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 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgement: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * from: @(#)vm_page.h 8.2 (Berkeley) 12/13/93 37 * 38 * 39 * Copyright (c) 1987, 1990 Carnegie-Mellon University. 40 * All rights reserved. 41 * 42 * Authors: Avadis Tevanian, Jr., Michael Wayne Young 43 * 44 * Permission to use, copy, modify and distribute this software and 45 * its documentation is hereby granted, provided that both the copyright 46 * notice and this permission notice appear in all copies of the 47 * software, derivative works or modified versions, and any portions 48 * thereof, and that both notices appear in supporting documentation. 49 * 50 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 51 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND 52 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 53 * 54 * Carnegie Mellon requests users of this software to return to 55 * 56 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 57 * School of Computer Science 58 * Carnegie Mellon University 59 * Pittsburgh PA 15213-3890 60 * 61 * any improvements or extensions that they make and grant Carnegie the 62 * rights to redistribute these changes. 63 * 64 * $FreeBSD: src/sys/vm/vm_page.h,v 1.75.2.8 2002/03/06 01:07:09 dillon Exp $ 65 */ 66 67 /* 68 * Resident memory system definitions. 69 */ 70 71 #ifndef _VM_VM_PAGE_H_ 72 #define _VM_VM_PAGE_H_ 73 74 #ifndef _SYS_TYPES_H_ 75 #include <sys/types.h> 76 #endif 77 #ifndef _SYS_TREE_H_ 78 #include <sys/tree.h> 79 #endif 80 #ifndef _MACHINE_PMAP_H_ 81 #include <machine/pmap.h> 82 #endif 83 #ifndef _VM_PMAP_H_ 84 #include <vm/pmap.h> 85 #endif 86 #include <machine/atomic.h> 87 88 #ifdef _KERNEL 89 90 #ifndef _SYS_SYSTM_H_ 91 #include <sys/systm.h> 92 #endif 93 #ifndef _SYS_THREAD2_H_ 94 #include <sys/thread2.h> 95 #endif 96 97 #ifdef __x86_64__ 98 #include <machine/vmparam.h> 99 #endif 100 101 #endif 102 103 typedef enum vm_page_event { VMEVENT_NONE, VMEVENT_COW } vm_page_event_t; 104 105 struct vm_page_action { 106 LIST_ENTRY(vm_page_action) entry; 107 struct vm_page *m; 108 vm_page_event_t event; 109 void (*func)(struct vm_page *, 110 struct vm_page_action *); 111 void *data; 112 }; 113 114 typedef struct vm_page_action *vm_page_action_t; 115 116 /* 117 * Management of resident (logical) pages. 118 * 119 * A small structure is kept for each resident 120 * page, indexed by page number. Each structure 121 * is an element of several lists: 122 * 123 * A hash table bucket used to quickly 124 * perform object/offset lookups 125 * 126 * A list of all pages for a given object, 127 * so they can be quickly deactivated at 128 * time of deallocation. 129 * 130 * An ordered list of pages due for pageout. 131 * 132 * In addition, the structure contains the object 133 * and offset to which this page belongs (for pageout), 134 * and sundry status bits. 135 * 136 * Fields in this structure are locked either by the lock on the 137 * object that the page belongs to (O) or by the lock on the page 138 * queues (P). 139 * 140 * The 'valid' and 'dirty' fields are distinct. A page may have dirty 141 * bits set without having associated valid bits set. This is used by 142 * NFS to implement piecemeal writes. 