sys/vm/vm_page.h

   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. Neither the name of the University nor the names of its contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  *
  32  *      from: @(#)vm_page.h     8.2 (Berkeley) 12/13/93
  33  *
  34  *
  35  * Copyright (c) 1987, 1990 Carnegie-Mellon University.
  36  * All rights reserved.
  37  *
  38  * Authors: Avadis Tevanian, Jr., Michael Wayne Young
  39  *
  40  * Permission to use, copy, modify and distribute this software and
  41  * its documentation is hereby granted, provided that both the copyright
  42  * notice and this permission notice appear in all copies of the
  43  * software, derivative works or modified versions, and any portions
  44  * thereof, and that both notices appear in supporting documentation.
  45  *
  46  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
  47  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
  48  * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
  49  *
  50  * Carnegie Mellon requests users of this software to return to
  51  *
  52  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
  53  *  School of Computer Science
  54  *  Carnegie Mellon University
  55  *  Pittsburgh PA 15213-3890
  56  *
  57  * any improvements or extensions that they make and grant Carnegie the
  58  * rights to redistribute these changes.
  59  *
  60  * $FreeBSD: src/sys/vm/vm_page.h,v 1.75.2.8 2002/03/06 01:07:09 dillon Exp $
  61  */
  62
  63 /*
  64  *      Resident memory system definitions.
  65  */
  66
  67 #ifndef _VM_VM_PAGE_H_
  68 #define _VM_VM_PAGE_H_
  69
  70 #ifndef _SYS_TYPES_H_
  71 #include <sys/types.h>
  72 #endif
  73 #ifndef _SYS_TREE_H_
  74 #include <sys/tree.h>
  75 #endif
  76 #ifndef _MACHINE_PMAP_H_
  77 #include <machine/pmap.h>
  78 #endif
  79 #ifndef _VM_PMAP_H_
  80 #include <vm/pmap.h>
  81 #endif
  82 #include <machine/atomic.h>
  83
  84 #ifdef _KERNEL
  85
  86 #ifndef _SYS_SYSTM_H_
  87 #include <sys/systm.h>
  88 #endif
  89 #ifndef _SYS_SPINLOCK_H_
  90 #include <sys/spinlock.h>
  91 #endif
  92 #ifndef _SYS_THREAD2_H_
  93 #include <sys/thread2.h>
  94 #endif
  95
  96 #ifdef __x86_64__
  97 #include <machine/vmparam.h>
  98 #endif
  99
 100 #endif
 101
 102 /*
 103  *      Management of resident (logical) pages.
 104  *
 105  *      A small structure is kept for each resident
 106  *      page, indexed by page number.  Each structure
 107  *      is an element of several lists:
 108  *
 109  *              A hash table bucket used to quickly
 110  *              perform object/offset lookups
 111  *
 112  *              A list of all pages for a given object,
 113  *              so they can be quickly deactivated at
 114  *              time of deallocation.
 115  *
 116  *              An ordered list of pages due for pageout.
 117  *
 118  *      In addition, the structure contains the object
 119  *      and offset to which this page belongs (for pageout),
 120  *      and sundry status bits.
 121  *
 122  *      Fields in this structure are locked either by the lock on the
 123  *      object that the page belongs to (O) or by the lock on the page
 124  *      queues (P).
 125  *
 126  *      The 'valid' and 'dirty' fields are distinct.  A page may have dirty
 127  *      bits set without having associated valid bits set.  This is used by
 128  *      NFS to implement piecemeal writes.
 129  */
 130
 131 TAILQ_HEAD(pglist, vm_page);
 132
 133 struct vm_object;
 134
 135 int rb_vm_page_compare(struct vm_page *, struct vm_page *);
 136
 137 struct vm_page_rb_tree;
 138 RB_PROTOTYPE2(vm_page_rb_tree, vm_page, rb_entry, rb_vm_page_compare, vm_pindex_t);
 139
 140 struct vm_page {
 141         TAILQ_ENTRY(vm_page) pageq;     /* vm_page_queues[] list (P)    */
 142         RB_ENTRY(vm_page) rb_entry;     /* Red-Black tree based at object */
 143         struct spinlock spin;
 144
 145         struct vm_object *object;       /* which object am I in (O,P)*/
 146         vm_pindex_t pindex;             /* offset into object (O,P) */
 147         vm_paddr_t phys_addr;           /* physical address of page */
 148         struct md_page md;              /* machine dependant stuff */
 149         u_short queue;                  /* page queue index */
 150         u_short pc;                     /* page color */
 151         u_char  act_count;              /* page usage count */
 152         u_char  busy;                   /* page busy count */
 153         u_char  pat_mode;               /* hardware page attribute */
 154         u_char  unused02;
 155         u_int32_t flags;                /* see below */
 156         u_int   wire_count;             /* wired down maps refs (P) */
 157         int     hold_count;             /* page hold count */
 158
 159         /*
 160          * NOTE that these must support one bit per DEV_BSIZE in a page!!!
