1 /*-
2 * Copyright (c) 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * The Mach Operating System project at Carnegie-Mellon University.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 * from: @(#)vm_page.h 8.2 (Berkeley) 12/13/93
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 * $FreeBSD$
61 */
62
63 /*
64 * Resident memory system definitions.
65 */
66
67 #ifndef _VM_PAGE_
68 #define _VM_PAGE_
69
70 #include <vm/pmap.h>
71
72 /*
73 * Management of resident (logical) pages.
74 *
75 * A small structure is kept for each resident
76 * page, indexed by page number. Each structure
77 * is an element of several collections:
78 *
79 * A radix tree used to quickly
80 * perform object/offset lookups
81 *
82 * A list of all pages for a given object,
83 * so they can be quickly deactivated at
84 * time of deallocation.
85 *
86 * An ordered list of pages due for pageout.
87 *
88 * In addition, the structure contains the object
89 * and offset to which this page belongs (for pageout),
90 * and sundry status bits.
91 *
92 * In general, operations on this structure's mutable fields are
93 * synchronized using either one of or a combination of the lock on the
94 * object that the page belongs to (O), the pool lock for the page (P),
95 * or the lock for either the free or paging queue (Q). If a field is
96 * annotated below with two of these locks, then holding either lock is
97 * sufficient for read access, but both locks are required for write
98 * access.
99 *
100 * In contrast, the synchronization of accesses to the page's
101 * dirty field is machine dependent (M). In the
102 * machine-independent layer, the lock on the object that the
103 * page belongs to must be held in order to operate on the field.
104 * However, the pmap layer is permitted to set all bits within
105 * the field without holding that lock. If the underlying
106 * architecture does not support atomic read-modify-write
107 * operations on the field's type, then the machine-independent
108 * layer uses a 32-bit atomic on the aligned 32-bit word that
109 * contains the dirty field. In the machine-independent layer,
110 * the implementation of read-modify-write operations on the
111 * field is encapsulated in vm_page_clear_dirty_mask().
112 */
113
114 #if PAGE_SIZE == 4096
115 #define VM_PAGE_BITS_ALL 0xffu
116 typedef uint8_t vm_page_bits_t;
117 #elif PAGE_SIZE == 8192
118 #define VM_PAGE_BITS_ALL 0xffffu
119 typedef uint16_t vm_page_bits_t;
120 #elif PAGE_SIZE == 16384
121 #define VM_PAGE_BITS_ALL 0xffffffffu
122 typedef uint32_t vm_page_bits_t;
123 #elif PAGE_SIZE == 32768
124 #define VM_PAGE_BITS_ALL 0xfffffffffffffffflu
125 typedef uint64_t vm_page_bits_t;
126 #endif
127
128 struct vm_page {
129 union {
130 TAILQ_ENTRY(vm_page) q; /* page queue or free list (Q) */
131 struct {
132 SLIST_ENTRY(vm_page) ss; /* private slists */
133 void *pv;
134 } s;
135 struct {
136 u_long p;
137 u_long v;
138 } memguard;
139 } plinks;
140 TAILQ_ENTRY(vm_page) listq; /* pages in same object (O) */
141 vm_object_t object; /* which object am I in (O,P) */
142 vm_pindex_t pindex; /* offset into object (O,P) */
143 vm_paddr_t phys_addr; /* physical address of page */
144 struct md_page md; /* machine dependant stuff */
145 u_int wire_count; /* wired down maps refs (P) */
146 volatile u_int busy_lock; /* busy owners lock */
147 uint16_t hold_count; /* page hold count (P) */
148 uint16_t flags; /* page PG_* flags (P) */
149 uint8_t aflags; /* access is atomic */
150 uint8_t oflags; /* page VPO_* flags (O) */
151 uint8_t queue; /* page queue index (P,Q) */
152 int8_t psind; /* pagesizes[] index (O) */
153 int8_t segind;
154 uint8_t order; /* index of the buddy queue */
155 uint8_t pool;
156 u_char act_count; /* page usage count (P) */
157 /* NOTE that these must support one bit per DEV_BSIZE in a page */
158 /* so, on normal X86 kernels, they must be at least 8 bits wide */
159 vm_page_bits_t valid; /* map of valid DEV_BSIZE chunks (O) */
160 vm_page_bits_t dirty; /* map of dirty DEV_BSIZE chunks (M) */
161 };
162
163 /*
164 * Page flags stored in oflags:
165 *
166 * Access to these page flags is synchronized by the lock on the object
167 * containing the page (O).
