osfmk/vm/vm_page.h

   1 /*
   2  * Copyright (c) 2000-2006 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28 /*
  29  * @OSF_COPYRIGHT@
  30  */
  31 /*
  32  * Mach Operating System
  33  * Copyright (c) 1991,1990,1989,1988 Carnegie Mellon University
  34  * All Rights Reserved.
  35  *
  36  * Permission to use, copy, modify and distribute this software and its
  37  * documentation is hereby granted, provided that both the copyright
  38  * notice and this permission notice appear in all copies of the
  39  * software, derivative works or modified versions, and any portions
  40  * thereof, and that both notices appear in supporting documentation.
  41  *
  42  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
  43  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
  44  * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
  45  *
  46  * Carnegie Mellon requests users of this software to return to
  47  *
  48  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
  49  *  School of Computer Science
  50  *  Carnegie Mellon University
  51  *  Pittsburgh PA 15213-3890
  52  *
  53  * any improvements or extensions that they make and grant Carnegie Mellon
  54  * the rights to redistribute these changes.
  55  */
  56 /*
  57  */
  58 /*
  59  *      File:   vm/vm_page.h
  60  *      Author: Avadis Tevanian, Jr., Michael Wayne Young
  61  *      Date:   1985
  62  *
  63  *      Resident memory system definitions.
  64  */
  65
  66 #ifndef _VM_VM_PAGE_H_
  67 #define _VM_VM_PAGE_H_
  68
  69 #include <debug.h>
  70
  71 #include <mach/boolean.h>
  72 #include <mach/vm_prot.h>
  73 #include <mach/vm_param.h>
  74 #include <vm/vm_object.h>
  75 #include <kern/queue.h>
  76 #include <kern/lock.h>
  77
  78 #include <kern/macro_help.h>
  79 #include <libkern/OSAtomic.h>
  80
  81
  82 /*
  83  * VM_PAGE_MIN_SPECULATIVE_AGE_Q through VM_PAGE_MAX_SPECULATIVE_AGE_Q
  84  * represents a set of aging bins that are 'protected'...
  85  *
  86  * VM_PAGE_SPECULATIVE_AGED_Q is a list of the speculative pages that have
  87  * not yet been 'claimed' but have been aged out of the protective bins
  88  * this occurs in vm_page_speculate when it advances to the next bin
  89  * and discovers that it is still occupied... at that point, all of the
  90  * pages in that bin are moved to the VM_PAGE_SPECULATIVE_AGED_Q.  the pages
  91  * in that bin are all guaranteed to have reached at least the maximum age
  92  * we allow for a protected page... they can be older if there is no
  93  * memory pressure to pull them from the bin, or there are no new speculative pages
  94  * being generated to push them out.
  95  * this list is the one that vm_pageout_scan will prefer when looking
  96  * for pages to move to the underweight free list
  97  *
  98  * VM_PAGE_MAX_SPECULATIVE_AGE_Q * VM_PAGE_SPECULATIVE_Q_AGE_MS
  99  * defines the amount of time a speculative page is normally
 100  * allowed to live in the 'protected' state (i.e. not available
 101  * to be stolen if vm_pageout_scan is running and looking for
 102  * pages)...  however, if the total number of speculative pages
 103  * in the protected state exceeds our limit (defined in vm_pageout.c)
 104  * and there are none available in VM_PAGE_SPECULATIVE_AGED_Q, then
 105  * vm_pageout_scan is allowed to steal pages from the protected
 106  * bucket even if they are underage.
 107  *
 108  * vm_pageout_scan is also allowed to pull pages from a protected
 109  * bin if the bin has reached the "age of consent" we've set
 110  */
 111 #define VM_PAGE_MAX_SPECULATIVE_AGE_Q   10
 112 #define VM_PAGE_MIN_SPECULATIVE_AGE_Q   1
 113 #define VM_PAGE_SPECULATIVE_AGED_Q      0
 114
 115 #define VM_PAGE_SPECULATIVE_Q_AGE_MS    500
 116
 117
 118 struct vm_speculative_age_q {
 119         /*
 120          * memory queue for speculative pages via clustered pageins
 121          */
 122         queue_head_t    age_q;
 123         mach_timespec_t age_ts;
 124 };
 125
 126
 127 extern
 128 struct vm_speculative_age_q     vm_page_queue_speculative[];
 129
 130 extern int                      speculative_steal_index;
 131 extern int                      speculative_age_index;
 132
 133
 134 /*
 135  *      Management of resident (logical) pages.