143 */ 144 145 TAILQ_HEAD(pglist, vm_page); 146 147 struct vm_object; 148 149 int rb_vm_page_compare(struct vm_page *, struct vm_page *); 150 151 struct vm_page_rb_tree; 152 RB_PROTOTYPE2(vm_page_rb_tree, vm_page, rb_entry, rb_vm_page_compare, vm_pindex_t); 153 154 struct vm_page { 155 TAILQ_ENTRY(vm_page) pageq; /* vm_page_queues[] list (P) */ 156 RB_ENTRY(vm_page) rb_entry; /* Red-Black tree based at object */ 157 158 struct vm_object *object; /* which object am I in (O,P)*/ 159 vm_pindex_t pindex; /* offset into object (O,P) */ 160 vm_paddr_t phys_addr; /* physical address of page */ 161 struct md_page md; /* machine dependant stuff */ 162 u_short queue; /* page queue index */ 163 u_short pc; /* page color */ 164 u_char act_count; /* page usage count */ 165 u_char busy; /* page busy count */ 166 u_char pat_mode; /* hardware page attribute */ 167 u_char unused02; 168 u_int32_t flags; /* see below */ 169 u_int wire_count; /* wired down maps refs (P) */ 170 int hold_count; /* page hold count */ 171 172 /* 173 * NOTE that these must support one bit per DEV_BSIZE in a page!!! 174 * so, on normal X86 kernels, they must be at least 8 bits wide. 175 */ 176 u_char valid; /* map of valid DEV_BSIZE chunks */ 177 u_char dirty; /* map of dirty DEV_BSIZE chunks */ 178 179 int ku_pagecnt; /* kmalloc helper */ 180 #ifdef VM_PAGE_DEBUG 181 const char *busy_func; 182 int busy_line; 183 #endif 184 }; 185 186 #ifdef VM_PAGE_DEBUG 187 #define VM_PAGE_DEBUG_EXT(name) name ## _debug 188 #define VM_PAGE_DEBUG_ARGS , const char *func, int lineno 189 #else 190 #define VM_PAGE_DEBUG_EXT(name) name 191 #define VM_PAGE_DEBUG_ARGS 192 #endif 193 194 #ifndef __VM_PAGE_T_DEFINED__ 195 #define __VM_PAGE_T_DEFINED__ 196 typedef struct vm_page *vm_page_t; 197 #endif 198 199 /* 200 * Page coloring parameters. We use generous parameters designed to 201 * statistically spread pages over available cpu cache space. This has 202 * become less important over time as cache associativity is higher 203 * in modern times but we still use the core algorithm to help reduce 204 * lock contention between cpus. 205 * 206 * Page coloring cannot be disabled. 207 */ 208 209 #define PQ_PRIME1 31 /* Prime number somewhat less than PQ_HASH_SIZE */ 210 #define PQ_PRIME2 23 /* Prime number somewhat less than PQ_HASH_SIZE */ 211 #define PQ_L2_SIZE 256 /* A number of colors opt for 1M cache */ 212 213 #if 0 214 #define PQ_PRIME1 31 /* Prime number somewhat less than PQ_HASH_SIZE */ 215 #define PQ_PRIME2 23 /* Prime number somewhat less than PQ_HASH_SIZE */ 216 #define PQ_L2_SIZE 128 /* A number of colors opt for 512K cache */ 217 218 #define PQ_PRIME1 13 /* Prime number somewhat less than PQ_HASH_SIZE */ 219 #define PQ_PRIME2 7 /* Prime number somewhat less than PQ_HASH_SIZE */ 220 #define PQ_L2_SIZE 64 /* A number of colors opt for 256K cache */ 221 222 #define PQ_PRIME1 9 /* Produces a good PQ_L2_SIZE/3 + PQ_PRIME1 */ 223 #define PQ_PRIME2 5 /* Prime number somewhat less than PQ_HASH_SIZE */ 224 #define PQ_L2_SIZE 32 /* A number of colors opt for 128k cache */ 225 226 #define PQ_PRIME1 5 /* Prime number somewhat less than PQ_HASH_SIZE */ 227 #define PQ_PRIME2 3 /* Prime number somewhat less than PQ_HASH_SIZE */ 228 #define PQ_L2_SIZE 16 /* A reasonable number of colors (opt for 64K cache) */ 229 #endif 230 231 #define PQ_L2_MASK (PQ_L2_SIZE - 1) 232 233 #define PQ_NONE 0 234 #define PQ_FREE (1 + 0*PQ_L2_SIZE) 235 #define PQ_INACTIVE (1 + 1*PQ_L2_SIZE) 236 #define PQ_ACTIVE (1 + 2*PQ_L2_SIZE) 237 #define PQ_CACHE (1 + 3*PQ_L2_SIZE) 238 #define PQ_HOLD (1 + 4*PQ_L2_SIZE) 239 #define PQ_COUNT (1 + 5*PQ_L2_SIZE) 240 241 /* 242 * Scan support 243 */ 244 struct vm_map; 245 246 struct rb_vm_page_scan_info { 247 vm_pindex_t start_pindex; 248 vm_pindex_t end_pindex; 249 int limit; 250 int desired; 251 int error; 252 int pagerflags; 253 vm_offset_t addr; 254 vm_pindex_t backing_offset_index; 255 struct vm_object *object; 256 struct vm_object *backing_object; 257 struct vm_page *mpte; 258 struct pmap *pmap; 259 struct vm_map *map; 260 }; 261 262 int rb_vm_page_scancmp(struct vm_page *, void *); 263 264 struct vpgqueues { 265 struct pglist pl; 266 int *cnt; 267 int lcnt; 268 int flipflop; /* probably not the best place */ 269 struct spinlock spin; 270 char unused[64 - sizeof(struct pglist) - 271 sizeof(int *) - sizeof(int) * 2]; 272 }; 273 274 extern struct vpgqueues vm_page_queues[PQ_COUNT]; 275 276 /* 277 * These are the flags defined for vm_page. 278 * 279 * PG_UNMANAGED (used by OBJT_PHYS) indicates that the page is 280 * not under PV management but otherwise should be treated as a 281 * normal page. Pages not under PV management cannot be paged out 282 * via the object/vm_page_t because there is no knowledge of their 283 * pte mappings, nor can they be removed from their objects via 284 * the object, and such pages are also not on any PQ queue. The 285 * PG_MAPPED and PG_WRITEABLE flags are not applicable. 286 * 287 * PG_MAPPED only applies to managed pages, indicating whether the page 288 * is mapped onto one or more pmaps. A page might still be mapped to 289 * special pmaps in an unmanaged fashion, for example when mapped into a 290 * buffer cache buffer, without setting PG_MAPPED. 291 * 292 * PG_WRITEABLE indicates that there may be a writeable managed pmap entry 293 * somewhere, and that the page can be dirtied by hardware at any time 294 * and may have to be tested for that. The modified bit in unmanaged 295 * mappings or in the special clean map is not tested. 296 * 297 * PG_SWAPPED indicates that the page is backed by a swap block. Any 298 * VM object type other than OBJT_DEFAULT can have swap-backed pages now. 299 * 300 * PG_SBUSY is set when m->busy != 0. PG_SBUSY and m->busy are only modified 301 * when the page is PG_BUSY. 302 */ 303 #define PG_BUSY 0x00000001 /* page is in transit (O) */ 304 #define PG_WANTED 0x00000002 /* someone is waiting for page (O) */ 305 #define PG_WINATCFLS 0x00000004 /* flush dirty page on inactive q */ 306 #define PG_FICTITIOUS 0x00000008 /* physical page doesn't exist (O) */ 307 #define PG_WRITEABLE 0x00000010 /* page is writeable */ 308 #define PG_MAPPED 0x00000020 /* page is mapped (managed) */ 309 #define PG_ZERO 0x00000040 /* page is zeroed */ 310 #define PG_REFERENCED 0x00000080 /* page has been referenced */ 311 #define PG_CLEANCHK 0x00000100 /* page will be checked for cleaning */ 312 #define PG_SWAPINPROG 0x00000200 /* swap I/O in progress on page */ 313 #define PG_NOSYNC 0x00000400 /* do not collect for syncer */ 314 #define PG_UNMANAGED 0x00000800 /* No PV management for page */ 315 #define PG_MARKER 0x00001000 /* special queue marker page */ 316 #define PG_RAM 0x00002000 /* read ahead mark */ 317 #define PG_SWAPPED 0x00004000 /* backed by swap */ 318 #define PG_NOTMETA 0x00008000 /* do not back with swap */ 319 #define PG_ACTIONLIST 0x00010000 /* lookaside action list present */ 320 #define PG_SBUSY 0x00020000 /* soft-busy also set */ 321 #define PG_NEED_COMMIT 0x00040000 /* clean page requires commit */ 322 323 /* 324 * Misc constants. 