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 struct vm_speculative_age_q {
 118         /*
 119          * memory queue for speculative pages via clustered pageins
 120          */
 121         queue_head_t    age_q;
 122         mach_timespec_t age_ts;
 123 };
 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 extern unsigned int             vm_page_speculative_q_age_ms;
 133
 134
 135 /*
 136  *      Management of resident (logical) pages.
 137  *
 138  *      A small structure is kept for each resident
 139  *      page, indexed by page number.  Each structure
 140  *      is an element of several lists:
 141  *
 142  *              A hash table bucket used to quickly
 143  *              perform object/offset lookups
 144  *
 145  *              A list of all pages for a given object,
 146  *              so they can be quickly deactivated at
 147  *              time of deallocation.
 148  *
 149  *              An ordered list of pages due for pageout.
 150  *
 151  *      In addition, the structure contains the object
 152  *      and offset to which this page belongs (for pageout),
 153  *      and sundry status bits.
 154  *
 155  *      Fields in this structure are locked either by the lock on the
 156  *      object that the page belongs to (O) or by the lock on the page
 157  *      queues (P).  [Some fields require that both locks be held to
 158  *      change that field; holding either lock is sufficient to read.]
 159  */
 160
 161 struct vm_page {
 162         queue_chain_t   pageq;          /* queue info for FIFO */
 163                                         /* queue or free list (P) */
 164
 165         queue_chain_t   listq;          /* all pages in same object (O) */
 166         struct vm_page  *next;          /* VP bucket link (O) */
 167
 168         vm_object_t     object;         /* which object am I in (O&P) */
 169         vm_object_offset_t offset;      /* offset into that object (O,P) */
 170
 171         /*
 172          * The following word of flags is protected
 173          * by the "page queues" lock.
 174          *
 175          * we use the 'wire_count' field to store the local
 176          * queue id if local queues are enabled...
 177          * see the comments at 'VM_PAGE_QUEUES_REMOVE' as to
 178          * why this is safe to do
 179          */
 180 #define local_id wire_count
 181         unsigned int    wire_count:16,  /* how many wired down maps use me? (O&P) */
 182         /* boolean_t */ inactive:1,     /* page is in inactive list (P) */
 183                         zero_fill:1,
 184                         active:1,       /* page is in active list (P) */
 185                         pageout_queue:1,/* page is on queue for pageout (P) */
 186                         speculative:1,  /* page is on speculative list (P) */
 187                         laundry:1,      /* page is being cleaned now (P)*/
 188                         free:1,         /* page is on free list (P) */
 189                         reference:1,    /* page has been used (P) */
 190                         gobbled:1,      /* page used internally (P) */
 191                         private:1,      /* Page should not be returned to
 192                                          *  the free list (P) */
 193                         throttled:1,    /* pager is not responding (P) */
 194                         local:1,
 195                         no_cache:1,     /* page is not to be cached and should
 196                                          * be reused ahead of other pages (P) */
 197                         __unused_pageq_bits:3;  /* 3 bits available here */
 198
 199         ppnum_t         phys_page;      /* Physical address of page, passed
 200                                          *  to pmap_enter (read-only) */
 201
 202         /*
 203          * The following word of flags is protected
 204          * by the "VM object" lock.
 205          */
 206         unsigned int
 207         /* boolean_t */ busy:1,         /* page is in transit (O) */
 208                         wanted:1,       /* someone is waiting for page (O) */
 209                         tabled:1,       /* page is in VP table (O) */
 210                         fictitious:1,   /* Physical page doesn't exist (O) */
 211         /*
 212          * IMPORTANT: the "pmapped" bit can be turned on while holding the
 213          * VM object "shared" lock.  See vm_fault_enter().
 214          * This is OK as long as it's the only bit in this bit field that
 215          * can be updated without holding the VM object "exclusive" lock.