168 *
169 * Note: VPO_UNMANAGED (used by OBJT_DEVICE, OBJT_PHYS and OBJT_SG)
170 * indicates that the page is not under PV management but
171 * otherwise should be treated as a normal page. Pages not
172 * under PV management cannot be paged out via the
173 * object/vm_page_t because there is no knowledge of their pte
174 * mappings, and such pages are also not on any PQ queue.
175 *
176 */
177 #define VPO_UNUSED01 0x01 /* --available-- */
178 #define VPO_SWAPSLEEP 0x02 /* waiting for swap to finish */
179 #define VPO_UNMANAGED 0x04 /* no PV management for page */
180 #define VPO_SWAPINPROG 0x08 /* swap I/O in progress on page */
181 #define VPO_NOSYNC 0x10 /* do not collect for syncer */
182
183 /*
184 * Busy page implementation details.
185 * The algorithm is taken mostly by rwlock(9) and sx(9) locks implementation,
186 * even if the support for owner identity is removed because of size
187 * constraints. Checks on lock recursion are then not possible, while the
188 * lock assertions effectiveness is somewhat reduced.
189 */
190 #define VPB_BIT_SHARED 0x01
191 #define VPB_BIT_EXCLUSIVE 0x02
192 #define VPB_BIT_WAITERS 0x04
193 #define VPB_BIT_FLAGMASK \
194 (VPB_BIT_SHARED | VPB_BIT_EXCLUSIVE | VPB_BIT_WAITERS)
195
196 #define VPB_SHARERS_SHIFT 3
197 #define VPB_SHARERS(x) \
198 (((x) & ~VPB_BIT_FLAGMASK) >> VPB_SHARERS_SHIFT)
199 #define VPB_SHARERS_WORD(x) ((x) << VPB_SHARERS_SHIFT | VPB_BIT_SHARED)
200 #define VPB_ONE_SHARER (1 << VPB_SHARERS_SHIFT)
201
202 #define VPB_SINGLE_EXCLUSIVER VPB_BIT_EXCLUSIVE
203
204 #define VPB_UNBUSIED VPB_SHARERS_WORD(0)
205
206 #define PQ_NONE 255
207 #define PQ_INACTIVE 0
208 #define PQ_ACTIVE 1
209 #define PQ_COUNT 2
210
211 #include <vm/vm_param.h> /* PA_LOCK_COUNT */
212
213 #define VM_ASSERT(exp) KASSERT(exp, (#exp))
214
215 TAILQ_HEAD(pglist, vm_page);
216 SLIST_HEAD(spglist, vm_page);
217
218 struct vm_pagequeue {
219 struct mtx pq_mutex;
220 struct pglist pq_pl;
221 int pq_cnt;
222 int * const pq_vcnt;
223 const char * const pq_name;
224 } __aligned(CACHE_LINE_SIZE);
225
226
227 struct vm_domain {
228 struct vm_pagequeue vmd_pagequeues[PQ_COUNT + PA_LOCK_COUNT];
229 u_int vmd_page_count;
230 u_int vmd_free_count;
231 long vmd_segs; /* bitmask of the segments */
232 boolean_t vmd_oom;
233 int vmd_pass; /* local pagedaemon pass */
234 int vmd_oom_seq;
235 int vmd_last_active_scan;
236 struct vm_page vmd_marker; /* marker for pagedaemon private use */
237 struct vm_page vmd_inacthead; /* marker for LRU-defeating insertions */
238 };
239 #define vm_page_queue_idx(m) \
240 (PQ_COUNT + (pa_index(VM_PAGE_TO_PHYS((m))) % PA_LOCK_COUNT))
241
242 extern struct vm_domain vm_dom[MAXMEMDOM];
243
244 #define vm_pagequeue_assert_locked(pq) mtx_assert(&(pq)->pq_mutex, MA_OWNED)
245 #define vm_pagequeue_lock(pq) mtx_lock(&(pq)->pq_mutex)
246 #define vm_pagequeue_trylock(pq) mtx_trylock(&(pq)->pq_mutex)
247 #define vm_pagequeue_unlock(pq) mtx_unlock(&(pq)->pq_mutex)
248
249 #ifdef _KERNEL
250 static __inline void
vm_pagequeue_cnt_add(struct vm_pagequeue * pq,int addend)251 vm_pagequeue_cnt_add(struct vm_pagequeue *pq, int addend)
252 {
253
254 #ifdef notyet
255 vm_pagequeue_assert_locked(pq);
