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_page.h 8.2 (Berkeley) 12/13/93 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 * Resident memory system definitions. 67 */ 68 69 #ifndef _VM_PAGE_ 70 #define _VM_PAGE_ 71 72 #include <vm/pmap.h> 73 74 /* 75 * Management of resident (logical) pages. 76 * 77 * A small structure is kept for each resident 78 * page, indexed by page number. Each structure 79 * is an element of several collections: 80 * 81 * A radix tree used to quickly 82 * perform object/offset lookups 83 * 84 * A list of all pages for a given object, 85 * so they can be quickly deactivated at 86 * time of deallocation. 87 * 88 * An ordered list of pages due for pageout. 89 * 90 * In addition, the structure contains the object 91 * and offset to which this page belongs (for pageout), 92 * and sundry status bits. 93 * 94 * In general, operations on this structure's mutable fields are 95 * synchronized using either one of or a combination of the lock on the 96 * object that the page belongs to (O), the page lock (P), 97 * the per-domain lock for the free queues (F), or the page's queue 98 * lock (Q). The physical address of a page is used to select its page 99 * lock from a pool. The queue lock for a page depends on the value of 100 * its queue field and described in detail below. If a field is 101 * annotated below with two of these locks, then holding either lock is 102 * sufficient for read access, but both locks are required for write 103 * access. An annotation of (C) indicates that the field is immutable. 104 * 105 * In contrast, the synchronization of accesses to the page's 106 * dirty field is machine dependent (M). In the 107 * machine-independent layer, the lock on the object that the 108 * page belongs to must be held in order to operate on the field. 109 * However, the pmap layer is permitted to set all bits within 110 * the field without holding that lock. If the underlying 111 * architecture does not support atomic read-modify-write 112 * operations on the field's type, then the machine-independent 113 * layer uses a 32-bit atomic on the aligned 32-bit word that 114 * contains the dirty field. In the machine-independent layer, 115 * the implementation of read-modify-write operations on the 116 * field is encapsulated in vm_page_clear_dirty_mask(). 117 * 118 * The page structure contains two counters which prevent page reuse. 119 * Both counters are protected by the page lock (P). The hold 120 * counter counts transient references obtained via a pmap lookup, and 121 * is also used to prevent page reclamation in situations where it is 122 * undesirable to block other accesses to the page. The wire counter 123 * is used to implement mlock(2) and is non-zero for pages containing 124 * kernel memory. Pages that are wired or held will not be reclaimed 125 * or laundered by the page daemon, but are treated differently during 126 * a page queue scan: held pages remain at their position in the queue, 127 * while wired pages are removed from the queue and must later be 128 * re-enqueued appropriately by the unwiring thread. It is legal to 129 * call vm_page_free() on a held page; doing so causes it to be removed 130 * from its object and page queue, and the page is released to the 131 * allocator once the last hold reference is dropped. In contrast, 132 * wired pages may not be freed. 133 * 134 * In some pmap implementations, the wire count of a page table page is 135 * used to track the number of populated entries. 