 136  *
 137  *      A small structure is kept for each resident
 138  *      page, indexed by page number.  Each structure
 139  *      is an element of several lists:
 140  *
 141  *              A hash table bucket used to quickly
 142  *              perform object/offset lookups
 143  *
 144  *              A list of all pages for a given object,
 145  *              so they can be quickly deactivated at
 146  *              time of deallocation.
 147  *
 148  *              An ordered list of pages due for pageout.
 149  *
 150  *      In addition, the structure contains the object
 151  *      and offset to which this page belongs (for pageout),
 152  *      and sundry status bits.
 153  *
 154  *      Fields in this structure are locked either by the lock on the
 155  *      object that the page belongs to (O) or by the lock on the page
 156  *      queues (P).  [Some fields require that both locks be held to
 157  *      change that field; holding either lock is sufficient to read.]
 158  */
 159
 160 struct vm_page {
 161         queue_chain_t   pageq;          /* queue info for FIFO */
 162                                         /* queue or free list (P) */
 163
 164         queue_chain_t   listq;          /* all pages in same object (O) */
 165         struct vm_page  *next;          /* VP bucket link (O) */
 166
 167         vm_object_t     object;         /* which object am I in (O&P) */
 168         vm_object_offset_t offset;      /* offset into that object (O,P) */
 169
 170         /*
 171          * The following word of flags is protected
 172          * by the "page queues" lock.
 173          *
 174          * we use the 'wire_count' field to store the local
 175          * queue id if local queues are enabled...
 176          * see the comments at 'VM_PAGE_QUEUES_REMOVE' as to
 177          * why this is safe to do
 178          */
 179 #define local_id wire_count
 180         unsigned int    wire_count:16,  /* how many wired down maps use me? (O&P) */
 181         /* boolean_t */ inactive:1,     /* page is in inactive list (P) */
 182                         active:1,       /* page is in active list (P) */
 183                         pageout_queue:1,/* page is on queue for pageout (P) */
 184                         speculative:1,  /* page is on speculative list (P) */
 185                         laundry:1,      /* page is being cleaned now (P)*/
 186                         free:1,         /* page is on free list (P) */
 187                         reference:1,    /* page has been used (P) */
 188                         gobbled:1,      /* page used internally (P) */
 189                         private:1,      /* Page should not be returned to
 190                                          *  the free list (P) */
 191                         throttled:1,    /* pager is not responding (P) */
 192                         local:1,
 193                         __unused_pageq_bits:5;  /* 5 bits available here */
 194
 195         ppnum_t         phys_page;      /* Physical address of page, passed
 196                                          *  to pmap_enter (read-only) */
 197
 198         /*
 199          * The following word of flags is protected
 200          * by the "VM object" lock.
 201          */
 202         unsigned int
 203         /* boolean_t */ busy:1,         /* page is in transit (O) */
 204                         wanted:1,       /* someone is waiting for page (O) */
 205                         tabled:1,       /* page is in VP table (O) */
 206                         fictitious:1,   /* Physical page doesn't exist (O) */
 207         /*
 208          * IMPORTANT: the "pmapped" bit can be turned on while holding the
 209          * VM object "shared" lock.  See vm_fault_enter().
 210          * This is OK as long as it's the only bit in this bit field that
 211          * can be updated without holding the VM object "exclusive" lock.