256 #endif
257 pq->pq_cnt += addend;
258 atomic_add_int(pq->pq_vcnt, addend);
259 }
260 #define vm_pagequeue_cnt_inc(pq) vm_pagequeue_cnt_add((pq), 1)
261 #define vm_pagequeue_cnt_dec(pq) vm_pagequeue_cnt_add((pq), -1)
262 #endif /* _KERNEL */
263
264 extern struct mtx_padalign vm_page_queue_free_mtx;
265 extern struct mtx_padalign pa_lock[];
266
267 #if defined(__arm__)
268 #define PDRSHIFT PDR_SHIFT
269 #elif !defined(PDRSHIFT)
270 #define PDRSHIFT 21
271 #endif
272
273 #define pa_index(pa) ((pa) >> PDRSHIFT)
274 #define PA_LOCKPTR(pa) ((struct mtx *)(&pa_lock[pa_index(pa) % PA_LOCK_COUNT]))
275 #define PA_LOCKOBJPTR(pa) ((struct lock_object *)PA_LOCKPTR((pa)))
276 #define PA_LOCK(pa) mtx_lock(PA_LOCKPTR(pa))
277 #define PA_TRYLOCK(pa) mtx_trylock(PA_LOCKPTR(pa))
278 #define PA_UNLOCK(pa) mtx_unlock(PA_LOCKPTR(pa))
279 #define PA_UNLOCK_COND(pa) \
280 do { \
281 if ((pa) != 0) { \
282 PA_UNLOCK((pa)); \
283 (pa) = 0; \
284 } \
285 } while (0)
286
287 #define PA_LOCK_ASSERT(pa, a) mtx_assert(PA_LOCKPTR(pa), (a))
288
289 #ifdef KLD_MODULE
290 #define vm_page_lock(m) vm_page_lock_KBI((m), LOCK_FILE, LOCK_LINE)
291 #define vm_page_unlock(m) vm_page_unlock_KBI((m), LOCK_FILE, LOCK_LINE)
292 #define vm_page_trylock(m) vm_page_trylock_KBI((m), LOCK_FILE, LOCK_LINE)
293 #else /* !KLD_MODULE */
294 #define vm_page_lockptr(m) (PA_LOCKPTR(VM_PAGE_TO_PHYS((m))))
295 #define vm_page_lock(m) mtx_lock(vm_page_lockptr((m)))
296 #define vm_page_unlock(m) mtx_unlock(vm_page_lockptr((m)))
297 #define vm_page_trylock(m) mtx_trylock(vm_page_lockptr((m)))
298 #endif
299 #if defined(INVARIANTS)
300 #define vm_page_assert_locked(m) \
301 vm_page_assert_locked_KBI((m), __FILE__, __LINE__)
302 #define vm_page_lock_assert(m, a) \
303 vm_page_lock_assert_KBI((m), (a), __FILE__, __LINE__)
304 #else
305 #define vm_page_assert_locked(m)
306 #define vm_page_lock_assert(m, a)
307 #endif
308
309 /*
310 * The vm_page's aflags are updated using atomic operations. To set or clear
311 * these flags, the functions vm_page_aflag_set() and vm_page_aflag_clear()
312 * must be used. Neither these flags nor these functions are part of the KBI.
313 *
314 * PGA_REFERENCED may be cleared only if the page is locked. It is set by
315 * both the MI and MD VM layers. However, kernel loadable modules should not
316 * directly set this flag. They should call vm_page_reference() instead.
317 *
318 * PGA_WRITEABLE is set exclusively on managed pages by pmap_enter().
319 * When it does so, the object must be locked, or the page must be
320 * exclusive busied. The MI VM layer must never access this flag
321 * directly. Instead, it should call pmap_page_is_write_mapped().
322 *
323 * PGA_EXECUTABLE may be set by pmap routines, and indicates that a page has
324 * at least one executable mapping. It is not consumed by the MI VM layer.
325 */
326 #define PGA_WRITEABLE 0x01 /* page may be mapped writeable */
327 #define PGA_REFERENCED 0x02 /* page has been referenced */
328 #define PGA_EXECUTABLE 0x04 /* page may be mapped executable */
329
330 /*
331 * Page flags. If changed at any other time than page allocation or
332 * freeing, the modification must be protected by the vm_page lock.
