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