 216          */
 217                         pmapped:1,      /* page has been entered at some
 218                                          * point into a pmap (O **shared**) */
 219                         wpmapped:1,     /* page has been entered at some
 220                                          * point into a pmap for write (O) */
 221                         pageout:1,      /* page wired & busy for pageout (O) */
 222                         absent:1,       /* Data has been requested, but is
 223                                          *  not yet available (O) */
 224                         error:1,        /* Data manager was unable to provide
 225                                          *  data due to error (O) */
 226                         dirty:1,        /* Page must be cleaned (O) */
 227                         cleaning:1,     /* Page clean has begun (O) */
 228                         precious:1,     /* Page is precious; data must be
 229                                          *  returned even if clean (O) */
 230                         clustered:1,    /* page is not the faulted page (O) */
 231                         overwriting:1,  /* Request to unlock has been made
 232                                          * without having data. (O)
 233                                          * [See vm_fault_page_overwrite] */
 234                         restart:1,      /* Page was pushed higher in shadow
 235                                            chain by copy_call-related pagers;
 236                                            start again at top of chain */
 237                         unusual:1,      /* Page is absent, error, restart or
 238                                            page locked */
 239                         encrypted:1,    /* encrypted for secure swap (O) */
 240                         encrypted_cleaning:1,   /* encrypting page */
 241                         list_req_pending:1, /* pagein/pageout alt mechanism */
 242                                             /* allows creation of list      */
 243                                             /* requests on pages that are   */
 244                                             /* actively being paged.        */
 245                         dump_cleaning:1,   /* set by the pageout daemon when */
 246                                            /* a page being cleaned is       */
 247                                            /* encountered and targeted as   */
 248                                            /* a pageout candidate           */
 249                         cs_validated:1,    /* code-signing: page was checked */
 250                         cs_tainted:1,      /* code-signing: page is tainted */
 251                         reusable:1,
 252                         lopage:1,
 253                         slid:1,
 254                         __unused_object_bits:7;  /* 7 bits available here */
 255
 256 #if __LP64__
 257         unsigned int __unused_padding;  /* Pad structure explicitly
 258                                         * to 8-byte multiple for LP64 */
 259 #endif
 260 };
 261
 262 #define DEBUG_ENCRYPTED_SWAP    1
 263 #if DEBUG_ENCRYPTED_SWAP
 264 #define ASSERT_PAGE_DECRYPTED(page)                                     \
 265         MACRO_BEGIN                                                     \
 266         if ((page)->encrypted) {                                        \
 267                 panic("VM page %p should not be encrypted here\n",      \
 268                       (page));                                          \
 269         }                                                               \
 270         MACRO_END
 271 #else   /* DEBUG_ENCRYPTED_SWAP */
 272 #define ASSERT_PAGE_DECRYPTED(page) assert(!(page)->encrypted)
 273 #endif  /* DEBUG_ENCRYPTED_SWAP */
 274
 275 typedef struct vm_page  *vm_page_t;
 276
 277
 278 typedef struct vm_locks_array {
 279         char    pad  __attribute__ ((aligned (64)));
 280         lck_mtx_t       vm_page_queue_lock2 __attribute__ ((aligned (64)));
 281         lck_mtx_t       vm_page_queue_free_lock2 __attribute__ ((aligned (64)));
 282         char    pad2  __attribute__ ((aligned (64)));
 283 } vm_locks_array_t;
 284
 285
 286 #define VM_PAGE_WIRED(m)        ((!(m)->local && (m)->wire_count))
 287 #define VM_PAGE_NULL            ((vm_page_t) 0)
 288 #define NEXT_PAGE(m)            ((vm_page_t) (m)->pageq.next)
 289 #define NEXT_PAGE_PTR(m)        ((vm_page_t *) &(m)->pageq.next)
 290
 291 /*
 292  * XXX  The unusual bit should not be necessary.  Most of the bit
 293  * XXX  fields above really want to be masks.
 294  */
 295
 296 /*
 297  *      For debugging, this macro can be defined to perform
 298  *      some useful check on a page structure.
 299  */
 300
 301 #define VM_PAGE_CHECK(mem)                      \
 302         MACRO_BEGIN                             \
 303         VM_PAGE_QUEUES_ASSERT(mem, 1);          \
 304         MACRO_END
 305
 306 /*     Page coloring:
 307  *
 308  *     The free page list is actually n lists, one per color,
 309  *     where the number of colors is a function of the machine's
 310  *     cache geometry set at system initialization.  To disable
 311  *     coloring, set vm_colors to 1 and vm_color_mask to 0.
 312  *     The boot-arg "colors" may be used to override vm_colors.
 313  *     Note that there is little harm in having more colors than needed.