333 */
334 #define PG_PAQUEUE 0x0001 /* page has inactivation pending */
335 #define PG_FICTITIOUS 0x0004 /* physical page doesn't exist */
336 #define PG_ZERO 0x0008 /* page is zeroed */
337 #define PG_MARKER 0x0010 /* special queue marker page */
338 #define PG_WINATCFLS 0x0040 /* flush dirty page on inactive q */
339 #define PG_NODUMP 0x0080 /* don't include this page in a dump */
340 #define PG_UNHOLDFREE 0x0100 /* delayed free of a held page */
341 #define PG_ATHEAD 0x0200 /* put page at head of inactive list after deferred handling */
342
343 /*
344 * Misc constants.
345 */
346 #define ACT_DECLINE 1
347 #define ACT_ADVANCE 3
348 #define ACT_INIT 5
349 #define ACT_MAX 64
350
351 #ifdef _KERNEL
352
353 #include <sys/systm.h>
354
355 #include <machine/atomic.h>
356
357 /*
358 * Each pageable resident page falls into one of four lists:
359 *
360 * free
361 * Available for allocation now.
362 *
363 * The following lists are LRU sorted:
364 *
365 * inactive
366 * Low activity, candidates for reclamation.
367 * This is the list of pages that should be
368 * paged out next.
369 *
370 * active
371 * Pages that are "active" i.e. they have been
372 * recently referenced.
373 *
374 */
375
376 extern int vm_page_zero_count;
377
378 extern vm_page_t vm_page_array; /* First resident page in table */
379 extern long vm_page_array_size; /* number of vm_page_t's */
380 extern long first_page; /* first physical page number */
381
382 #define VM_PAGE_TO_PHYS(entry) ((entry)->phys_addr)
383
384 /*
385 * PHYS_TO_VM_PAGE() returns the vm_page_t object that represents a memory
386 * page to which the given physical address belongs. The correct vm_page_t
387 * object is returned for addresses that are not page-aligned.
388 */
389 vm_page_t PHYS_TO_VM_PAGE(vm_paddr_t pa);
390
391 /*
392 * Page allocation parameters for vm_page for the functions
393 * vm_page_alloc(), vm_page_grab(), vm_page_alloc_contig() and
394 * vm_page_alloc_freelist(). Some functions support only a subset
395 * of the flags, and ignore others, see the flags legend.
396 *
397 * Bits 0 - 1 define class.
398 * Bits 2 - 15 dedicated for flags.
399 * Legend:
400 * (a) - vm_page_alloc() supports the flag.
401 * (c) - vm_page_alloc_contig() supports the flag.
402 * (f) - vm_page_alloc_freelist() supports the flag.
403 * (g) - vm_page_grab() supports the flag.
404 * Bits above 15 define the count of additional pages that the caller
405 * intends to allocate.
406 */
407 #define VM_ALLOC_NORMAL 0
408 #define VM_ALLOC_INTERRUPT 1
409 #define VM_ALLOC_SYSTEM 2
410 #define VM_ALLOC_CLASS_MASK 3
411 #define VM_ALLOC_WIRED 0x0020 /* (acfg) Allocate non pageable page */
412 #define VM_ALLOC_ZERO 0x0040 /* (acfg) Try to obtain a zeroed page */
413 #define VM_ALLOC_NOOBJ 0x0100 /* (acg) No associated object */
414 #define VM_ALLOC_NOBUSY 0x0200 /* (acg) Do not busy the page */
415 #define VM_ALLOC_UNUSED11 0x0400 /* unused */
416 #define VM_ALLOC_UNUSED12 0x0800 /* unused */
417 #define VM_ALLOC_IGN_SBUSY 0x1000 /* (g) Ignore shared busy flag */
418 #define VM_ALLOC_NODUMP 0x2000 /* (ag) don't include in dump */
419 #define VM_ALLOC_SBUSY 0x4000 /* (acg) Shared busy the page */
420 #define VM_ALLOC_NOWAIT 0x8000 /* (g) Do not sleep, return NULL */
421 #define VM_ALLOC_COUNT_SHIFT 16