136 * 137 * The busy lock is an embedded reader-writer lock which protects the 138 * page's contents and identity (i.e., its <object, pindex> tuple) and 139 * interlocks with the object lock (O). In particular, a page may be 140 * busied or unbusied only with the object write lock held. To avoid 141 * bloating the page structure, the busy lock lacks some of the 142 * features available to the kernel's general-purpose synchronization 143 * primitives. As a result, busy lock ordering rules are not verified, 144 * lock recursion is not detected, and an attempt to xbusy a busy page 145 * or sbusy an xbusy page results will trigger a panic rather than 146 * causing the thread to block. vm_page_sleep_if_busy() can be used to 147 * sleep until the page's busy state changes, after which the caller 148 * must re-lookup the page and re-evaluate its state. 149 * 150 * The queue field is the index of the page queue containing the 151 * page, or PQ_NONE if the page is not enqueued. The queue lock of a 152 * page is the page queue lock corresponding to the page queue index, 153 * or the page lock (P) for the page if it is not enqueued. To modify 154 * the queue field, the queue lock for the old value of the field must 155 * be held. It is invalid for a page's queue field to transition 156 * between two distinct page queue indices. That is, when updating 157 * the queue field, either the new value or the old value must be 158 * PQ_NONE. 159 * 160 * To avoid contention on page queue locks, page queue operations 161 * (enqueue, dequeue, requeue) are batched using per-CPU queues. 162 * A deferred operation is requested by inserting an entry into a 163 * batch queue; the entry is simply a pointer to the page, and the 164 * request type is encoded in the page's aflags field using the values 165 * in PGA_QUEUE_STATE_MASK. The type-stability of struct vm_pages is 166 * crucial to this scheme since the processing of entries in a given 167 * batch queue may be deferred indefinitely. In particular, a page 168 * may be freed before its pending batch queue entries have been 169 * processed. The page lock (P) must be held to schedule a batched 170 * queue operation, and the page queue lock must be held in order to 171 * process batch queue entries for the page queue. 172 */ 173 174 #if PAGE_SIZE == 4096 175 #define VM_PAGE_BITS_ALL 0xffu 176 typedef uint8_t vm_page_bits_t; 177 #elif PAGE_SIZE == 8192 178 #define VM_PAGE_BITS_ALL 0xffffu 179 typedef uint16_t vm_page_bits_t; 180 #elif PAGE_SIZE == 16384 181 #define VM_PAGE_BITS_ALL 0xffffffffu 182 typedef uint32_t vm_page_bits_t; 183 #elif PAGE_SIZE == 32768 184 #define VM_PAGE_BITS_ALL 0xfffffffffffffffflu 185 typedef uint64_t vm_page_bits_t; 186 #endif 187 188 struct vm_page { 189 union { 190 TAILQ_ENTRY(vm_page) q; /* page queue or free list (Q) */ 191 struct { 192 SLIST_ENTRY(vm_page) ss; /* private slists */ 193 void *pv; 194 } s; 195 struct { 196 u_long p; 197 u_long v; 198 } memguard; 199 } plinks; 200 TAILQ_ENTRY(vm_page) listq; /* pages in same object (O) */ 201 vm_object_t object; /* which object am I in (O,P) */ 202 vm_pindex_t pindex; /* offset into object (O,P) */ 203 vm_paddr_t phys_addr; /* physical address of page (C) */ 204 struct md_page md; /* machine dependent stuff */ 205 u_int wire_count; /* wired down maps refs (P) */ 206 volatile u_int busy_lock; /* busy owners lock */ 207 uint16_t flags; /* page PG_* flags (P) */ 208 uint8_t order; /* index of the buddy queue (F) */ 209 uint8_t pool; /* vm_phys freepool index (F) */ 210 uint8_t aflags; /* access is atomic */ 211 uint8_t oflags; /* page VPO_* flags (O) */ 212 uint8_t queue; /* page queue index (Q) */ 213 int8_t psind; /* pagesizes[] index (O) */ 214 int8_t segind; /* vm_phys segment index (C) */ 215 u_char act_count; /* page usage count (P) */ 216 /* NOTE that these must support one bit per DEV_BSIZE in a page */ 217 /* so, on normal X86 kernels, they must be at least 8 bits wide */ 218 vm_page_bits_t valid; /* map of valid DEV_BSIZE chunks (O) */ 219 vm_page_bits_t dirty; /* map of dirty DEV_BSIZE chunks (M) */ 220 }; 221 222 /* 223 * Page flags stored in oflags: 224 * 225 * Access to these page flags is synchronized by the lock on the object 226 * containing the page (O). 