325 */ 326 327 #define ACT_DECLINE 1 328 #define ACT_ADVANCE 3 329 #define ACT_INIT 5 330 #define ACT_MAX 64 331 332 #ifdef _KERNEL 333 /* 334 * Each pageable resident page falls into one of four lists: 335 * 336 * free 337 * Available for allocation now. 338 * 339 * The following are all LRU sorted: 340 * 341 * cache 342 * Almost available for allocation. Still in an 343 * object, but clean and immediately freeable at 344 * non-interrupt times. 345 * 346 * inactive 347 * Low activity, candidates for reclamation. 348 * This is the list of pages that should be 349 * paged out next. 350 * 351 * active 352 * Pages that are "active" i.e. they have been 353 * recently referenced. 354 * 355 * zero 356 * Pages that are really free and have been pre-zeroed 357 * 358 */ 359 360 extern int vm_page_zero_count; 361 extern struct vm_page *vm_page_array; /* First resident page in table */ 362 extern int vm_page_array_size; /* number of vm_page_t's */ 363 extern long first_page; /* first physical page number */ 364 365 #define VM_PAGE_TO_PHYS(entry) \ 366 ((entry)->phys_addr) 367 368 #define PHYS_TO_VM_PAGE(pa) \ 369 (&vm_page_array[atop(pa) - first_page]) 370 371 372 #if PAGE_SIZE == 4096 373 #define VM_PAGE_BITS_ALL 0xff 374 #endif 375 376 /* 377 * Note: the code will always use nominally free pages from the free list 378 * before trying other flag-specified sources. 379 * 380 * At least one of VM_ALLOC_NORMAL|VM_ALLOC_SYSTEM|VM_ALLOC_INTERRUPT 381 * must be specified. VM_ALLOC_RETRY may only be specified if VM_ALLOC_NORMAL 382 * is also specified. 383 */ 384 #define VM_ALLOC_NORMAL 0x0001 /* ok to use cache pages */ 385 #define VM_ALLOC_SYSTEM 0x0002 /* ok to exhaust most of free list */ 386 #define VM_ALLOC_INTERRUPT 0x0004 /* ok to exhaust entire free list */ 387 #define VM_ALLOC_ZERO 0x0008 /* req pre-zero'd memory if avail */ 388 #define VM_ALLOC_QUICK 0x0010 /* like NORMAL but do not use cache */ 389 #define VM_ALLOC_FORCE_ZERO 0x0020 /* zero page even if already valid */ 390 #define VM_ALLOC_NULL_OK 0x0040 /* ok to return NULL on collision */ 391 #define VM_ALLOC_RETRY 0x0080 /* indefinite block (vm_page_grab()) */ 392 #define VM_ALLOC_USE_GD 0x0100 /* use per-gd cache */ 393 394 void vm_page_queue_spin_lock(vm_page_t); 395 void vm_page_queues_spin_lock(u_short); 396 void vm_page_and_queue_spin_lock(vm_page_t); 397 398 void vm_page_queue_spin_unlock(vm_page_t); 399 void vm_page_queues_spin_unlock(u_short); 400 void vm_page_and_queue_spin_unlock(vm_page_t m); 401 402 void vm_page_init(vm_page_t m); 403 void vm_page_io_finish(vm_page_t m); 404 void vm_page_io_start(vm_page_t m); 405 void vm_page_need_commit(vm_page_t m); 406 void vm_page_clear_commit(vm_page_t m); 407 void vm_page_wakeup(vm_page_t m); 408 void vm_page_hold(vm_page_t); 409 void vm_page_unhold(vm_page_t); 410 void vm_page_activate (vm_page_t); 411 void vm_page_pcpu_cache(void); 412 413 vm_size_t vm_contig_avail_pages(void); 414 