 212          */
 213                         pmapped:1,      /* page has been entered at some
 214                                          * point into a pmap (O **shared**) */
 215                         wpmapped:1,     /* page has been entered at some
 216                                          * point into a pmap for write (O) */
 217                         pageout:1,      /* page wired & busy for pageout (O) */
 218                         absent:1,       /* Data has been requested, but is
 219                                          *  not yet available (O) */
 220                         error:1,        /* Data manager was unable to provide
 221                                          *  data due to error (O) */
 222                         dirty:1,        /* Page must be cleaned (O) */
 223                         cleaning:1,     /* Page clean has begun (O) */
 224                         precious:1,     /* Page is precious; data must be
 225                                          *  returned even if clean (O) */
 226                         clustered:1,    /* page is not the faulted page (O) */
 227                         overwriting:1,  /* Request to unlock has been made
 228                                          * without having data. (O)
 229                                          * [See vm_fault_page_overwrite] */
 230                         restart:1,      /* Page was pushed higher in shadow
 231                                            chain by copy_call-related pagers;
 232                                            start again at top of chain */
 233                         unusual:1,      /* Page is absent, error, restart or
 234                                            page locked */
 235                         encrypted:1,    /* encrypted for secure swap (O) */
 236                         encrypted_cleaning:1,   /* encrypting page */
 237                         list_req_pending:1, /* pagein/pageout alt mechanism */
 238                                             /* allows creation of list      */
 239                                             /* requests on pages that are   */
 240                                             /* actively being paged.        */
 241                         dump_cleaning:1,   /* set by the pageout daemon when */
 242                                            /* a page being cleaned is       */
 243                                            /* encountered and targeted as   */
 244                                            /* a pageout candidate           */
 245                         cs_validated:1,    /* code-signing: page was checked */
 246                         cs_tainted:1,      /* code-signing: page is tainted */
 247                         no_cache:1,        /* page is not to be cached and */
 248                                            /* should be reused ahead of    */
 249                                            /* other pages                  */
 250                         zero_fill:1,
 251                         reusable:1,
 252                         lopage:1,
 253                         __unused_object_bits:6;  /* 6 bits available here */
 254
 255 #if __LP64__
 256         unsigned int __unused_padding;  /* Pad structure explicitly
 257                                         * to 8-byte multiple for LP64 */
 258 #endif
 259 };
 260
 261 #define DEBUG_ENCRYPTED_SWAP    1
 262 #if DEBUG_ENCRYPTED_SWAP
 263 #define ASSERT_PAGE_DECRYPTED(page)                                     \
 264         MACRO_BEGIN                                                     \
 265         if ((page)->encrypted) {                                        \
 266                 panic("VM page %p should not be encrypted here\n",      \
 267                       (page));                                          \
 268         }                                                               \
 269         MACRO_END
 270 #else   /* DEBUG_ENCRYPTED_SWAP */
 271 #define ASSERT_PAGE_DECRYPTED(page) assert(!(page)->encrypted)
 272 #endif  /* DEBUG_ENCRYPTED_SWAP */
 273
 274 typedef struct vm_page  *vm_page_t;
 275
 276
 277 typedef struct vm_locks_array {
 278         char    pad  __attribute__ ((aligned (64)));
 279         lck_mtx_t       vm_page_queue_lock2 __attribute__ ((aligned (64)));
 280         lck_mtx_t       vm_page_queue_free_lock2 __attribute__ ((aligned (64)));
 281         char    pad2  __attribute__ ((aligned (64)));
 282 } vm_locks_array_t;
 283
 284
 285 #define VM_PAGE_WIRED(m)        ((!(m)->local && (m)->wire_count))
 286 #define VM_PAGE_NULL            ((vm_page_t) 0)
 287 #define NEXT_PAGE(m)            ((vm_page_t) (m)->pageq.next)
 288 #define NEXT_PAGE_PTR(m)        ((vm_page_t *) &(m)->pageq.next)
 289
 290 /*
 291  * XXX  The unusual bit should not be necessary.  Most of the bit
 292  * XXX  fields above really want to be masks.
 293  */
 294
 295 /*
 296  *      For debugging, this macro can be defined to perform
 297  *      some useful check on a page structure.
 298  */
 299
 300 #define VM_PAGE_CHECK(mem)                      \
 301         MACRO_BEGIN                             \
 302         VM_PAGE_QUEUES_ASSERT(mem, 1);          \
 303         MACRO_END
 304
 305 /*     Page coloring:
 306  *
 307  *     The free page list is actually n lists, one per color,
 308  *     where the number of colors is a function of the machine's
 309  *     cache geometry set at system initialization.  To disable
 310  *     coloring, set vm_colors to 1 and vm_color_mask to 0.
 311  *     The boot-arg "colors" may be used to override vm_colors.
 312  *     Note that there is little harm in having more colors than needed.