227 * 228 * Note: VPO_UNMANAGED (used by OBJT_DEVICE, OBJT_PHYS and OBJT_SG) 229 * indicates that the page is not under PV management but 230 * otherwise should be treated as a normal page. Pages not 231 * under PV management cannot be paged out via the 232 * object/vm_page_t because there is no knowledge of their pte 233 * mappings, and such pages are also not on any PQ queue. 234 * 235 */ 236 #define VPO_KMEM_EXEC 0x01 /* kmem mapping allows execution */ 237 #define VPO_SWAPSLEEP 0x02 /* waiting for swap to finish */ 238 #define VPO_UNMANAGED 0x04 /* no PV management for page */ 239 #define VPO_SWAPINPROG 0x08 /* swap I/O in progress on page */ 240 #define VPO_NOSYNC 0x10 /* do not collect for syncer */ 241 242 /* 243 * Busy page implementation details. 244 * The algorithm is taken mostly by rwlock(9) and sx(9) locks implementation, 245 * even if the support for owner identity is removed because of size 246 * constraints. Checks on lock recursion are then not possible, while the 247 * lock assertions effectiveness is someway reduced. 248 */ 249 #define VPB_BIT_SHARED 0x01 250 #define VPB_BIT_EXCLUSIVE 0x02 251 #define VPB_BIT_WAITERS 0x04 252 #define VPB_BIT_FLAGMASK \ 253 (VPB_BIT_SHARED | VPB_BIT_EXCLUSIVE | VPB_BIT_WAITERS) 254 255 #define VPB_SHARERS_SHIFT 3 256 #define VPB_SHARERS(x) \ 257 (((x) & ~VPB_BIT_FLAGMASK) >> VPB_SHARERS_SHIFT) 258 #define VPB_SHARERS_WORD(x) ((x) << VPB_SHARERS_SHIFT | VPB_BIT_SHARED) 259 #define VPB_ONE_SHARER (1 << VPB_SHARERS_SHIFT) 260 261 #define VPB_SINGLE_EXCLUSIVER VPB_BIT_EXCLUSIVE 262 263 #define VPB_UNBUSIED VPB_SHARERS_WORD(0) 264 265 #define PQ_NONE 255 266 #define PQ_INACTIVE 0 267 #define PQ_ACTIVE 1 268 #define PQ_LAUNDRY 2 269 #define PQ_UNSWAPPABLE 3 270 #define PQ_COUNT 4 271 272 #ifndef VM_PAGE_HAVE_PGLIST 273 TAILQ_HEAD(pglist, vm_page); 274 #define VM_PAGE_HAVE_PGLIST 275 #endif 276 SLIST_HEAD(spglist, vm_page); 277 278 #ifdef _KERNEL 279 extern vm_page_t bogus_page; 280 #endif /* _KERNEL */ 281 282 extern struct mtx_padalign pa_lock[]; 283 284 #if defined(__arm__) 285 #define PDRSHIFT PDR_SHIFT 286 #elif !defined(PDRSHIFT) 287 #define PDRSHIFT 21 288 #endif 289 290 #define pa_index(pa) ((pa) >> PDRSHIFT) 291 #define PA_LOCKPTR(pa) ((struct mtx *)(&pa_lock[pa_index(pa) % PA_LOCK_COUNT])) 292 #define PA_LOCKOBJPTR(pa) ((struct lock_object *)PA_LOCKPTR((pa))) 293 #define PA_LOCK(pa) mtx_lock(PA_LOCKPTR(pa)) 294 #define PA_TRYLOCK(pa) mtx_trylock(PA_LOCKPTR(pa)) 295 #define PA_UNLOCK(pa) mtx_unlock(PA_LOCKPTR(pa)) 296 #define PA_UNLOCK_COND(pa) \ 297 do { \ 298 if ((pa) != 0) { \ 299 PA_UNLOCK((pa)); \ 300 (pa) = 0; \ 301 } \ 302 } while (0) 303 304 #define PA_LOCK_ASSERT(pa, a) mtx_assert(PA_LOCKPTR(pa), (a)) 305 306 #if defined(KLD_MODULE) && !defined(KLD_TIED) 307 #define vm_page_lock(m) vm_page_lock_KBI((m), LOCK_FILE, LOCK_LINE) 308 #define vm_page_unlock(m) vm_page_unlock_KBI((m), LOCK_FILE, LOCK_LINE) 309 #define vm_page_trylock(m) vm_page_trylock_KBI((m), LOCK_FILE, LOCK_LINE) 310 #else /* !KLD_MODULE */ 311 #define vm_page_lockptr(m) (PA_LOCKPTR(VM_PAGE_TO_PHYS((m)))) 312 #define vm_page_lock(m) mtx_lock(vm_page_lockptr((m))) 313 #define vm_page_unlock(m) mtx_unlock(vm_page_lockptr((m))) 314 #define vm_page_trylock(m) mtx_trylock(vm_page_lockptr((m))) 315 #endif 316 #if defined(INVARIANTS) 317 #define vm_page_assert_locked(m) \ 318 vm_page_assert_locked_KBI((m), __FILE__, __LINE__) 319 #define vm_page_lock_assert(m, a) \ 320 vm_page_lock_assert_KBI((m), (a), __FILE__, __LINE__) 321 #else 322 #define vm_page_assert_locked(m) 323 #define vm_page_lock_assert(m, a) 324 #endif 325 326 /* 327 * The vm_page's aflags are updated using atomic operations. To set or clear 328 * these flags, the functions vm_page_aflag_set() and vm_page_aflag_clear() 329 * must be used. Neither these flags nor these functions are part of the KBI. 330 * 331 * PGA_REFERENCED may be cleared only if the page is locked. It is set by 332 * both the MI and MD VM layers. However, kernel loadable modules should not 333 * directly set this flag. They should call vm_page_reference() instead. 