master
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
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/_blockcount.h>
72#include <sys/_lock.h>
73#include <sys/_mutex.h>
74#include <sys/_pctrie.h>
75#include <sys/_rwlock.h>
76#include <sys/_domainset.h>
77
78#include <vm/_vm_radix.h>
79
80/*
81 * Types defined:
82 *
83 * vm_object_t Virtual memory object.
84 *
85 * List of locks
86 * (a) atomic
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
94TAILQ_HEAD(pglist, vm_page);
95#define VM_PAGE_HAVE_PGLIST
96#endif
97
98struct 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 volatile int generation; /* generation ID */
108 int cleangeneration; /* Generation at clean time */
109 volatile u_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 pg_color; /* (c) color of first page in obj */
114 u_int flags; /* see below */
115 blockcount_t paging_in_progress; /* (a) Paging (in or out) so don't collapse or destroy */
116 blockcount_t busy; /* (a) object is busy, disallow page busy. */
117 int resident_page_count; /* number of resident pages */
118 struct vm_object *backing_object; /* object that I'm a shadow of */
119 vm_ooffset_t backing_object_offset;/* Offset in backing object */
120 TAILQ_ENTRY(vm_object) pager_object_list; /* list of all objects of this pager type */
121 LIST_HEAD(, vm_reserv) rvq; /* list of reservations */
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 const struct cdev_pager_ops *ops;
142 void *handle;
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_priv - pager-private.
158 * swp_blks - pc-trie of the allocated swap blocks.
159 * writemappings - count of bytes mapped for write
160 *
161 */
162 struct {
163 void *swp_priv;
164 struct pctrie swp_blks;
165 vm_ooffset_t writemappings;
166 } swp;
167
168 /*
169 * Phys pager
170 */
171 struct {
172 const struct phys_pager_ops *ops;
173 union {
174 void *data_ptr;
175 uintptr_t data_val;
176 };
177 void *phys_priv;
178 } phys;
179 } un_pager;
180 struct ucred *cred;
181 vm_ooffset_t charge;
182 void *umtx_data;
183};
184
185/*
186 * Flags
187 */
188#define OBJ_FICTITIOUS 0x00000001 /* (c) contains fictitious pages */
189#define OBJ_UNMANAGED 0x00000002 /* (c) contains unmanaged pages */
190#define OBJ_POPULATE 0x00000004 /* pager implements populate() */
191#define OBJ_DEAD 0x00000008 /* dead objects (during rundown) */
192#define OBJ_ANON 0x00000010 /* (c) contains anonymous memory */
193#define OBJ_UMTXDEAD 0x00000020 /* umtx pshared was terminated */
194#define OBJ_SIZEVNLOCK 0x00000040 /* lock vnode to check obj size */
195#define OBJ_PG_DTOR 0x00000080 /* do not reset object, leave that
196 for dtor */
197#define OBJ_SHADOWLIST 0x00000100 /* Object is on the shadow list. */
198#define OBJ_SWAP 0x00000200 /* object swaps, type will be OBJT_SWAP
199 or dynamically registered */
200#define OBJ_SPLIT 0x00000400 /* object is being split */
201#define OBJ_COLLAPSING 0x00000800 /* Parent of collapse. */
202#define OBJ_COLORED 0x00001000 /* pg_color is defined */
203#define OBJ_ONEMAPPING 0x00002000 /* One USE (a single, non-forked)
204 mapping flag */
205#define OBJ_PAGERPRIV1 0x00004000 /* Pager private */
206#define OBJ_PAGERPRIV2 0x00008000 /* Pager private */
207#define OBJ_SYSVSHM 0x00010000 /* SysV SHM */
208#define OBJ_POSIXSHM 0x00020000 /* Posix SHM */
209#define OBJ_CDEVH 0x00040000 /* OBJT_DEVICE handle is cdev */
210
211/*
212 * Helpers to perform conversion between vm_object page indexes and offsets.
213 * IDX_TO_OFF() converts an index into an offset.
214 * OFF_TO_IDX() converts an offset into an index.
215 * OBJ_MAX_SIZE specifies the maximum page index corresponding to the
216 * maximum unsigned offset.