 313  */
 314
 315 #define MAX_COLORS      128
 316 #define DEFAULT_COLORS  32
 317
 318 extern
 319 unsigned int    vm_colors;              /* must be in range 1..MAX_COLORS */
 320 extern
 321 unsigned int    vm_color_mask;          /* must be (vm_colors-1) */
 322 extern
 323 unsigned int    vm_cache_geometry_colors; /* optimal #colors based on cache geometry */
 324
 325 /*
 326  * Wired memory is a very limited resource and we can't let users exhaust it
 327  * and deadlock the entire system.  We enforce the following limits:
 328  *
 329  * vm_user_wire_limit (default: all memory minus vm_global_no_user_wire_amount)
 330  *      how much memory can be user-wired in one user task
 331  *
 332  * vm_global_user_wire_limit (default: same as vm_user_wire_limit)
 333  *      how much memory can be user-wired in all user tasks
 334  *
 335  * vm_global_no_user_wire_amount (default: VM_NOT_USER_WIREABLE)
 336  *      how much memory must remain user-unwired at any time
 337  */
 338 #define VM_NOT_USER_WIREABLE (64*1024*1024)     /* 64MB */
 339 extern
 340 vm_map_size_t   vm_user_wire_limit;
 341 extern
 342 vm_map_size_t   vm_global_user_wire_limit;
 343 extern
 344 vm_map_size_t   vm_global_no_user_wire_amount;
 345
 346 /*
 347  *      Each pageable resident page falls into one of three lists:
 348  *
 349  *      free
 350  *              Available for allocation now.  The free list is
 351  *              actually an array of lists, one per color.
 352  *      inactive
 353  *              Not referenced in any map, but still has an
 354  *              object/offset-page mapping, and may be dirty.
 355  *              This is the list of pages that should be
 356  *              paged out next.  There are actually two
 357  *              inactive lists, one for pages brought in from
 358  *              disk or other backing store, and another
 359  *              for "zero-filled" pages.  See vm_pageout_scan()
 360  *              for the distinction and usage.
 361  *      active
 362  *              A list of pages which have been placed in
 363  *              at least one physical map.  This list is
 364  *              ordered, in LRU-like fashion.
 365  */
 366
 367
 368 #define VPL_LOCK_SPIN 1
 369
 370 struct vpl {
 371         unsigned int    vpl_count;
 372         queue_head_t    vpl_queue;
 373 #ifdef  VPL_LOCK_SPIN
 374         lck_spin_t      vpl_lock;
 375 #else
 376         lck_mtx_t       vpl_lock;
 377         lck_mtx_ext_t   vpl_lock_ext;
 378 #endif
 379 };
 380
 381 struct  vplq {
 382         union {
 383                 char   cache_line_pad[128];
 384                 struct vpl vpl;
 385         } vpl_un;
 386 };
 387 extern
 388 unsigned int    vm_page_local_q_count;
 389 extern
 390 struct vplq     *vm_page_local_q;
 391 extern
 392 unsigned int    vm_page_local_q_soft_limit;
 393 extern
 394 unsigned int    vm_page_local_q_hard_limit;
 395 extern
 396 vm_locks_array_t vm_page_locks;
 397
 398 extern
 399 queue_head_t    vm_page_queue_free[MAX_COLORS]; /* memory free queue */
 400 extern
 401 queue_head_t    vm_lopage_queue_free;           /* low memory free queue */
 402 extern
 403 vm_page_t       vm_page_queue_fictitious;       /* fictitious free queue */
 404 extern
 405 queue_head_t    vm_page_queue_active;   /* active memory queue */
 406 extern
 407 queue_head_t    vm_page_queue_inactive; /* inactive memory queue for normal pages */
 408 extern
 409 queue_head_t    vm_page_queue_zf;       /* inactive memory queue for zero fill */
 410 extern
 411 queue_head_t    vm_page_queue_throttled;        /* memory queue for throttled pageout pages */
 412
 413 extern
 414 vm_offset_t     first_phys_addr;        /* physical address for first_page */
 415 extern
 416 vm_offset_t     last_phys_addr;         /* physical address for last_page */
 417
 418 extern
 419 unsigned int    vm_page_free_count;     /* How many pages are free? (sum of all colors) */
 420 extern
 421 unsigned int    vm_page_fictitious_count;/* How many fictitious pages are free? */
 422 extern
 423 unsigned int    vm_page_active_count;   /* How many pages are active? */
 424 extern
 425 unsigned int    vm_page_inactive_count; /* How many pages are inactive? */
 426 extern
 427 unsigned int    vm_page_throttled_count;/* How many inactives are throttled */
 428 extern
 429 unsigned int    vm_page_speculative_count;      /* How many speculative pages are unclaimed? */
 430 extern
 431 unsigned int    vm_page_wire_count;     /* How many pages are wired? */
 432 extern
 433 unsigned int    vm_page_free_target;    /* How many do we want free? */
 434 extern
 435 unsigned int    vm_page_free_min;       /* When to wakeup pageout */
 436 extern
 437 unsigned int    vm_page_throttle_limit; /* When to throttle new page creation */