334 * 335 * PGA_WRITEABLE is set exclusively on managed pages by pmap_enter(). 336 * When it does so, the object must be locked, or the page must be 337 * exclusive busied. The MI VM layer must never access this flag 338 * directly. Instead, it should call pmap_page_is_write_mapped(). 339 * 340 * PGA_EXECUTABLE may be set by pmap routines, and indicates that a page has 341 * at least one executable mapping. It is not consumed by the MI VM layer. 342 * 343 * PGA_ENQUEUED is set and cleared when a page is inserted into or removed 344 * from a page queue, respectively. It determines whether the plinks.q field 345 * of the page is valid. To set or clear this flag, the queue lock for the 346 * page must be held: the page queue lock corresponding to the page's "queue" 347 * field if its value is not PQ_NONE, and the page lock otherwise. 348 * 349 * PGA_DEQUEUE is set when the page is scheduled to be dequeued from a page 350 * queue, and cleared when the dequeue request is processed. A page may 351 * have PGA_DEQUEUE set and PGA_ENQUEUED cleared, for instance if a dequeue 352 * is requested after the page is scheduled to be enqueued but before it is 353 * actually inserted into the page queue. For allocated pages, the page lock 354 * must be held to set this flag, but it may be set by vm_page_free_prep() 355 * without the page lock held. The page queue lock must be held to clear the 356 * PGA_DEQUEUE flag. 357 * 358 * PGA_REQUEUE is set when the page is scheduled to be enqueued or requeued 359 * in its page queue. The page lock must be held to set this flag, and the 360 * queue lock for the page must be held to clear it. 361 * 362 * PGA_REQUEUE_HEAD is a special flag for enqueuing pages near the head of 363 * the inactive queue, thus bypassing LRU. The page lock must be held to 364 * set this flag, and the queue lock for the page must be held to clear it. 365 */ 366 #define PGA_WRITEABLE 0x01 /* page may be mapped writeable */ 367 #define PGA_REFERENCED 0x02 /* page has been referenced */ 368 #define PGA_EXECUTABLE 0x04 /* page may be mapped executable */ 369 #define PGA_ENQUEUED 0x08 /* page is enqueued in a page queue */ 370 #define PGA_DEQUEUE 0x10 /* page is due to be dequeued */ 371 #define PGA_REQUEUE 0x20 /* page is due to be requeued */ 372 #define PGA_REQUEUE_HEAD 0x40 /* page requeue should bypass LRU */ 373 374 #define PGA_QUEUE_STATE_MASK (PGA_ENQUEUED | PGA_DEQUEUE | PGA_REQUEUE | \ 375 PGA_REQUEUE_HEAD) 376 377 /* 378 * Page flags. If changed at any other time than page allocation or 379 * freeing, the modification must be protected by the vm_page lock. 380 * 381 * The PG_PCPU_CACHE flag is set at allocation time if the page was 382 * allocated from a per-CPU cache. It is cleared the next time that the 383 * page is allocated from the physical memory allocator. 384 */ 385 #define PG_PCPU_CACHE 0x0001 /* was allocated from per-CPU caches */ 386 #define PG_FICTITIOUS 0x0004 /* physical page doesn't exist */ 387 #define PG_ZERO 0x0008 /* page is zeroed */ 388 #define PG_MARKER 0x0010 /* special queue marker page */ 389 #define PG_NODUMP 0x0080 /* don't include this page in a dump */ 390 391 /* 392 * Misc constants. 