217 */
218#define IDX_TO_OFF(idx) (((vm_ooffset_t)(idx)) << PAGE_SHIFT)
219#define OFF_TO_IDX(off) ((vm_pindex_t)(((vm_ooffset_t)(off)) >> PAGE_SHIFT))
220#define OBJ_MAX_SIZE (OFF_TO_IDX(UINT64_MAX) + 1)
221
222#ifdef _KERNEL
223
224#define OBJPC_SYNC 0x1 /* sync I/O */
225#define OBJPC_INVAL 0x2 /* invalidate */
226#define OBJPC_NOSYNC 0x4 /* skip if PGA_NOSYNC */
227
228/*
229 * The following options are supported by vm_object_page_remove().
230 */
231#define OBJPR_CLEANONLY 0x1 /* Don't remove dirty pages. */
232#define OBJPR_NOTMAPPED 0x2 /* Don't unmap pages. */
233#define OBJPR_VALIDONLY 0x4 /* Ignore invalid pages. */
234
235TAILQ_HEAD(object_q, vm_object);
236
237extern struct object_q vm_object_list; /* list of allocated objects */
238extern struct mtx vm_object_list_mtx; /* lock for object list and count */
239
240extern struct vm_object kernel_object_store;
241
242/* kernel and kmem are aliased for backwards KPI compat. */
243#define kernel_object (&kernel_object_store)
244#define kmem_object (&kernel_object_store)
245
246#define VM_OBJECT_ASSERT_LOCKED(object) \
247 rw_assert(&(object)->lock, RA_LOCKED)
248#define VM_OBJECT_ASSERT_RLOCKED(object) \
249 rw_assert(&(object)->lock, RA_RLOCKED)
250#define VM_OBJECT_ASSERT_WLOCKED(object) \
251 rw_assert(&(object)->lock, RA_WLOCKED)
252#define VM_OBJECT_ASSERT_UNLOCKED(object) \
253 rw_assert(&(object)->lock, RA_UNLOCKED)
254#define VM_OBJECT_LOCK_DOWNGRADE(object) \
255 rw_downgrade(&(object)->lock)
256#define VM_OBJECT_RLOCK(object) \
257 rw_rlock(&(object)->lock)
258#define VM_OBJECT_RUNLOCK(object) \
259 rw_runlock(&(object)->lock)
260#define VM_OBJECT_SLEEP(object, wchan, pri, wmesg, timo) \
261 rw_sleep((wchan), &(object)->lock, (pri), (wmesg), (timo))
262#define VM_OBJECT_TRYRLOCK(object) \
263 rw_try_rlock(&(object)->lock)
264#define VM_OBJECT_TRYWLOCK(object) \
265 rw_try_wlock(&(object)->lock)
266#define VM_OBJECT_TRYUPGRADE(object) \
267 rw_try_upgrade(&(object)->lock)
268#define VM_OBJECT_WLOCK(object) \
269 rw_wlock(&(object)->lock)
270#define VM_OBJECT_WOWNED(object) \
271 rw_wowned(&(object)->lock)
272#define VM_OBJECT_WUNLOCK(object) \
273 rw_wunlock(&(object)->lock)
274#define VM_OBJECT_UNLOCK(object) \
275 rw_unlock(&(object)->lock)
276#define VM_OBJECT_DROP(object) \
277 lock_class_rw.lc_unlock(&(object)->lock.lock_object)
278#define VM_OBJECT_PICKUP(object, state) \
279 lock_class_rw.lc_lock(&(object)->lock.lock_object, (state))
280
281#define VM_OBJECT_ASSERT_PAGING(object) \
282 KASSERT(blockcount_read(&(object)->paging_in_progress) != 0, \
283 ("vm_object %p is not paging", object))
284#define VM_OBJECT_ASSERT_REFERENCE(object) \
285 KASSERT((object)->reference_count != 0, \
286 ("vm_object %p is not referenced", object))
287
288struct vnode;
289
290/*
291 * The object must be locked or thread private.
292 */
293static __inline void
294vm_object_set_flag(vm_object_t object, u_int bits)
295{
296
297 object->flags |= bits;
298}
299
300/*
301 * Conditionally set the object's color, which (1) enables the allocation
302 * of physical memory reservations for anonymous objects and larger-than-
303 * superpage-sized named objects and (2) determines the first page offset
304 * within the object at which a reservation may be allocated. In other
305 * words, the color determines the alignment of the object with respect
306 * to the largest superpage boundary. When mapping named objects, like
307 * files or POSIX shared memory objects, the color should be set to zero
308 * before a virtual address is selected for the mapping. In contrast,
309 * for anonymous objects, the color may be set after the virtual address
310 * is selected.