 161          * so, on normal X86 kernels, they must be at least 8 bits wide.
 162          */
 163         u_char  valid;                  /* map of valid DEV_BSIZE chunks */
 164         u_char  dirty;                  /* map of dirty DEV_BSIZE chunks */
 165
 166         int     ku_pagecnt;             /* kmalloc helper */
 167 #ifdef VM_PAGE_DEBUG
 168         const char *busy_func;
 169         int     busy_line;
 170 #endif
 171 };
 172
 173 #ifdef VM_PAGE_DEBUG
 174 #define VM_PAGE_DEBUG_EXT(name) name ## _debug
 175 #define VM_PAGE_DEBUG_ARGS      , const char *func, int lineno
 176 #else
 177 #define VM_PAGE_DEBUG_EXT(name) name
 178 #define VM_PAGE_DEBUG_ARGS
 179 #endif
 180
 181 #ifndef __VM_PAGE_T_DEFINED__
 182 #define __VM_PAGE_T_DEFINED__
 183 typedef struct vm_page *vm_page_t;
 184 #endif
 185
 186 /*
 187  * Page coloring parameters.  We use generous parameters designed to
 188  * statistically spread pages over available cpu cache space.  This has
 189  * become less important over time as cache associativity is higher
 190  * in modern times but we still use the core algorithm to help reduce
 191  * lock contention between cpus.
 192  *
 193  * Page coloring cannot be disabled.
 194  */
 195
 196 #define PQ_PRIME1 31    /* Prime number somewhat less than PQ_HASH_SIZE */
 197 #define PQ_PRIME2 23    /* Prime number somewhat less than PQ_HASH_SIZE */
 198 #define PQ_L2_SIZE 512  /* A number of colors opt for 1M cache */
 199 #define PQ_L2_MASK      (PQ_L2_SIZE - 1)
 200
 201 #define PQ_NONE         0
 202 #define PQ_FREE         (1 + 0*PQ_L2_SIZE)
 203 #define PQ_INACTIVE     (1 + 1*PQ_L2_SIZE)
 204 #define PQ_ACTIVE       (1 + 2*PQ_L2_SIZE)
 205 #define PQ_CACHE        (1 + 3*PQ_L2_SIZE)
 206 #define PQ_HOLD         (1 + 4*PQ_L2_SIZE)
 207 #define PQ_COUNT        (1 + 5*PQ_L2_SIZE)
 208
 209 /*
 210  * Scan support
 211  */
 212 struct vm_map;
 213
 214 struct rb_vm_page_scan_info {
 215         vm_pindex_t     start_pindex;
 216         vm_pindex_t     end_pindex;
 217         int             limit;
 218         int             desired;
 219         int             error;
 220         int             pagerflags;
 221         int             count;
 222         int             unused01;
 223         vm_offset_t     addr;
 224         vm_pindex_t     backing_offset_index;
 225         struct vm_object *object;
 226         struct vm_object *backing_object;
 227         struct vm_page  *mpte;
 228         struct pmap     *pmap;
 229         struct vm_map   *map;
 230 };
 231
 232 int rb_vm_page_scancmp(struct vm_page *, void *);
 233
 234 struct vpgqueues {
 235         struct spinlock spin;
 236         struct pglist pl;
 237         int     cnt_offset;     /* offset into vmstats structure (int) */
 238         int     lcnt;
 239         int     flipflop;       /* probably not the best place */
 240         int     unused00;
 241         int     unused01;
 242         char    unused[64 - sizeof(struct pglist) -
 243                         sizeof(int *) - sizeof(int) * 4];
 244 };
 245
 246 extern struct vpgqueues vm_page_queues[PQ_COUNT];
 247
 248 /*
 249  * These are the flags defined for vm_page.
 250  *
 251  *  PG_UNMANAGED (used by OBJT_PHYS) indicates that the page is
 252  *  not under PV management but otherwise should be treated as a
 253  *  normal page.  Pages not under PV management cannot be paged out
 254  *  via the object/vm_page_t because there is no knowledge of their
 255  *  pte mappings, nor can they be removed from their objects via
 256  *  the object, and such pages are also not on any PQ queue.  The
 257  *  PG_MAPPED and PG_WRITEABLE flags are not applicable.