393 */ 394 #define ACT_DECLINE 1 395 #define ACT_ADVANCE 3 396 #define ACT_INIT 5 397 #define ACT_MAX 64 398 399 #ifdef _KERNEL 400 401 #include <sys/systm.h> 402 403 #include <machine/atomic.h> 404 405 /* 406 * Each pageable resident page falls into one of five lists: 407 * 408 * free 409 * Available for allocation now. 410 * 411 * inactive 412 * Low activity, candidates for reclamation. 413 * This list is approximately LRU ordered. 414 * 415 * laundry 416 * This is the list of pages that should be 417 * paged out next. 418 * 419 * unswappable 420 * Dirty anonymous pages that cannot be paged 421 * out because no swap device is configured. 422 * 423 * active 424 * Pages that are "active", i.e., they have been 425 * recently referenced. 426 * 427 */ 428 429 extern vm_page_t vm_page_array; /* First resident page in table */ 430 extern long vm_page_array_size; /* number of vm_page_t's */ 431 extern long first_page; /* first physical page number */ 432 433 #define VM_PAGE_TO_PHYS(entry) ((entry)->phys_addr) 434 435 /* 436 * PHYS_TO_VM_PAGE() returns the vm_page_t object that represents a memory 437 * page to which the given physical address belongs. The correct vm_page_t 438 * object is returned for addresses that are not page-aligned. 439 */ 440 vm_page_t PHYS_TO_VM_PAGE(vm_paddr_t pa); 441 442 /* 443 * Page allocation parameters for vm_page for the functions 444 * vm_page_alloc(), vm_page_grab(), vm_page_alloc_contig() and 445 * vm_page_alloc_freelist(). Some functions support only a subset 446 * of the flags, and ignore others, see the flags legend. 447 * 448 * The meaning of VM_ALLOC_ZERO differs slightly between the vm_page_alloc*() 449 * and the vm_page_grab*() functions. See these functions for details. 450 * 451 * Bits 0 - 1 define class. 452 * Bits 2 - 15 dedicated for flags. 453 * Legend: 454 * (a) - vm_page_alloc() supports the flag. 455 * (c) - vm_page_alloc_contig() supports the flag. 456 * (f) - vm_page_alloc_freelist() supports the flag. 457 * (g) - vm_page_grab() supports the flag. 458 * (p) - vm_page_grab_pages() supports the flag. 459 * Bits above 15 define the count of additional pages that the caller 460 * intends to allocate. 461 */ 462 #define VM_ALLOC_NORMAL 0 463 #define VM_ALLOC_INTERRUPT 1 464 #define VM_ALLOC_SYSTEM 2 465 #define VM_ALLOC_CLASS_MASK 3 466 #define VM_ALLOC_WAITOK 0x0008 /* (acf) Sleep and retry */ 467 #define VM_ALLOC_WAITFAIL 0x0010 /* (acf) Sleep and return error */ 468 #define VM_ALLOC_WIRED 0x0020 /* (acfgp) Allocate a wired page */ 469 #define VM_ALLOC_ZERO 0x0040 /* (acfgp) Allocate a prezeroed page */ 470 #define VM_ALLOC_NOOBJ 0x0100 /* (acg) No associated object */ 471 #define VM_ALLOC_NOBUSY 0x0200 /* (acgp) Do not excl busy the page */ 472 #define VM_ALLOC_IGN_SBUSY 0x1000 /* (gp) Ignore shared busy flag */ 473 #define VM_ALLOC_NODUMP 0x2000 /* (ag) don't include in dump */ 474 #define VM_ALLOC_SBUSY 0x4000 /* (acgp) Shared busy the page */ 475 #define VM_ALLOC_NOWAIT 0x8000 /* (acfgp) Do not sleep */ 476 #define VM_ALLOC_COUNT_SHIFT 16 477 #define VM_ALLOC_COUNT(count) ((count) << VM_ALLOC_COUNT_SHIFT) 478 479 #ifdef M_NOWAIT 480 static inline int 481 malloc2vm_flags(int malloc_flags) 482 { 483 int pflags; 484 485 KASSERT((malloc_flags & M_USE_RESERVE) == 0 || 486 (malloc_flags & M_NOWAIT) != 0, 487 ("M_USE_RESERVE requires M_NOWAIT")); 488 pflags = (malloc_flags & M_USE_RESERVE) != 0 ? VM_ALLOC_INTERRUPT : 489 VM_ALLOC_SYSTEM; 490 if ((malloc_flags & M_ZERO) != 0) 491 pflags |= VM_ALLOC_ZERO; 492 if ((malloc_flags & M_NODUMP) != 0) 493 pflags |= VM_ALLOC_NODUMP; 494 if ((malloc_flags & M_NOWAIT)) 495 pflags |= VM_ALLOC_NOWAIT; 496 if ((malloc_flags & M_WAITOK)) 497 pflags |= VM_ALLOC_WAITOK; 498 return (pflags); 499 } 500 #endif 501 502 /* 503 * Predicates supported by vm_page_ps_test(): 504 * 505 * PS_ALL_DIRTY is true only if the entire (super)page is dirty. 