311 *
312 * The object must be locked.
313 */
314static __inline void
315vm_object_color(vm_object_t object, u_short color)
316{
317
318 if ((object->flags & OBJ_COLORED) == 0) {
319 object->pg_color = color;
320 vm_object_set_flag(object, OBJ_COLORED);
321 }
322}
323
324static __inline bool
325vm_object_reserv(vm_object_t object)
326{
327
328 if (object != NULL &&
329 (object->flags & (OBJ_COLORED | OBJ_FICTITIOUS)) == OBJ_COLORED) {
330 return (true);
331 }
332 return (false);
333}
334
335void vm_object_clear_flag(vm_object_t object, u_short bits);
336void vm_object_pip_add(vm_object_t object, short i);
337void vm_object_pip_wakeup(vm_object_t object);
338void vm_object_pip_wakeupn(vm_object_t object, short i);
339void vm_object_pip_wait(vm_object_t object, const char *waitid);
340void vm_object_pip_wait_unlocked(vm_object_t object, const char *waitid);
341
342void vm_object_busy(vm_object_t object);
343void vm_object_unbusy(vm_object_t object);
344void vm_object_busy_wait(vm_object_t object, const char *wmesg);
345
346static inline bool
347vm_object_busied(vm_object_t object)
348{
349
350 return (blockcount_read(&object->busy) != 0);
351}
352#define VM_OBJECT_ASSERT_BUSY(object) MPASS(vm_object_busied((object)))
353
354void umtx_shm_object_init(vm_object_t object);
355void umtx_shm_object_terminated(vm_object_t object);
356extern int umtx_shm_vnobj_persistent;
357
358vm_object_t vm_object_allocate (objtype_t, vm_pindex_t);
359vm_object_t vm_object_allocate_anon(vm_pindex_t, vm_object_t, struct ucred *,
360 vm_size_t);
361vm_object_t vm_object_allocate_dyn(objtype_t, vm_pindex_t, u_short);
362boolean_t vm_object_coalesce(vm_object_t, vm_ooffset_t, vm_size_t, vm_size_t,
363 boolean_t);
364void vm_object_collapse (vm_object_t);
365void vm_object_deallocate (vm_object_t);
366void vm_object_destroy (vm_object_t);
367void vm_object_terminate (vm_object_t);
368void vm_object_set_writeable_dirty (vm_object_t);
369void vm_object_set_writeable_dirty_(vm_object_t object);
370bool vm_object_mightbedirty(vm_object_t object);
371bool vm_object_mightbedirty_(vm_object_t object);
372void vm_object_init (void);
373int vm_object_kvme_type(vm_object_t object, struct vnode **vpp);
374void vm_object_madvise(vm_object_t, vm_pindex_t, vm_pindex_t, int);
375boolean_t vm_object_page_clean(vm_object_t object, vm_ooffset_t start,
376 vm_ooffset_t end, int flags);
377void vm_object_page_noreuse(vm_object_t object, vm_pindex_t start,
378 vm_pindex_t end);
379void vm_object_page_remove(vm_object_t object, vm_pindex_t start,
380 vm_pindex_t end, int options);
381boolean_t vm_object_populate(vm_object_t, vm_pindex_t, vm_pindex_t);
382void vm_object_print(long addr, boolean_t have_addr, long count, char *modif);
383void vm_object_reference (vm_object_t);
384void vm_object_reference_locked(vm_object_t);
385int vm_object_set_memattr(vm_object_t object, vm_memattr_t memattr);
386void vm_object_shadow(vm_object_t *, vm_ooffset_t *, vm_size_t, struct ucred *,
387 bool);
388void vm_object_split(vm_map_entry_t);
389boolean_t vm_object_sync(vm_object_t, vm_ooffset_t, vm_size_t, boolean_t,
390 boolean_t);
391void vm_object_unwire(vm_object_t object, vm_ooffset_t offset,
392 vm_size_t length, uint8_t queue);
393struct vnode *vm_object_vnode(vm_object_t object);
394bool vm_object_is_active(vm_object_t obj);
395#endif /* _KERNEL */
396
397#endif /* _VM_OBJECT_ */