 438 extern
 439 uint32_t        vm_page_creation_throttle;      /* When to throttle new page creation */
 440 extern
 441 unsigned int    vm_page_inactive_target;/* How many do we want inactive? */
 442 extern
 443 unsigned int    vm_page_inactive_min;   /* When do wakeup pageout */
 444 extern
 445 unsigned int    vm_page_free_reserved;  /* How many pages reserved to do pageout */
 446 extern
 447 unsigned int    vm_page_throttle_count; /* Count of page allocations throttled */
 448 extern
 449 unsigned int    vm_page_gobble_count;
 450
 451 #if DEVELOPMENT || DEBUG
 452 extern
 453 unsigned int    vm_page_speculative_used;
 454 #endif
 455
 456 extern
 457 unsigned int    vm_page_purgeable_count;/* How many pages are purgeable now ? */
 458 extern
 459 unsigned int    vm_page_purgeable_wired_count;/* How many purgeable pages are wired now ? */
 460 extern
 461 uint64_t        vm_page_purged_count;   /* How many pages got purged so far ? */
 462
 463 extern unsigned int     vm_page_free_wanted;
 464                                 /* how many threads are waiting for memory */
 465
 466 extern unsigned int     vm_page_free_wanted_privileged;
 467                                 /* how many VM privileged threads are waiting for memory */
 468
 469 extern ppnum_t  vm_page_fictitious_addr;
 470                                 /* (fake) phys_addr of fictitious pages */
 471
 472 extern ppnum_t  vm_page_guard_addr;
 473                                 /* (fake) phys_addr of guard pages */
 474
 475
 476 extern boolean_t        vm_page_deactivate_hint;
 477
 478 /*
 479    0 = all pages avail ( default. )
 480    1 = disable high mem ( cap max pages to 4G)
 481    2 = prefer himem
 482 */
 483 extern int              vm_himemory_mode;
 484
 485 extern boolean_t        vm_lopage_needed;
 486 extern uint32_t         vm_lopage_free_count;
 487 extern uint32_t         vm_lopage_free_limit;
 488 extern uint32_t         vm_lopage_lowater;
 489 extern boolean_t        vm_lopage_refill;
 490 extern uint64_t         max_valid_dma_address;
 491 extern ppnum_t          max_valid_low_ppnum;
 492
 493 /*
 494  * Prototypes for functions exported by this module.
 495  */
 496 extern void             vm_page_bootstrap(
 497                                         vm_offset_t     *startp,
 498                                         vm_offset_t     *endp) __attribute__((section("__TEXT, initcode")));
 499
 500 extern void             vm_page_module_init(void) __attribute__((section("__TEXT, initcode")));
 501
 502 extern void             vm_page_init_local_q(void);
 503
 504 extern void             vm_page_create(
 505                                         ppnum_t         start,
 506                                         ppnum_t         end);
 507
 508 extern vm_page_t        vm_page_lookup(
 509                                         vm_object_t             object,
 510                                         vm_object_offset_t      offset);
 511
 512 extern vm_page_t        vm_page_grab_fictitious(void);
 513
 514 extern vm_page_t        vm_page_grab_guard(void);
 515
 516 extern void             vm_page_release_fictitious(
 517                                         vm_page_t page);
 518
 519 extern void             vm_page_more_fictitious(void);
 520
 521 extern int              vm_pool_low(void);
 522
 523 extern vm_page_t        vm_page_grab(void);
 524
 525 extern vm_page_t        vm_page_grablo(void);
 526
 527 extern void             vm_page_release(
 528                                         vm_page_t       page);
 529
 530 extern boolean_t        vm_page_wait(
 531                                         int             interruptible );
 532
 533 extern vm_page_t        vm_page_alloc(
 534                                         vm_object_t             object,
 535                                         vm_object_offset_t      offset);
 536
 537 extern vm_page_t        vm_page_alloclo(
 538                                         vm_object_t             object,
 539                                         vm_object_offset_t      offset);
 540
 541 extern vm_page_t        vm_page_alloc_guard(
 542         vm_object_t             object,
 543         vm_object_offset_t      offset);
 544
 545 extern void             vm_page_init(
 546                                         vm_page_t       page,
 547                                         ppnum_t         phys_page,
 548                                         boolean_t       lopage);
 549
 550 extern void             vm_page_free(
 551                                         vm_page_t       page);
 552
 553 extern void             vm_page_free_unlocked(
 554                                         vm_page_t       page,
 555                                         boolean_t       remove_from_hash);