 314  */
 315
 316 #define MAX_COLORS      128
 317 #define DEFAULT_COLORS  32
 318
 319 extern
 320 unsigned int    vm_colors;              /* must be in range 1..MAX_COLORS */
 321 extern
 322 unsigned int    vm_color_mask;          /* must be (vm_colors-1) */
 323 extern
 324 unsigned int    vm_cache_geometry_colors; /* optimal #colors based on cache geometry */
 325
 326 /*
 327  * Wired memory is a very limited resource and we can't let users exhaust it
 328  * and deadlock the entire system.  We enforce the following limits:
 329  *
 330  * vm_user_wire_limit (default: all memory minus vm_global_no_user_wire_amount)
 331  *      how much memory can be user-wired in one user task
 332  *
 333  * vm_global_user_wire_limit (default: same as vm_user_wire_limit)
 334  *      how much memory can be user-wired in all user tasks
 335  *
 336  * vm_global_no_user_wire_amount (default: VM_NOT_USER_WIREABLE)
 337  *      how much memory must remain user-unwired at any time
 338  */
 339 #define VM_NOT_USER_WIREABLE (64*1024*1024)     /* 64MB */
 340 extern
 341 vm_map_size_t   vm_user_wire_limit;
 342 extern
 343 vm_map_size_t   vm_global_user_wire_limit;
 344 extern
 345 vm_map_size_t   vm_global_no_user_wire_amount;
 346
 347 /*
 348  *      Each pageable resident page falls into one of three lists:
 349  *
 350  *      free
 351  *              Available for allocation now.  The free list is
 352  *              actually an array of lists, one per color.
 353  *      inactive
 354  *              Not referenced in any map, but still has an
 355  *              object/offset-page mapping, and may be dirty.
 356  *              This is the list of pages that should be
 357  *              paged out next.  There are actually two
 358  *              inactive lists, one for pages brought in from
 359  *              disk or other backing store, and another
 360  *              for "zero-filled" pages.  See vm_pageout_scan()
 361  *              for the distinction and usage.
 362  *      active
 363  *              A list of pages which have been placed in
 364  *              at least one physical map.  This list is
 365  *              ordered, in LRU-like fashion.
 366  */
 367
 368
 369 #define VPL_LOCK_SPIN 1
 370
 371 struct vpl {
 372         unsigned int    vpl_count;
 373         queue_head_t    vpl_queue;
 374 #ifdef  VPL_LOCK_SPIN
 375         lck_spin_t      vpl_lock;
 376 #else
 377         lck_mtx_t       vpl_lock;
 378         lck_mtx_ext_t   vpl_lock_ext;
 379 #endif
 380 };
 381
 382 struct  vplq {
 383         union {
 384                 char   cache_line_pad[128];
 385                 struct vpl vpl;
 386         } vpl_un;
 387 };
 388 extern
 389 unsigned int    vm_page_local_q_count;
 390 extern
 391 struct vplq     *vm_page_local_q;
 392 extern
 393 unsigned int    vm_page_local_q_soft_limit;
 394 extern
 395 unsigned int    vm_page_local_q_hard_limit;
 396 extern
 397 vm_locks_array_t vm_page_locks;
 398
 399 extern
 400 queue_head_t    vm_page_queue_free[MAX_COLORS]; /* memory free queue */
 401 extern
 402 queue_head_t    vm_lopage_queue_free;           /* low memory free queue */
 403 extern
 404 queue_head_t    vm_page_queue_active;   /* active memory queue */
 405 extern
 406 queue_head_t    vm_page_queue_inactive; /* inactive memory queue for normal pages */
 407 extern
 408 queue_head_t    vm_page_queue_zf;       /* inactive memory queue for zero fill */
 409 extern
 410 queue_head_t    vm_page_queue_throttled;        /* memory queue for throttled pageout pages */
 411
 412 extern
 413 vm_offset_t     first_phys_addr;        /* physical address for first_page */
 414 extern
 415 vm_offset_t     last_phys_addr;         /* physical address for last_page */
 416
 417 extern
 418 unsigned int    vm_page_free_count;     /* How many pages are free? (sum of all colors) */
 419 extern
 420 unsigned int    vm_page_fictitious_count;/* How many fictitious pages are free? */
 421 extern
 422 unsigned int    vm_page_active_count;   /* How many pages are active? */
 423 extern
 424 unsigned int    vm_page_inactive_count; /* How many pages are inactive? */
 425 extern
 426 unsigned int    vm_page_throttled_count;/* How many inactives are throttled */
 427 extern
 428 unsigned int    vm_page_speculative_count;      /* How many speculative pages are unclaimed? */
 429 extern
 430 unsigned int    vm_page_wire_count;     /* How many pages are wired? */
 431 extern
 432 unsigned int    vm_page_free_target;    /* How many do we want free? */
 433 extern
 434 unsigned int    vm_page_free_min;       /* When to wakeup pageout */
 435 extern
 436 unsigned int    vm_page_throttle_limit; /* When to throttle new page creation */
 437 extern
 438 uint32_t        vm_page_creation_throttle;      /* When to throttle new page creation */