506 * However, it can be spuriously false when the (super)page has become 507 * dirty in the pmap but that information has not been propagated to the 508 * machine-independent layer. 509 */ 510 #define PS_ALL_DIRTY 0x1 511 #define PS_ALL_VALID 0x2 512 #define PS_NONE_BUSY 0x4 513 514 void vm_page_busy_downgrade(vm_page_t m); 515 void vm_page_busy_sleep(vm_page_t m, const char *msg, bool nonshared); 516 void vm_page_flash(vm_page_t m); 517 void vm_page_free(vm_page_t m); 518 void vm_page_free_zero(vm_page_t m); 519 520 void vm_page_activate (vm_page_t); 521 void vm_page_advise(vm_page_t m, int advice); 522 vm_page_t vm_page_alloc(vm_object_t, vm_pindex_t, int); 523 vm_page_t vm_page_alloc_domain(vm_object_t, vm_pindex_t, int, int); 524 vm_page_t vm_page_alloc_after(vm_object_t, vm_pindex_t, int, vm_page_t); 525 vm_page_t vm_page_alloc_domain_after(vm_object_t, vm_pindex_t, int, int, 526 vm_page_t); 527 vm_page_t vm_page_alloc_contig(vm_object_t object, vm_pindex_t pindex, int req, 528 u_long npages, vm_paddr_t low, vm_paddr_t high, u_long alignment, 529 vm_paddr_t boundary, vm_memattr_t memattr); 530 vm_page_t vm_page_alloc_contig_domain(vm_object_t object, 531 vm_pindex_t pindex, int domain, int req, u_long npages, vm_paddr_t low, 532 vm_paddr_t high, u_long alignment, vm_paddr_t boundary, 533 vm_memattr_t memattr); 534 vm_page_t vm_page_alloc_freelist(int, int); 535 vm_page_t vm_page_alloc_freelist_domain(int, int, int); 536 bool vm_page_blacklist_add(vm_paddr_t pa, bool verbose); 537 void vm_page_change_lock(vm_page_t m, struct mtx **mtx); 538 vm_page_t vm_page_grab (vm_object_t, vm_pindex_t, int); 539 int vm_page_grab_pages(vm_object_t object, vm_pindex_t pindex, int allocflags, 540 vm_page_t *ma, int count); 541 void vm_page_deactivate(vm_page_t); 542 void vm_page_deactivate_noreuse(vm_page_t); 543 void vm_page_dequeue(vm_page_t m); 544 void vm_page_dequeue_deferred(vm_page_t m); 545 void vm_page_drain_pqbatch(void); 546 vm_page_t vm_page_find_least(vm_object_t, vm_pindex_t); 547 bool vm_page_free_prep(vm_page_t m); 548 vm_page_t vm_page_getfake(vm_paddr_t paddr, vm_memattr_t memattr); 549 void vm_page_initfake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr); 550 int vm_page_insert (vm_page_t, vm_object_t, vm_pindex_t); 551 void vm_page_launder(vm_page_t m); 552 vm_page_t vm_page_lookup (vm_object_t, vm_pindex_t); 553 vm_page_t vm_page_next(vm_page_t m); 554 int vm_page_pa_tryrelock(pmap_t, vm_paddr_t, vm_paddr_t *); 555 struct vm_pagequeue *vm_page_pagequeue(vm_page_t m); 556 vm_page_t vm_page_prev(vm_page_t m); 557 bool vm_page_ps_test(vm_page_t m, int flags, vm_page_t skip_m); 558 void vm_page_putfake(vm_page_t m); 559 void vm_page_readahead_finish(vm_page_t m); 560 bool vm_page_reclaim_contig(int req, u_long npages, vm_paddr_t low, 561 vm_paddr_t high, u_long alignment, vm_paddr_t boundary); 562 bool vm_page_reclaim_contig_domain(int domain, int req, u_long npages, 563 vm_paddr_t low, vm_paddr_t high, u_long alignment, vm_paddr_t boundary); 564 void vm_page_reference(vm_page_t m); 565 #define VPR_TRYFREE 0x01 566 #define VPR_NOREUSE 0x02 567 void vm_page_release(vm_page_t m, int flags); 568 void vm_page_release_locked(vm_page_t m, int flags); 569 bool vm_page_remove(vm_page_t); 570 int vm_page_rename(vm_page_t, vm_object_t, vm_pindex_t); 571 vm_page_t vm_page_replace(vm_page_t mnew, vm_object_t object, 572 vm_pindex_t pindex); 573 void vm_page_requeue(vm_page_t m); 574 int vm_page_sbusied(vm_page_t m); 575 vm_page_t vm_page_scan_contig(u_long npages, vm_page_t m_start, 576 vm_page_t m_end, u_long alignment, vm_paddr_t boundary, int options); 577 void vm_page_set_valid_range(vm_page_t m, int base, int size); 578 int