 556
 557 extern void             vm_page_activate(
 558                                         vm_page_t       page);
 559
 560 extern void             vm_page_deactivate(
 561                                         vm_page_t       page);
 562
 563 extern void             vm_page_deactivate_internal(
 564                                         vm_page_t       page,
 565                                         boolean_t       clear_hw_reference);
 566
 567 extern void             vm_page_lru(
 568                                         vm_page_t       page);
 569
 570 extern void             vm_page_speculate(
 571                                         vm_page_t       page,
 572                                         boolean_t       new);
 573
 574 extern void             vm_page_speculate_ageit(
 575                                         struct vm_speculative_age_q *aq);
 576
 577 extern void             vm_page_reactivate_all_throttled(void);
 578
 579 extern void             vm_page_reactivate_local(uint32_t lid, boolean_t force, boolean_t nolocks);
 580
 581 extern void             vm_page_rename(
 582                                         vm_page_t               page,
 583                                         vm_object_t             new_object,
 584                                         vm_object_offset_t      new_offset,
 585                                         boolean_t               encrypted_ok);
 586
 587 extern void             vm_page_insert(
 588                                         vm_page_t               page,
 589                                         vm_object_t             object,
 590                                         vm_object_offset_t      offset);
 591
 592 extern void             vm_page_insert_internal(
 593                                         vm_page_t               page,
 594                                         vm_object_t             object,
 595                                         vm_object_offset_t      offset,
 596                                         boolean_t               queues_lock_held,
 597                                         boolean_t               insert_in_hash);
 598
 599 extern void             vm_page_replace(
 600                                         vm_page_t               mem,
 601                                         vm_object_t             object,
 602                                         vm_object_offset_t      offset);
 603
 604 extern void             vm_page_remove(
 605                                         vm_page_t       page,
 606                                         boolean_t       remove_from_hash);
 607
 608 extern void             vm_page_zero_fill(
 609                                         vm_page_t       page);
 610
 611 extern void             vm_page_part_zero_fill(
 612                                         vm_page_t       m,
 613                                         vm_offset_t     m_pa,
 614                                         vm_size_t       len);
 615
 616 extern void             vm_page_copy(
 617                                         vm_page_t       src_page,
 618                                         vm_page_t       dest_page);
 619
 620 extern void             vm_page_part_copy(
 621                                         vm_page_t       src_m,
 622                                         vm_offset_t     src_pa,
 623                                         vm_page_t       dst_m,
 624                                         vm_offset_t     dst_pa,
 625                                         vm_size_t       len);
 626
 627 extern void             vm_page_wire(
 628                                         vm_page_t       page);
 629
 630 extern void             vm_page_unwire(
 631                                         vm_page_t       page,
 632                                         boolean_t       queueit);
 633
 634 extern void             vm_set_page_size(void);
 635
 636 extern void             vm_page_gobble(
 637                                         vm_page_t      page);
 638
 639 extern void             vm_page_validate_cs(vm_page_t   page);
 640 extern void             vm_page_validate_cs_mapped(
 641         vm_page_t       page,
 642         const void      *kaddr);
 643
 644 extern void             vm_page_free_prepare_queues(
 645                                         vm_page_t       page);
 646
 647 extern void             vm_page_free_prepare_object(
 648                                         vm_page_t       page,
 649                                         boolean_t       remove_from_hash);
 650
 651 /*
 652  *      Functions implemented as macros. m->wanted and m->busy are
 653  *      protected by the object lock.