 439 extern
 440 unsigned int    vm_page_inactive_target;/* How many do we want inactive? */
 441 extern
 442 unsigned int    vm_page_inactive_min;   /* When do wakeup pageout */
 443 extern
 444 unsigned int    vm_page_free_reserved;  /* How many pages reserved to do pageout */
 445 extern
 446 unsigned int    vm_page_throttle_count; /* Count of page allocations throttled */
 447 extern
 448 unsigned int    vm_page_gobble_count;
 449
 450 #if DEVELOPMENT || DEBUG
 451 extern
 452 unsigned int    vm_page_speculative_used;
 453 #endif
 454
 455 extern
 456 unsigned int    vm_page_purgeable_count;/* How many pages are purgeable now ? */
 457 extern
 458 unsigned int    vm_page_purgeable_wired_count;/* How many purgeable pages are wired now ? */
 459 extern
 460 uint64_t        vm_page_purged_count;   /* How many pages got purged so far ? */
 461
 462 extern unsigned int     vm_page_free_wanted;
 463                                 /* how many threads are waiting for memory */
 464
 465 extern unsigned int     vm_page_free_wanted_privileged;
 466                                 /* how many VM privileged threads are waiting for memory */
 467
 468 extern ppnum_t  vm_page_fictitious_addr;
 469                                 /* (fake) phys_addr of fictitious pages */
 470
 471 extern ppnum_t  vm_page_guard_addr;
 472                                 /* (fake) phys_addr of guard pages */
 473
 474
 475 extern boolean_t        vm_page_deactivate_hint;
 476
 477 /*
 478    0 = all pages avail ( default. )
 479    1 = disable high mem ( cap max pages to 4G)
 480    2 = prefer himem
 481 */
 482 extern int              vm_himemory_mode;
 483
 484 extern boolean_t        vm_lopage_needed;
 485 extern uint32_t         vm_lopage_free_count;
 486 extern uint32_t         vm_lopage_free_limit;
 487 extern uint32_t         vm_lopage_lowater;
 488 extern boolean_t        vm_lopage_refill;
 489 extern uint64_t         max_valid_dma_address;
 490 extern ppnum_t          max_valid_low_ppnum;
 491
 492 /*
 493  * Prototypes for functions exported by this module.
 494  */
 495 extern void             vm_page_bootstrap(
 496                                         vm_offset_t     *startp,
 497                                         vm_offset_t     *endp) __attribute__((section("__TEXT, initcode")));
 498
 499 extern void             vm_page_module_init(void) __attribute__((section("__TEXT, initcode")));
 500
 501 extern void             vm_page_init_local_q(void);
 502
 503 extern void             vm_page_create(
 504                                         ppnum_t         start,
 505                                         ppnum_t         end);
 506
 507 extern vm_page_t        vm_page_lookup(
 508                                         vm_object_t             object,
 509                                         vm_object_offset_t      offset);
 510
 511 extern vm_page_t        vm_page_grab_fictitious(void);
 512
 513 extern vm_page_t        vm_page_grab_guard(void);
 514
 515 extern void             vm_page_release_fictitious(
 516                                         vm_page_t page);
 517
 518 extern void             vm_page_more_fictitious(void);
 519
 520 extern int              vm_pool_low(void);
 521
 522 extern vm_page_t        vm_page_grab(void);
 523
 524 extern vm_page_t        vm_page_grablo(void);
 525
 526 extern void             vm_page_release(
 527                                         vm_page_t       page);
 528
 529 extern boolean_t        vm_page_wait(
 530                                         int             interruptible );
 531
 532 extern vm_page_t        vm_page_alloc(
 533                                         vm_object_t             object,
 534                                         vm_object_offset_t      offset);
 535
 536 extern vm_page_t        vm_page_alloclo(
 537                                         vm_object_t             object,
 538                                         vm_object_offset_t      offset);
 539
 540 extern vm_page_t        vm_page_alloc_guard(
 541         vm_object_t             object,
 542         vm_object_offset_t      offset);
 543
 544 extern void             vm_page_init(
 545                                         vm_page_t       page,
 546                                         ppnum_t         phys_page,
 547                                         boolean_t       lopage);
 548
 549 extern void             vm_page_free(
 550                                         vm_page_t       page);
 551
 552 extern void             vm_page_free_unlocked(
 553                                         vm_page_t       page,
 554                                         boolean_t       remove_from_hash);
 555
 556 extern void             vm_page_activate(