422 #define VM_ALLOC_COUNT(count) ((count) << VM_ALLOC_COUNT_SHIFT)
423
424 #ifdef M_NOWAIT
425 static inline int
malloc2vm_flags(int malloc_flags)426 malloc2vm_flags(int malloc_flags)
427 {
428 int pflags;
429
430 KASSERT((malloc_flags & M_USE_RESERVE) == 0 ||
431 (malloc_flags & M_NOWAIT) != 0,
432 ("M_USE_RESERVE requires M_NOWAIT"));
433 pflags = (malloc_flags & M_USE_RESERVE) != 0 ? VM_ALLOC_INTERRUPT :
434 VM_ALLOC_SYSTEM;
435 if ((malloc_flags & M_ZERO) != 0)
436 pflags |= VM_ALLOC_ZERO;
437 if ((malloc_flags & M_NODUMP) != 0)
438 pflags |= VM_ALLOC_NODUMP;
439 return (pflags);
440 }
441 #endif
442
443 void vm_page_busy_downgrade(vm_page_t m);
444 void vm_page_busy_sleep(vm_page_t m, const char *msg);
445 void vm_page_flash(vm_page_t m);
446 void vm_page_hold(vm_page_t mem);
447 void vm_page_unhold(vm_page_t mem);
448 void vm_page_free(vm_page_t m);
449 void vm_page_free_zero(vm_page_t m);
450
451 void vm_page_activate (vm_page_t);
452 void vm_page_advise(vm_page_t m, int advice);
453 vm_page_t vm_page_alloc (vm_object_t, vm_pindex_t, int);
454 vm_page_t vm_page_alloc_contig(vm_object_t object, vm_pindex_t pindex, int req,
455 u_long npages, vm_paddr_t low, vm_paddr_t high, u_long alignment,
456 vm_paddr_t boundary, vm_memattr_t memattr);
457 vm_page_t vm_page_alloc_freelist(int, int);
458 vm_page_t vm_page_grab (vm_object_t, vm_pindex_t, int);
459 int vm_page_try_to_free (vm_page_t);
460 void vm_page_deactivate (vm_page_t);
461 void vm_page_deactivate_noreuse(vm_page_t);
462 void vm_page_dequeue(vm_page_t m);
463 void vm_page_dequeue_locked(vm_page_t m);
464 vm_page_t vm_page_find_least(vm_object_t, vm_pindex_t);
465 vm_page_t vm_page_getfake(vm_paddr_t paddr, vm_memattr_t memattr);
466 void vm_page_initfake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr);
467 int vm_page_insert (vm_page_t, vm_object_t, vm_pindex_t);
468 vm_page_t vm_page_lookup (vm_object_t, vm_pindex_t);
469 vm_page_t vm_page_next(vm_page_t m);
470 int vm_page_pa_tryrelock(pmap_t, vm_paddr_t, vm_paddr_t *);
471 struct vm_pagequeue *vm_page_pagequeue(vm_page_t m);
472 struct vm_pagequeue *vm_page_pagequeue_deferred(vm_page_t m);
473 vm_page_t vm_page_prev(vm_page_t m);
474 boolean_t vm_page_ps_is_valid(vm_page_t m);
475 void vm_page_putfake(vm_page_t m);
476 void vm_page_readahead_finish(vm_page_t m);
477 bool vm_page_reclaim_contig(int req, u_long npages, vm_paddr_t low,
478 vm_paddr_t high, u_long alignment, vm_paddr_t boundary);
479 void vm_page_reference(vm_page_t m);
480 void vm_page_remove (vm_page_t);
481 int vm_page_rename (vm_page_t, vm_object_t, vm_pindex_t);
482 vm_page_t vm_page_replace(vm_page_t mnew, vm_object_t object,
483 vm_pindex_t pindex);
484 void vm_page_requeue(vm_page_t m);
485 void vm_page_requeue_locked(vm_page_t m);
486 int vm_page_sbusied(vm_page_t m);
487 vm_page_t vm_page_scan_contig(u_long npages, vm_page_t m_start,
488 vm_page_t m_end, u_long alignment, vm_paddr_t boundary, int options);
489 void vm_page_set_valid_range(vm_page_t m, int base, int size);
490 int vm_page_sleep_if_busy(vm_page_t m, const char *msg);
491 vm_offset_t vm_page_startup(vm_offset_t vaddr);
492 void vm_page_sunbusy(vm_page_t m);
493 int vm_page_trysbusy(vm_page_t m);
494 void vm_page_unhold_pages(vm_page_t *ma, int count);