 654  */
 655
 656 #define PAGE_ASSERT_WAIT(m, interruptible)                      \
 657                 (((m)->wanted = TRUE),                          \
 658                  assert_wait((event_t) (m), (interruptible)))
 659
 660 #define PAGE_SLEEP(o, m, interruptible)                         \
 661                 (((m)->wanted = TRUE),                          \
 662                  thread_sleep_vm_object((o), (m), (interruptible)))
 663
 664 #define PAGE_WAKEUP_DONE(m)                                     \
 665                 MACRO_BEGIN                                     \
 666                 (m)->busy = FALSE;                              \
 667                 if ((m)->wanted) {                              \
 668                         (m)->wanted = FALSE;                    \
 669                         thread_wakeup((event_t) (m));           \
 670                 }                                               \
 671                 MACRO_END
 672
 673 #define PAGE_WAKEUP(m)                                          \
 674                 MACRO_BEGIN                                     \
 675                 if ((m)->wanted) {                              \
 676                         (m)->wanted = FALSE;                    \
 677                         thread_wakeup((event_t) (m));           \
 678                 }                                               \
 679                 MACRO_END
 680
 681 #define VM_PAGE_FREE(p)                         \
 682                 MACRO_BEGIN                     \
 683                 vm_page_free_unlocked(p, TRUE); \
 684                 MACRO_END
 685
 686 #define VM_PAGE_GRAB_FICTITIOUS(M)                                      \
 687                 MACRO_BEGIN                                             \
 688                 while ((M = vm_page_grab_fictitious()) == VM_PAGE_NULL) \
 689                         vm_page_more_fictitious();                      \
 690                 MACRO_END
 691
 692 #define VM_PAGE_WAIT()          ((void)vm_page_wait(THREAD_UNINT))
 693
 694 #define vm_page_queue_lock (vm_page_locks.vm_page_queue_lock2)
 695 #define vm_page_queue_free_lock (vm_page_locks.vm_page_queue_free_lock2)
 696
 697 #define vm_page_lock_queues()   lck_mtx_lock(&vm_page_queue_lock)
 698 #define vm_page_unlock_queues() lck_mtx_unlock(&vm_page_queue_lock)
 699
 700 #define vm_page_lockspin_queues()       lck_mtx_lock_spin(&vm_page_queue_lock)
 701 #define vm_page_trylockspin_queues()    lck_mtx_try_lock_spin(&vm_page_queue_lock)
 702 #define vm_page_lockconvert_queues()    lck_mtx_convert_spin(&vm_page_queue_lock)
 703
 704 #ifdef  VPL_LOCK_SPIN
 705 #define VPL_LOCK_INIT(vlq, vpl_grp, vpl_attr) lck_spin_init(&vlq->vpl_lock, vpl_grp, vpl_attr)
 706 #define VPL_LOCK(vpl) lck_spin_lock(vpl)
 707 #define VPL_UNLOCK(vpl) lck_spin_unlock(vpl)
 708 #else
 709 #define VPL_LOCK_INIT(vlq, vpl_grp, vpl_attr) lck_mtx_init_ext(&vlq->vpl_lock, &vlq->vpl_lock_ext, vpl_grp, vpl_attr)
 710 #define VPL_LOCK(vpl) lck_mtx_lock_spin(vpl)
 711 #define VPL_UNLOCK(vpl) lck_mtx_unlock(vpl)
 712 #endif
 713
 714 #if MACH_ASSERT
 715 extern void vm_page_queues_assert(vm_page_t mem, int val);
 716 #define VM_PAGE_QUEUES_ASSERT(mem, val) vm_page_queues_assert((mem), (val))
 717 #else
 718 #define VM_PAGE_QUEUES_ASSERT(mem, val)
 719 #endif
 720
 721
 722 /*
 723  * 'vm_fault_enter' will place newly created pages (zero-fill and COW) onto the
 724  * local queues if they exist... its the only spot in the system where we add pages
 725  * to those queues...  once on those queues, those pages can only move to one of the
 726  * global page queues or the free queues... they NEVER move from local q to local q.
 727  * the 'local' state is stable when VM_PAGE_QUEUES_REMOVE is called since we're behind
 728  * the global vm_page_queue_lock at this point...  we still need to take the local lock
 729  * in case this operation is being run on a different CPU then the local queue's identity,
 730  * but we don't have to worry about the page moving to a global queue or becoming wired
 731  * while we're grabbing the local lock since those operations would require the global
 732  * vm_page_queue_lock to be held, and we already own it.
 733  *
 734  * this is why its safe to utilze the wire_count field in the vm_page_t as the local_id...
 735  * 'wired' and local are ALWAYS mutually exclusive conditions.