 557                                         vm_page_t       page);
 558
 559 extern void             vm_page_deactivate(
 560                                         vm_page_t       page);
 561
 562 extern void             vm_page_deactivate_internal(
 563                                         vm_page_t       page,
 564                                         boolean_t       clear_hw_reference);
 565
 566 extern void             vm_page_lru(
 567                                         vm_page_t       page);
 568
 569 extern void             vm_page_speculate(
 570                                         vm_page_t       page,
 571                                         boolean_t       new);
 572
 573 extern void             vm_page_speculate_ageit(
 574                                         struct vm_speculative_age_q *aq);
 575
 576 extern void             vm_page_reactivate_all_throttled(void);
 577
 578 extern void             vm_page_reactivate_local(uint32_t lid, boolean_t force, boolean_t nolocks);
 579
 580 extern void             vm_page_rename(
 581                                         vm_page_t               page,
 582                                         vm_object_t             new_object,
 583                                         vm_object_offset_t      new_offset,
 584                                         boolean_t               encrypted_ok);
 585
 586 extern void             vm_page_insert(
 587                                         vm_page_t               page,
 588                                         vm_object_t             object,
 589                                         vm_object_offset_t      offset);
 590
 591 extern void             vm_page_insert_internal(
 592                                         vm_page_t               page,
 593                                         vm_object_t             object,
 594                                         vm_object_offset_t      offset,
 595                                         boolean_t               queues_lock_held,
 596                                         boolean_t               insert_in_hash);
 597
 598 extern void             vm_page_replace(
 599                                         vm_page_t               mem,
 600                                         vm_object_t             object,
 601                                         vm_object_offset_t      offset);
 602
 603 extern void             vm_page_remove(
 604                                         vm_page_t       page,
 605                                         boolean_t       remove_from_hash);
 606
 607 extern void             vm_page_zero_fill(
 608                                         vm_page_t       page);
 609
 610 extern void             vm_page_part_zero_fill(
 611                                         vm_page_t       m,
 612                                         vm_offset_t     m_pa,
 613                                         vm_size_t       len);
 614
 615 extern void             vm_page_copy(
 616                                         vm_page_t       src_page,
 617                                         vm_page_t       dest_page);
 618
 619 extern void             vm_page_part_copy(
 620                                         vm_page_t       src_m,
 621                                         vm_offset_t     src_pa,
 622                                         vm_page_t       dst_m,
 623                                         vm_offset_t     dst_pa,
 624                                         vm_size_t       len);
 625
 626 extern void             vm_page_wire(
 627                                         vm_page_t       page);
 628
 629 extern void             vm_page_unwire(
 630                                         vm_page_t       page,
 631                                         boolean_t       queueit);
 632
 633 extern void             vm_set_page_size(void);
 634
 635 extern void             vm_page_gobble(
 636                                         vm_page_t      page);
 637
 638 extern void             vm_page_validate_cs(vm_page_t   page);
 639 extern void             vm_page_validate_cs_mapped(
 640         vm_page_t       page,
 641         const void      *kaddr);
 642
 643 extern void             vm_page_free_prepare_queues(
 644                                         vm_page_t       page);
 645
 646 extern void             vm_page_free_prepare_object(
 647                                         vm_page_t       page,
 648                                         boolean_t       remove_from_hash);
 649
 650 extern void             vm_check_memorystatus(void);
 651
 652
 653 /*
 654  *      Functions implemented as macros. m->wanted and m->busy are
 655  *      protected by the object lock.
