VM_OBJECT_ASSERT_REFERENCE(object) \
284 KASSERT((object)->reference_count != 0, \
285 ("vm_object %p is not referenced", object))
286
287 struct vnode;
288
289 /*
290 * The object must be locked or thread private.
291 */
292 static __inline void
vm_object_set_flag(vm_object_t object,u_int bits)293 vm_object_set_flag(vm_object_t object, u_int bits)
294 {
295
296 object->flags |= bits;
297 }
298
299 /*
300 * Conditionally set the object's color, which (1) enables the allocation
301 * of physical memory reservations for anonymous objects and larger-than-
302 * superpage-sized named objects and (2) determines the first page offset
303 * within the object at which a reservation may be allocated. In other
304 * words, the color determines the alignment of the object with respect
305 * to the largest superpage boundary. When mapping named objects, like
306 * files or POSIX shared memory objects, the color should be set to zero
307 * before a virtual address is selected for the mapping. In contrast,
308 * for anonymous objects, the color may be set after the virtual address
309 * is selected.
310 *
311 * The object must be locked.
312 */
313 static __inline void
vm_object_color(vm_object_t object,u_short color)314 vm_object_color(vm_object_t object, u_short color)
315 {
316
317 if ((object->flags & OBJ_COLORED) == 0) {
318 object->pg_color = color;
319 vm_object_set_flag(object, OBJ_COLORED);
320 }
321 }
322
323 static __inline bool
vm_object_reserv(vm_object_t object)324 vm_object_reserv(vm_object_t object)
325 {
326
327 if (object != NULL &&
328 (object->flags & (OBJ_COLORED | OBJ_FICTITIOUS)) == OBJ_COLORED) {
329 return (true);
330 }
331 return (false);
332 }
333
334 void vm_object_clear_flag(vm_object_t object, u_short bits);
335 void vm_object_pip_add(vm_object_t object, short i);
336 void vm_object_pip_wakeup(vm_object_t object);
337 void vm_object_pip_wakeupn(vm_object_t object, short i);
338 void vm_object_pip_wait(vm_object_t object, const char *waitid);
339 void vm_object_pip_wait_unlocked(vm_object_t object, const char *waitid);
340
341 void vm_object_busy(vm_object_t object);
342 void vm_object_unbusy(vm_object_t object);
343 void vm_object_busy_wait(vm_object_t object, const char *wmesg);
344
345 static inline bool
vm_object_busied(vm_object_t object)346 vm_object_busied(vm_object_t object)
347 {
348
349 return (blockcount_read(&object->busy) != 0);
350 }
351 #define VM_OBJECT_ASSERT_BUSY(object) MPASS(vm_object_busied((object)))
352
353 void umtx_shm_object_init(vm_object_t object);
354 void umtx_shm_object_terminated(vm_object_t object);
355 extern int umtx_shm_vnobj_persistent;
356
357 vm_object_t vm_object_allocate (objtype_t, vm_pindex_t);
358 vm_object_t vm_object_allocate_anon(vm_pindex_t, vm_object_t, struct ucred *);
359 vm_object_t vm_object_allocate_dyn(objtype_t, vm_pindex_t, u_short);
360 boolean_t vm_object_coalesce(vm_object_t, vm_ooffset_t, vm_size_t, vm_size_t,
361 int);
362 void vm_object_collapse (vm_object_t);
363 void vm_object_deallocate (vm_object_t);
364 void vm_object_destroy (vm_object_t);
365 void vm_object_terminate (vm_object_t);
366 void vm_object_set_writeable_dirty (vm_object_t);
367 void vm_object_set_writeable_dirty_(vm_object_t object);
368 bool vm_object_mightbedirty(vm_object_t object);
369 bool vm_object_mightbedirty_(vm_object_t object);
370 void vm_object_init (void);
371 int vm_object_kvme_type(vm_object_t object, struct vnode **vpp);
372 void vm_object_madvise(vm_object_t, vm_pindex_t, vm_pindex_t, int);
373 boolean_t vm_object_page_clean(vm_object_t object, vm_ooffset_t start,
374 vm_ooffset_t end, int flags);
375 void vm_object_page_noreuse(vm_object_t object, vm_pindex_t start,
376 vm_pindex_t end);
377 void vm_object_page_remove(vm_object_t object, vm_pindex_t start,
378 vm_pindex_t end, int options);
379 boolean_t vm_object_populate(vm_object_t, vm_pindex_t, vm_pindex_t);
380 void vm_object_prepare_buf_pages(vm_object_t object, vm_page_t *ma_dst,
381 int count, int *rbehind, int *rahead, vm_page_t *ma_src);
382 void vm_object_print(long addr, boolean_t have_addr, long count, char *modif);
383 void vm_object_reference (vm_object_t);
384 void vm_object_reference_locked(vm_object_t);
385 int vm_object_set_memattr(vm_object_t object, vm_memattr_t memattr);
386 void vm_object_shadow(vm_object_t *, vm_ooffset_t *, vm_size_t, struct ucred *,
387 bool);
388 void vm_object_split(vm_map_entry_t);
389 boolean_t vm_object_sync(vm_object_t, vm_ooffset_t, vm_size_t, boolean_t,
390 boolean_t);
391 void vm_object_unwire(vm_object_t object, vm_ooffset_t offset,
392 vm_size_t length, uint8_t queue);
393 struct vnode *vm_object_vnode(vm_object_t object);
394 bool vm_object_is_active(vm_object_t obj);
395 #endif /* _KERNEL */
396
397 #endif /* _VM_OBJECT_ */
398