 736  */
 737 #define VM_PAGE_QUEUES_REMOVE(mem)                              \
 738         MACRO_BEGIN                                             \
 739         VM_PAGE_QUEUES_ASSERT(mem, 1);                          \
 740         assert(!mem->laundry);                                  \
 741         assert(!mem->pageout_queue);                            \
 742         if (mem->local) {                                       \
 743                 struct vpl      *lq;                            \
 744                 assert(mem->object != kernel_object);           \
 745                 assert(!mem->inactive && !mem->speculative);    \
 746                 assert(!mem->active && !mem->throttled);        \
 747                 lq = &vm_page_local_q[mem->local_id].vpl_un.vpl;        \
 748                 VPL_LOCK(&lq->vpl_lock);                        \
 749                 queue_remove(&lq->vpl_queue,                    \
 750                              mem, vm_page_t, pageq);            \
 751                 mem->local = FALSE;                             \
 752                 mem->local_id = 0;                              \
 753                 lq->vpl_count--;                                \
 754                 VPL_UNLOCK(&lq->vpl_lock);                      \
 755         }                                                       \
 756         if (mem->active) {                                      \
 757                 assert(mem->object != kernel_object);           \
 758                 assert(!mem->inactive && !mem->speculative);    \
 759                 assert(!mem->throttled);                        \
 760                 queue_remove(&vm_page_queue_active,             \
 761                         mem, vm_page_t, pageq);                 \
 762                 mem->active = FALSE;                            \
 763                 if (!mem->fictitious) {                         \
 764                         vm_page_active_count--;                 \
 765                 } else {                                        \
 766                         assert(mem->phys_page ==                \
 767                                vm_page_fictitious_addr);        \
 768                 }                                               \
 769         }                                                       \
 770                                                                 \
 771         else if (mem->inactive) {                               \
 772                 assert(mem->object != kernel_object);           \
 773                 assert(!mem->active && !mem->speculative);      \
 774                 assert(!mem->throttled);                        \
 775                 if (mem->zero_fill) {                           \
 776                         queue_remove(&vm_page_queue_zf,         \
 777                         mem, vm_page_t, pageq);                 \
 778                         vm_zf_queue_count--;                    \
 779                 } else {                                        \
 780                         queue_remove(&vm_page_queue_inactive,   \
 781                         mem, vm_page_t, pageq);                 \
 782                 }                                               \
 783                 mem->inactive = FALSE;                          \
 784                 if (!mem->fictitious) {                         \
 785                         vm_page_inactive_count--;               \
 786                         vm_purgeable_q_advance_all();           \
 787                 } else {                                        \
 788                         assert(mem->phys_page ==                \
 789                                vm_page_fictitious_addr);        \
 790                 }                                               \
 791         }                                                       \
 792                                                                 \
 793         else if (mem->throttled) {                              \
 794                 assert(!mem->active && !mem->inactive);         \
 795                 assert(!mem->speculative);                      \
 796                 queue_remove(&vm_page_queue_throttled,          \
 797                              mem, vm_page_t, pageq);            \
 798                 mem->throttled = FALSE;                         \
 799                 if (!mem->fictitious)                           \
 800                         vm_page_throttled_count--;              \
 801         }                                                       \
 802                                                                 \
 803         else if (mem->speculative) {                            \
 804                 assert(!mem->active && !mem->inactive);         \
 805                 assert(!mem->throttled);                        \
 806                 assert(!mem->fictitious);                       \
 807                 remque(&mem->pageq);                            \
 808                 mem->speculative = FALSE;                       \
 809                 vm_page_speculative_count--;                    \
 810         }                                                       \
 811         mem->pageq.next = NULL;                                 \
 812         mem->pageq.prev = NULL;                                 \
 813         VM_PAGE_QUEUES_ASSERT(mem, 0);                          \
 814         MACRO_END
 815
 816
 817 #if DEVELOPMENT || DEBUG
 818 #define VM_PAGE_SPECULATIVE_USED_ADD()                          \
 819         MACRO_BEGIN                                             \
 820         OSAddAtomic(1, &vm_page_speculative_used);      \
 821         MACRO_END
 822 #else
 823 #define VM_PAGE_SPECULATIVE_USED_ADD()
 824 #endif
 825
 826
 827 #define VM_PAGE_CONSUME_CLUSTERED(mem)                          \
 828         MACRO_BEGIN                                             \
 829         if (mem->clustered) {                                   \
 830                 assert(mem->object);                            \
 831                 mem->object->pages_used++;                      \
 832                 mem->clustered = FALSE;                         \
 833                 VM_PAGE_SPECULATIVE_USED_ADD();                 \
 834         }                                                       \
 835         MACRO_END
 836
 837 #endif  /* _VM_VM_PAGE_H_ */