 656  */
 657
 658 #define PAGE_ASSERT_WAIT(m, interruptible)                      \
 659                 (((m)->wanted = TRUE),                          \
 660                  assert_wait((event_t) (m), (interruptible)))
 661
 662 #define PAGE_SLEEP(o, m, interruptible)                         \
 663                 (((m)->wanted = TRUE),                          \
 664                  thread_sleep_vm_object((o), (m), (interruptible)))
 665
 666 #define PAGE_WAKEUP_DONE(m)                                     \
 667                 MACRO_BEGIN                                     \
 668                 (m)->busy = FALSE;                              \
 669                 if ((m)->wanted) {                              \
 670                         (m)->wanted = FALSE;                    \
 671                         thread_wakeup((event_t) (m));           \
 672                 }                                               \
 673                 MACRO_END
 674
 675 #define PAGE_WAKEUP(m)                                          \
 676                 MACRO_BEGIN                                     \
 677                 if ((m)->wanted) {                              \
 678                         (m)->wanted = FALSE;                    \
 679                         thread_wakeup((event_t) (m));           \
 680                 }                                               \
 681                 MACRO_END
 682
 683 #define VM_PAGE_FREE(p)                         \
 684                 MACRO_BEGIN                     \
 685                 vm_page_free_unlocked(p, TRUE); \
 686                 MACRO_END
 687
 688 #define VM_PAGE_GRAB_FICTITIOUS(M)                                      \
 689                 MACRO_BEGIN                                             \
 690                 while ((M = vm_page_grab_fictitious()) == VM_PAGE_NULL) \
 691                         vm_page_more_fictitious();                      \
 692                 MACRO_END
 693
 694 #define VM_PAGE_WAIT()          ((void)vm_page_wait(THREAD_UNINT))
 695
 696 #define vm_page_queue_lock (vm_page_locks.vm_page_queue_lock2)
 697 #define vm_page_queue_free_lock (vm_page_locks.vm_page_queue_free_lock2)
 698
 699 #define vm_page_lock_queues()   lck_mtx_lock(&vm_page_queue_lock)
 700 #define vm_page_unlock_queues() lck_mtx_unlock(&vm_page_queue_lock)
 701
 702 #define vm_page_lockspin_queues()       lck_mtx_lock_spin(&vm_page_queue_lock)
 703 #define vm_page_trylockspin_queues()    lck_mtx_try_lock_spin(&vm_page_queue_lock)
 704 #define vm_page_lockconvert_queues()    lck_mtx_convert_spin(&vm_page_queue_lock)
 705
 706 #ifdef  VPL_LOCK_SPIN
 707 #define VPL_LOCK_INIT(vlq, vpl_grp, vpl_attr) lck_spin_init(&vlq->vpl_lock, vpl_grp, vpl_attr)
 708 #define VPL_LOCK(vpl) lck_spin_lock(vpl)
 709 #define VPL_UNLOCK(vpl) lck_spin_unlock(vpl)
 710 #else
 711 #define VPL_LOCK_INIT(vlq, vpl_grp, vpl_attr) lck_mtx_init_ext(&vlq->vpl_lock, &vlq->vpl_lock_ext, vpl_grp, vpl_attr)
 712 #define VPL_LOCK(vpl) lck_mtx_lock_spin(vpl)
 713 #define VPL_UNLOCK(vpl) lck_mtx_unlock(vpl)
 714 #endif
 715
 716 #if MACH_ASSERT
 717 extern void vm_page_queues_assert(vm_page_t mem, int val);
 718 #define VM_PAGE_QUEUES_ASSERT(mem, val) vm_page_queues_assert((mem), (val))
 719 #else
 720 #define VM_PAGE_QUEUES_ASSERT(mem, val)
 721 #endif
 722
 723
 724 /*
 725  * 'vm_fault_enter' will place newly created pages (zero-fill and COW) onto the
 726  * local queues if they exist... its the only spot in the system where we add pages
 727  * to those queues...  once on those queues, those pages can only move to one of the
 728  * global page queues or the free queues... they NEVER move from local q to local q.
 729  * the 'local' state is stable when VM_PAGE_QUEUES_REMOVE is called since we're behind
 730  * the global vm_page_queue_lock at this point...  we still need to take the local lock
 731  * in case this operation is being run on a different CPU then the local queue's identity,
 732  * but we don't have to worry about the page moving to a global queue or becoming wired
 733  * while we're grabbing the local lock since those operations would require the global
 734  * vm_page_queue_lock to be held, and we already own it.
 735  *
 736  * this is why its safe to utilze the wire_count field in the vm_page_t as the local_id...
 737  * 'wired' and local are ALWAYS mutually exclusive conditions.