vm_page_sleep_if_busy(vm_page_t m, const char *msg); 579 vm_offset_t vm_page_startup(vm_offset_t vaddr); 580 void vm_page_sunbusy(vm_page_t m); 581 int vm_page_trysbusy(vm_page_t m); 582 void vm_page_unhold_pages(vm_page_t *ma, int count); 583 void vm_page_unswappable(vm_page_t m); 584 bool vm_page_unwire(vm_page_t m, uint8_t queue); 585 bool vm_page_unwire_noq(vm_page_t m); 586 void vm_page_updatefake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr); 587 void vm_page_wire (vm_page_t); 588 void vm_page_xunbusy_hard(vm_page_t m); 589 void vm_page_xunbusy_maybelocked(vm_page_t m); 590 void vm_page_set_validclean (vm_page_t, int, int); 591 void vm_page_clear_dirty (vm_page_t, int, int); 592 void vm_page_set_invalid (vm_page_t, int, int); 593 int vm_page_is_valid (vm_page_t, int, int); 594 void vm_page_test_dirty (vm_page_t); 595 vm_page_bits_t vm_page_bits(int base, int size); 596 void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid); 597 void vm_page_free_toq(vm_page_t m); 598 void vm_page_free_pages_toq(struct spglist *free, bool update_wire_count); 599 600 void vm_page_dirty_KBI(vm_page_t m); 601 void vm_page_lock_KBI(vm_page_t m, const char *file, int line); 602 void vm_page_unlock_KBI(vm_page_t m, const char *file, int line); 603 int vm_page_trylock_KBI(vm_page_t m, const char *file, int line); 604 #if defined(INVARIANTS) || defined(INVARIANT_SUPPORT) 605 void vm_page_assert_locked_KBI(vm_page_t m, const char *file, int line); 606 void vm_page_lock_assert_KBI(vm_page_t m, int a, const char *file, int line); 607 #endif 608 609 #define vm_page_assert_sbusied(m) \ 610 KASSERT(vm_page_sbusied(m), \ 611 ("vm_page_assert_sbusied: page %p not shared busy @ %s:%d", \ 612 (m), __FILE__, __LINE__)) 613 614 #define vm_page_assert_unbusied(m) \ 615 KASSERT(!vm_page_busied(m), \ 616 ("vm_page_assert_unbusied: page %p busy @ %s:%d", \ 617 (m), __FILE__, __LINE__)) 618 619 #define vm_page_assert_xbusied(m) \ 620 KASSERT(vm_page_xbusied(m), \ 621 ("vm_page_assert_xbusied: page %p not exclusive busy @ %s:%d", \ 622 (m), __FILE__, __LINE__)) 623 624 #define vm_page_busied(m) \ 625 ((m)->busy_lock != VPB_UNBUSIED) 626 627 #define vm_page_sbusy(m) do { \ 628 if (!vm_page_trysbusy(m)) \ 629 panic("%s: page %p failed shared busying", __func__, \ 630 (m)); \ 631 } while (0) 632 633 #define vm_page_tryxbusy(m) \ 634 (atomic_cmpset_acq_int(&(m)->busy_lock, VPB_UNBUSIED, \ 635 VPB_SINGLE_EXCLUSIVER)) 636 637 #define vm_page_xbusied(m) \ 638 (((m)->busy_lock & VPB_SINGLE_EXCLUSIVER) != 0) 639 640 #define vm_page_xbusy(m) do { \ 641 if (!vm_page_tryxbusy(m)) \ 642 panic("%s: page %p failed exclusive busying", __func__, \ 643 (m)); \ 644 } while (0) 645 646 /* Note: page m's lock must not be owned by the caller. */ 647 #define vm_page_xunbusy(m) do { \ 648 if (!atomic_cmpset_rel_int(&(m)->busy_lock, \ 649 VPB_SINGLE_EXCLUSIVER, VPB_UNBUSIED)) \ 650 vm_page_xunbusy_hard(m); \ 651 } while (0) 652 653 #ifdef INVARIANTS 654 void vm_page_object_lock_assert(vm_page_t m); 655 #define VM_PAGE_OBJECT_LOCK_ASSERT(m) vm_page_object_lock_assert(m) 656 void vm_page_assert_pga_writeable(vm_page_t m, uint8_t bits); 657 #define VM_PAGE_ASSERT_PGA_WRITEABLE(m, bits) \ 658 vm_page_assert_pga_writeable(m, bits) 659 #else 660 #define VM_PAGE_OBJECT_LOCK_ASSERT(m) (void)0 661 #define VM_PAGE_ASSERT_PGA_WRITEABLE(m, bits) (void)0 662 #endif 663 664 /* 665 * We want to use atomic updates for the aflags field, which is 8 bits wide. 666 * However, not all architectures support atomic operations on 8-bit 667 * destinations. In order that we can easily use a 32-bit operation, we 668 * require that the aflags field be 32-bit aligned. 