495 boolean_t vm_page_unwire(vm_page_t m, uint8_t queue);
496 void vm_page_updatefake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr);
497 void vm_page_wire (vm_page_t);
498 void vm_page_xunbusy_hard(vm_page_t m);
499 void vm_page_set_validclean (vm_page_t, int, int);
500 void vm_page_clear_dirty (vm_page_t, int, int);
501 void vm_page_set_invalid (vm_page_t, int, int);
502 int vm_page_is_valid (vm_page_t, int, int);
503 void vm_page_test_dirty (vm_page_t);
504 vm_page_bits_t vm_page_bits(int base, int size);
505 void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid);
506 void vm_page_free_toq(vm_page_t m);
507 void vm_page_zero_idle_wakeup(void);
508
509 int vm_page_queue_fixup(struct vm_page *m);
510 int vm_page_queue_fixup_locked(struct vm_page *m);
511 void vm_page_dirty_KBI(vm_page_t m);
512 void vm_page_lock_KBI(vm_page_t m, const char *file, int line);
513 void vm_page_unlock_KBI(vm_page_t m, const char *file, int line);
514 int vm_page_trylock_KBI(vm_page_t m, const char *file, int line);
515 #if defined(INVARIANTS) || defined(INVARIANT_SUPPORT)
516 void vm_page_assert_locked_KBI(vm_page_t m, const char *file, int line);
517 void vm_page_lock_assert_KBI(vm_page_t m, int a, const char *file, int line);
518 #endif
519
520 #define vm_page_assert_sbusied(m) \
521 KASSERT(vm_page_sbusied(m), \
522 ("vm_page_assert_sbusied: page %p not shared busy @ %s:%d", \
523 (m), __FILE__, __LINE__))
524
525 #define vm_page_assert_unbusied(m) \
526 KASSERT(!vm_page_busied(m), \
527 ("vm_page_assert_unbusied: page %p busy @ %s:%d", \
528 (m), __FILE__, __LINE__))
529
530 #define vm_page_assert_xbusied(m) \
531 KASSERT(vm_page_xbusied(m), \
532 ("vm_page_assert_xbusied: page %p not exclusive busy @ %s:%d", \
533 (m), __FILE__, __LINE__))
534
535 #define vm_page_busied(m) \
536 ((m)->busy_lock != VPB_UNBUSIED)
537
538 #define vm_page_sbusy(m) do { \
539 if (!vm_page_trysbusy(m)) \
540 panic("%s: page %p failed shared busying", __func__, \
541 (m)); \
542 } while (0)
543
544 #define vm_page_tryxbusy(m) \
545 (atomic_cmpset_acq_int(&(m)->busy_lock, VPB_UNBUSIED, \
546 VPB_SINGLE_EXCLUSIVER))
547
548 #define vm_page_xbusied(m) \
549 (((m)->busy_lock & VPB_SINGLE_EXCLUSIVER) != 0)
550
551 #define vm_page_xbusy(m) do { \
552 if (!vm_page_tryxbusy(m)) \
553 panic("%s: page %p failed exclusive busying", __func__, \
554 (m)); \
555 } while (0)
556
557 #define vm_page_xunbusy(m) do { \
558 if (!atomic_cmpset_rel_int(&(m)->busy_lock, \
559 VPB_SINGLE_EXCLUSIVER, VPB_UNBUSIED)) \
560 vm_page_xunbusy_hard(m); \
561 } while (0)
562
563 #ifdef INVARIANTS
564 void vm_page_object_lock_assert(vm_page_t m);
565 #define VM_PAGE_OBJECT_LOCK_ASSERT(m) vm_page_object_lock_assert(m)
566 void vm_page_assert_pga_writeable(vm_page_t m, uint8_t bits);
567 #define VM_PAGE_ASSERT_PGA_WRITEABLE(m, bits) \
568 vm_page_assert_pga_writeable(m, bits)
569 #else
570 #define VM_PAGE_OBJECT_LOCK_ASSERT(m) (void)0
571 #define VM_PAGE_ASSERT_PGA_WRITEABLE(m, bits) (void)0
572 #endif
573
574 /*
575 * We want to use atomic updates for the aflags field, which is 8 bits wide.
576 * However, not all architectures support atomic operations on 8-bit
577 * destinations. In order that we can easily use a 32-bit operation, we
578 * require that the aflags field be 32-bit aligned.