 738  */
 739 #define VM_PAGE_QUEUES_REMOVE(mem)                              \
 740         MACRO_BEGIN                                             \
 741         VM_PAGE_QUEUES_ASSERT(mem, 1);                          \
 742         assert(!mem->laundry);                                  \
 743         assert(!mem->pageout_queue);                            \
 744         if (mem->local) {                                       \
 745                 struct vpl      *lq;                            \
 746                 assert(mem->object != kernel_object);           \
 747                 assert(!mem->inactive && !mem->speculative);    \
 748                 assert(!mem->active && !mem->throttled);        \
 749                 assert(!mem->fictitious);                       \
 750                 lq = &vm_page_local_q[mem->local_id].vpl_un.vpl;        \
 751                 VPL_LOCK(&lq->vpl_lock);                        \
 752                 queue_remove(&lq->vpl_queue,                    \
 753                              mem, vm_page_t, pageq);            \
 754                 mem->local = FALSE;                             \
 755                 mem->local_id = 0;                              \
 756                 lq->vpl_count--;                                \
 757                 VPL_UNLOCK(&lq->vpl_lock);                      \
 758         }                                                       \
 759                                                                 \
 760         else if (mem->active) {                                 \
 761                 assert(mem->object != kernel_object);           \
 762                 assert(!mem->inactive && !mem->speculative);    \
 763                 assert(!mem->throttled);                        \
 764                 assert(!mem->fictitious);                       \
 765                 queue_remove(&vm_page_queue_active,             \
 766                         mem, vm_page_t, pageq);                 \
 767                 mem->active = FALSE;                            \
 768                 vm_page_active_count--;                         \
 769         }                                                       \
 770                                                                 \
 771         else if (mem->inactive) {                               \
 772                 assert(mem->object != kernel_object);           \
 773                 assert(!mem->active && !mem->speculative);      \
 774                 assert(!mem->throttled);                        \
 775                 assert(!mem->fictitious);                       \
 776                 if (mem->zero_fill) {                           \
 777                         queue_remove(&vm_page_queue_zf,         \
 778                         mem, vm_page_t, pageq);                 \
 779                         vm_zf_queue_count--;                    \
 780                 } else {                                        \
 781                         queue_remove(&vm_page_queue_inactive,   \
 782                         mem, vm_page_t, pageq);                 \
 783                 }                                               \
 784                 mem->inactive = FALSE;                          \
 785                 vm_page_inactive_count--;                       \
 786                 vm_purgeable_q_advance_all();                   \
 787         }                                                       \
 788                                                                 \
 789         else if (mem->throttled) {                              \
 790                 assert(!mem->active && !mem->inactive);         \
 791                 assert(!mem->speculative);                      \
 792                 assert(!mem->fictitious);                       \
 793                 queue_remove(&vm_page_queue_throttled,          \
 794                              mem, vm_page_t, pageq);            \
 795                 mem->throttled = FALSE;                         \
 796                 vm_page_throttled_count--;                      \
 797         }                                                       \
 798                                                                 \
 799         else if (mem->speculative) {                            \
 800                 assert(!mem->active && !mem->inactive);         \
 801                 assert(!mem->throttled);                        \
 802                 assert(!mem->fictitious);                       \
 803                 remque(&mem->pageq);                            \
 804                 mem->speculative = FALSE;                       \
 805                 vm_page_speculative_count--;                    \
 806         }                                                       \
 807                                                                 \
 808         else if (mem->pageq.next || mem->pageq.prev)            \
 809                 panic("VM_PAGE_QUEUES_REMOVE: unmarked page on Q");     \
 810         mem->pageq.next = NULL;                                 \
 811         mem->pageq.prev = NULL;                                 \
 812         VM_PAGE_QUEUES_ASSERT(mem, 0);                          \
 813         MACRO_END
 814
 815
 816 #define VM_PAGE_ENQUEUE_INACTIVE(mem, first)                    \
 817         MACRO_BEGIN                                             \
 818         VM_PAGE_QUEUES_ASSERT(mem, 0);                          \