669 */ 670 CTASSERT(offsetof(struct vm_page, aflags) % sizeof(uint32_t) == 0); 671 672 /* 673 * Clear the given bits in the specified page. 674 */ 675 static inline void 676 vm_page_aflag_clear(vm_page_t m, uint8_t bits) 677 { 678 uint32_t *addr, val; 679 680 /* 681 * The PGA_REFERENCED flag can only be cleared if the page is locked. 682 */ 683 if ((bits & PGA_REFERENCED) != 0) 684 vm_page_assert_locked(m); 685 686 /* 687 * Access the whole 32-bit word containing the aflags field with an 688 * atomic update. Parallel non-atomic updates to the other fields 689 * within this word are handled properly by the atomic update. 690 */ 691 addr = (void *)&m->aflags; 692 KASSERT(((uintptr_t)addr & (sizeof(uint32_t) - 1)) == 0, 693 ("vm_page_aflag_clear: aflags is misaligned")); 694 val = bits; 695 #if BYTE_ORDER == BIG_ENDIAN 696 val <<= 24; 697 #endif 698 atomic_clear_32(addr, val); 699 } 700 701 /* 702 * Set the given bits in the specified page. 703 */ 704 static inline void 705 vm_page_aflag_set(vm_page_t m, uint8_t bits) 706 { 707 uint32_t *addr, val; 708 709 VM_PAGE_ASSERT_PGA_WRITEABLE(m, bits); 710 711 /* 712 * Access the whole 32-bit word containing the aflags field with an 713 * atomic update. Parallel non-atomic updates to the other fields 714 * within this word are handled properly by the atomic update. 715 */ 716 addr = (void *)&m->aflags; 717 KASSERT(((uintptr_t)addr & (sizeof(uint32_t) - 1)) == 0, 718 ("vm_page_aflag_set: aflags is misaligned")); 719 val = bits; 720 #if BYTE_ORDER == BIG_ENDIAN 721 val <<= 24; 722 #endif 723 atomic_set_32(addr, val); 724 } 725 726 /* 727 * vm_page_dirty: 728 * 729 * Set all bits in the page's dirty field. 730 * 731 * The object containing the specified page must be locked if the 732 * call is made from the machine-independent layer. 733 * 734 * See vm_page_clear_dirty_mask(). 735 */ 736 static __inline void 737 vm_page_dirty(vm_page_t m) 738 { 739 740 /* Use vm_page_dirty_KBI() under INVARIANTS to save memory. */ 741 #if (defined(KLD_MODULE) && !defined(KLD_TIED)) || defined(INVARIANTS) 742 vm_page_dirty_KBI(m); 743 #else 744 m->dirty = VM_PAGE_BITS_ALL; 745 #endif 746 } 747 748 /* 749 * vm_page_undirty: 750 * 751 * Set page to not be dirty. Note: does not clear pmap modify bits 752 */ 753 static __inline void 754 vm_page_undirty(vm_page_t m) 755 { 756 757 VM_PAGE_OBJECT_LOCK_ASSERT(m); 758 m->dirty = 0; 759 } 760 761 static inline void 762 vm_page_replace_checked(vm_page_t mnew, vm_object_t object, vm_pindex_t pindex, 763 vm_page_t mold) 764 { 765 vm_page_t mret; 766 767 mret = vm_page_replace(mnew, object, pindex); 768 KASSERT(mret == mold, 769 ("invalid page replacement, mold=%p, mret=%p", mold, mret)); 770 771 /* Unused if !INVARIANTS. */ 772 (void)mold; 773 (void)mret; 774 } 775 776 /* 777 * vm_page_queue: 778 * 779 * Return the index of the queue containing m. This index is guaranteed 780 * not to change while the page lock is held. 781 */ 782 static inline uint8_t 783 vm_page_queue(vm_page_t m) 784 { 785 786 vm_page_assert_locked(m); 787 788 if ((m->aflags & PGA_DEQUEUE) != 0) 789 return (PQ_NONE); 790 atomic_thread_fence_acq(); 791 return (m->queue); 792 } 793 794 static inline bool 795 vm_page_active(vm_page_t m) 796 { 797 798 return (vm_page_queue(m) == PQ_ACTIVE); 799 } 800 801 static inline bool 802 vm_page_inactive(vm_page_t m) 803 { 804 805 return (vm_page_queue(m) == PQ_INACTIVE); 806 } 807 808 static inline bool 809 vm_page_in_laundry(vm_page_t m) 810 { 811 uint8_t queue; 812 813 queue = vm_page_queue(m); 814 return (queue == PQ_LAUNDRY || queue == PQ_UNSWAPPABLE); 815 } 816 817 /* 818 * vm_page_wired: 819 * 820 * Return true if a reference prevents the page from being reclaimable. 821 */ 822 static inline bool 823 vm_page_wired(vm_page_t m) 824 { 825 826 return (m->wire_count > 0); 827 } 828 829 #endif /* _KERNEL */ 830 #endif /* !_VM_PAGE_ */ 831