vm_page_t vm_page_alloc (struct vm_object *, vm_pindex_t, int); 415 vm_page_t vm_page_alloc_contig(vm_paddr_t low, vm_paddr_t high, 416 unsigned long alignment, unsigned long boundary, 417 unsigned long size, vm_memattr_t memattr); 418 419 vm_page_t vm_page_grab (struct vm_object *, vm_pindex_t, int); 420 void vm_page_cache (vm_page_t); 421 int vm_page_try_to_cache (vm_page_t); 422 int vm_page_try_to_free (vm_page_t); 423 void vm_page_dontneed (vm_page_t); 424 void vm_page_deactivate (vm_page_t); 425 void vm_page_deactivate_locked (vm_page_t); 426 void vm_page_initfake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr); 427 int vm_page_insert (vm_page_t, struct vm_object *, vm_pindex_t); 428 vm_page_t vm_page_lookup (struct vm_object *, vm_pindex_t); 429 vm_page_t VM_PAGE_DEBUG_EXT(vm_page_lookup_busy_wait)( 430 struct vm_object *, vm_pindex_t, int, const char * 431 VM_PAGE_DEBUG_ARGS); 432 vm_page_t VM_PAGE_DEBUG_EXT(vm_page_lookup_busy_try)( 433 struct vm_object *, vm_pindex_t, int, int * 434 VM_PAGE_DEBUG_ARGS); 435 void vm_page_remove (vm_page_t); 436 void vm_page_rename (vm_page_t, struct vm_object *, vm_pindex_t); 437 void vm_page_startup (void); 438 void vm_page_unmanage (vm_page_t); 439 void vm_page_unwire (vm_page_t, int); 440 void vm_page_wire (vm_page_t); 441 void vm_page_unqueue (vm_page_t); 442 void vm_page_unqueue_nowakeup (vm_page_t); 443 vm_page_t vm_page_next (vm_page_t); 444 void vm_page_set_validclean (vm_page_t, int, int); 445 void vm_page_set_validdirty (vm_page_t, int, int); 446 void vm_page_set_valid (vm_page_t, int, int); 447 void vm_page_set_dirty (vm_page_t, int, int); 448 void vm_page_clear_dirty (vm_page_t, int, int); 449 void vm_page_set_invalid (vm_page_t, int, int); 450 int vm_page_is_valid (vm_page_t, int, int); 451 void vm_page_test_dirty (vm_page_t); 452 int vm_page_bits (int, int); 453 vm_page_t vm_page_list_find(int basequeue, int index, boolean_t prefer_zero); 454 void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid); 455 void vm_page_free_toq(vm_page_t m); 456 void vm_page_free_contig(vm_page_t m, unsigned long size); 457 vm_page_t vm_page_free_fromq_fast(void); 458 void vm_page_event_internal(vm_page_t, vm_page_event_t); 459 void vm_page_dirty(vm_page_t m); 460 void vm_page_register_action(vm_page_action_t action, vm_page_event_t event); 461 void vm_page_unregister_action(vm_page_action_t action); 462 void vm_page_sleep_busy(vm_page_t m, int also_m_busy, const char *msg); 463 void VM_PAGE_DEBUG_EXT(vm_page_busy_wait)(vm_page_t m, int also_m_busy, const char *wmsg VM_PAGE_DEBUG_ARGS); 464 int VM_PAGE_DEBUG_EXT(vm_page_busy_try)(vm_page_t m, int also_m_busy VM_PAGE_DEBUG_ARGS); 465 466 #ifdef VM_PAGE_DEBUG 467 468 #define vm_page_lookup_busy_wait(object, pindex, alsob, msg) \ 469 vm_page_lookup_busy_wait_debug(object, pindex, alsob, msg, \ 470 __func__, __LINE__) 471 472 #define vm_page_lookup_busy_try(object, pindex, alsob, errorp) \ 473 vm_page_lookup_busy_try_debug(object, pindex, alsob, errorp, \ 474 __func__, __LINE__) 475 476 #define vm_page_busy_wait(m, alsob, msg) \ 477 vm_page_busy_wait_debug(m, alsob, msg, __func__, __LINE__) 478 479 #define vm_page_busy_try(m, alsob) \ 480 vm_page_busy_try_debug(m, alsob, __func__, __LINE__) 481 482 #endif 483 484 #endif /* _KERNEL */ 485 #endif /* !_VM_VM_PAGE_H_ */ 486