 819         assert(!mem->fictitious);                               \
 820         assert(!mem->laundry);                                  \
 821         assert(!mem->pageout_queue);                            \
 822         if (mem->zero_fill) {                                   \
 823                 if (first == TRUE)                              \
 824                         queue_enter_first(&vm_page_queue_zf, mem, vm_page_t, pageq);    \
 825                 else                                            \
 826                         queue_enter(&vm_page_queue_zf, mem, vm_page_t, pageq);          \
 827                 vm_zf_queue_count++;                            \
 828         } else {                                                \
 829                 if (first == TRUE)                              \
 830                         queue_enter_first(&vm_page_queue_inactive, mem, vm_page_t, pageq); \
 831                 else                                            \
 832                         queue_enter(&vm_page_queue_inactive, mem, vm_page_t, pageq);    \
 833         }                                                       \
 834         mem->inactive = TRUE;                                   \
 835         vm_page_inactive_count++;                               \
 836         token_new_pagecount++;                                  \
 837         MACRO_END
 838
 839
 840 #if DEVELOPMENT || DEBUG
 841 #define VM_PAGE_SPECULATIVE_USED_ADD()                          \
 842         MACRO_BEGIN                                             \
 843         OSAddAtomic(1, &vm_page_speculative_used);      \
 844         MACRO_END
 845 #else
 846 #define VM_PAGE_SPECULATIVE_USED_ADD()
 847 #endif
 848
 849
 850 #define VM_PAGE_CONSUME_CLUSTERED(mem)                          \
 851         MACRO_BEGIN                                             \
 852         if (mem->clustered) {                                   \
 853                 assert(mem->object);                            \
 854                 mem->object->pages_used++;                      \
 855                 mem->clustered = FALSE;                         \
 856                 VM_PAGE_SPECULATIVE_USED_ADD();                 \
 857         }                                                       \
 858         MACRO_END
 859
 860
 861
 862 #define DW_vm_page_unwire               0x01
 863 #define DW_vm_page_wire                 0x02
 864 #define DW_vm_page_free                 0x04
 865 #define DW_vm_page_activate             0x08
 866 #define DW_vm_page_deactivate_internal  0x10
 867 #define DW_vm_page_speculate            0x20
 868 #define DW_vm_page_lru                  0x40
 869 #define DW_vm_pageout_throttle_up       0x80
 870 #define DW_PAGE_WAKEUP                  0x100
 871 #define DW_clear_busy                   0x200
 872 #define DW_clear_reference              0x400
 873 #define DW_set_reference                0x800
 874 #define DW_move_page                    0x1000
 875 #define DW_VM_PAGE_QUEUES_REMOVE        0x2000
 876 #define DW_set_list_req_pending         0x4000
 877
 878 struct vm_page_delayed_work {
 879         vm_page_t       dw_m;
 880         int             dw_mask;
 881 };
 882
 883 void vm_page_do_delayed_work(vm_object_t object, struct vm_page_delayed_work *dwp, int dw_count);
 884
 885 extern unsigned int vm_max_delayed_work_limit;
 886
 887 #define DEFAULT_DELAYED_WORK_LIMIT      32
 888
 889 #define DELAYED_WORK_LIMIT(max) ((vm_max_delayed_work_limit >= max ? max : vm_max_delayed_work_limit))
 890
 891 /*
 892  * vm_page_do_delayed_work may need to drop the object lock...
 893  * if it does, we need the pages it's looking at to
 894  * be held stable via the busy bit, so if busy isn't already
 895  * set, we need to set it and ask vm_page_do_delayed_work
 896  * to clear it and wakeup anyone that might have blocked on
 897  * it once we're done processing the page.
 898  *
 899  * additionally, we can't call vm_page_do_delayed_work with
 900  * list_req_pending == TRUE since it may need to
 901  * drop the object lock before dealing
 902  * with this page and because list_req_pending == TRUE,
 903  * busy == TRUE will NOT protect this page from being stolen
 904  * so clear list_req_pending and ask vm_page_do_delayed_work
 905  * to re-set it once it holds both the pageq and object locks
 906  */
 907
 908 #define VM_PAGE_ADD_DELAYED_WORK(dwp, mem, dw_cnt)              \
 909         MACRO_BEGIN                                             \
 910         if (mem->busy == FALSE) {                               \
 911                 mem->busy = TRUE;                               \
 912                 if ( !(dwp->dw_mask & DW_vm_page_free))         \
 913                         dwp->dw_mask |= (DW_clear_busy | DW_PAGE_WAKEUP); \
 914         }                                                       \
 915         if (mem->list_req_pending) {                            \
 916                 mem->list_req_pending = FALSE;                  \
 917                 dwp->dw_mask |= DW_set_list_req_pending;        \
 918         }                                                       \
 919         dwp->dw_m = mem;                                        \
 920         dwp++;                                                  \
 921         dw_count++;                                             \
 922         MACRO_END
 923
 924 extern vm_page_t vm_object_page_grab(vm_object_t);
 925
 926
 927 #endif  /* _VM_VM_PAGE_H_ */