579 */
580 CTASSERT(offsetof(struct vm_page, aflags) % sizeof(uint32_t) == 0);
581
582 /*
583 * Clear the given bits in the specified page.
584 */
585 static inline void
vm_page_aflag_clear(vm_page_t m,uint8_t bits)586 vm_page_aflag_clear(vm_page_t m, uint8_t bits)
587 {
588 uint32_t *addr, val;
589
590 /*
591 * The PGA_REFERENCED flag can only be cleared if the page is locked.
592 */
593 if ((bits & PGA_REFERENCED) != 0)
594 vm_page_assert_locked(m);
595
596 /*
597 * Access the whole 32-bit word containing the aflags field with an
598 * atomic update. Parallel non-atomic updates to the other fields
599 * within this word are handled properly by the atomic update.
600 */
601 addr = (void *)&m->aflags;
602 KASSERT(((uintptr_t)addr & (sizeof(uint32_t) - 1)) == 0,
603 ("vm_page_aflag_clear: aflags is misaligned"));
604 val = bits;
605 #if BYTE_ORDER == BIG_ENDIAN
606 val <<= 24;
607 #endif
608 atomic_clear_32(addr, val);
609 }
610
611 /*
612 * Set the given bits in the specified page.
613 */
614 static inline void
vm_page_aflag_set(vm_page_t m,uint8_t bits)615 vm_page_aflag_set(vm_page_t m, uint8_t bits)
616 {
617 uint32_t *addr, val;
618
619 VM_PAGE_ASSERT_PGA_WRITEABLE(m, bits);
620
621 /*
622 * Access the whole 32-bit word containing the aflags field with an
623 * atomic update. Parallel non-atomic updates to the other fields
624 * within this word are handled properly by the atomic update.
625 */
626 addr = (void *)&m->aflags;
627 KASSERT(((uintptr_t)addr & (sizeof(uint32_t) - 1)) == 0,
628 ("vm_page_aflag_set: aflags is misaligned"));
629 val = bits;
630 #if BYTE_ORDER == BIG_ENDIAN
631 val <<= 24;
632 #endif
633 atomic_set_32(addr, val);
634 }
635
636 /*
637 * vm_page_dirty:
638 *
639 * Set all bits in the page's dirty field.
640 *
641 * The object containing the specified page must be locked if the
642 * call is made from the machine-independent layer.
643 *
644 * See vm_page_clear_dirty_mask().
645 */
646 static __inline void
vm_page_dirty(vm_page_t m)647 vm_page_dirty(vm_page_t m)
648 {
649
650 /* Use vm_page_dirty_KBI() under INVARIANTS to save memory. */
651 #if defined(KLD_MODULE) || defined(INVARIANTS)
652 vm_page_dirty_KBI(m);
653 #else
654 m->dirty = VM_PAGE_BITS_ALL;
655 #endif
656 }
657
658 /*
659 * vm_page_remque:
660 *
661 * If the given page is in a page queue, then remove it from that page
662 * queue.
663 *
664 * The page must be locked.
665 */
666 static inline void
vm_page_remque(vm_page_t m)667 vm_page_remque(vm_page_t m)
668 {
669
670 if (m->queue != PQ_NONE)
671 vm_page_dequeue(m);
672 }
673
674 /*
675 * vm_page_undirty:
676 *
677 * Set page to not be dirty. Note: does not clear pmap modify bits
678 */
679 static __inline void
vm_page_undirty(vm_page_t m)680 vm_page_undirty(vm_page_t m)
681 {
682
683 VM_PAGE_OBJECT_LOCK_ASSERT(m);
684 m->dirty = 0;
685 }
686
687 static inline void
vm_page_replace_checked(vm_page_t mnew,vm_object_t object,vm_pindex_t pindex,vm_page_t mold)688 vm_page_replace_checked(vm_page_t mnew, vm_object_t object, vm_pindex_t pindex,
689 vm_page_t mold)
690 {
691 vm_page_t mret;
692
693 mret = vm_page_replace(mnew, object, pindex);
694 KASSERT(mret == mold,
695 ("invalid page replacement, mold=%p, mret=%p", mold, mret));
696
697 /* Unused if !INVARIANTS. */
698 (void)mold;
699 (void)mret;
700 }
701
702 #endif /* _KERNEL */
703 #endif /* !_VM_PAGE_ */
704