 258  *
 259  *  PG_MAPPED only applies to managed pages, indicating whether the page
 260  *  is mapped onto one or more pmaps.  A page might still be mapped to
 261  *  special pmaps in an unmanaged fashion, for example when mapped into a
 262  *  buffer cache buffer, without setting PG_MAPPED.
 263  *
 264  *  PG_WRITEABLE indicates that there may be a writeable managed pmap entry
 265  *  somewhere, and that the page can be dirtied by hardware at any time
 266  *  and may have to be tested for that.  The modified bit in unmanaged
 267  *  mappings or in the special clean map is not tested.
 268  *
 269  *  PG_SWAPPED indicates that the page is backed by a swap block.  Any
 270  *  VM object type other than OBJT_DEFAULT can have swap-backed pages now.
 271  *
 272  *  PG_SBUSY is set when m->busy != 0.  PG_SBUSY and m->busy are only modified
 273  *  when the page is PG_BUSY.
 274  */
 275 #define PG_BUSY         0x00000001      /* page is in transit (O) */
 276 #define PG_WANTED       0x00000002      /* someone is waiting for page (O) */
 277 #define PG_WINATCFLS    0x00000004      /* flush dirty page on inactive q */
 278 #define PG_FICTITIOUS   0x00000008      /* physical page doesn't exist (O) */
 279 #define PG_WRITEABLE    0x00000010      /* page is writeable */
 280 #define PG_MAPPED       0x00000020      /* page is mapped (managed) */
 281 #define PG_UNUSED0040   0x00000040
 282 #define PG_REFERENCED   0x00000080      /* page has been referenced */
 283 #define PG_CLEANCHK     0x00000100      /* page will be checked for cleaning */
 284 #define PG_SWAPINPROG   0x00000200      /* swap I/O in progress on page      */
 285 #define PG_NOSYNC       0x00000400      /* do not collect for syncer */
 286 #define PG_UNMANAGED    0x00000800      /* No PV management for page */
 287 #define PG_MARKER       0x00001000      /* special queue marker page */
 288 #define PG_RAM          0x00002000      /* read ahead mark */
 289 #define PG_SWAPPED      0x00004000      /* backed by swap */
 290 #define PG_NOTMETA      0x00008000      /* do not back with swap */
 291 #define PG_UNUSED10000  0x00010000
 292 #define PG_SBUSY        0x00020000      /* soft-busy also set */
 293 #define PG_NEED_COMMIT  0x00040000      /* clean page requires commit */
 294
 295 #define PG_KEEP_NEWPAGE_MASK    (PG_BUSY | PG_SBUSY | PG_WANTED)
 296
 297
 298 /*
 299  * Misc constants.
 300  */
 301
 302 #define ACT_DECLINE             1
 303 #define ACT_ADVANCE             3
 304 #define ACT_INIT                5
 305 #define ACT_MAX                 64
 306
 307 #ifdef _KERNEL
 308 /*
 309  * Each pageable resident page falls into one of four lists:
 310  *
 311  *      free
 312  *              Available for allocation now.
 313  *
 314  * The following are all LRU sorted:
 315  *
 316  *      cache
 317  *              Almost available for allocation. Still in an
 318  *              object, but clean and immediately freeable at
 319  *              non-interrupt times.
 320  *
 321  *      inactive
 322  *              Low activity, candidates for reclamation.
 323  *              This is the list of pages that should be
 324  *              paged out next.
 325  *
 326  *      active
 327  *              Pages that are "active" i.e. they have been
 328  *              recently referenced.
 329  *
 330  *      zero
 331  *              Pages that are really free and have been pre-zeroed
 332  *
 333  */
 334
 335 extern struct vm_page *vm_page_array;   /* First resident page in table */
 336 extern int vm_page_array_size;          /* number of vm_page_t's */
 337 extern long first_page;                 /* first physical page number */
 338
 339 #define VM_PAGE_TO_PHYS(entry)  \
 340                 ((entry)->phys_addr)
 341
 342 #define PHYS_TO_VM_PAGE(pa)     \
 343                 (&vm_page_array[atop(pa) - first_page])
 344
 345
 346 #if PAGE_SIZE == 4096
 347 #define VM_PAGE_BITS_ALL 0xff
 348 #endif
 349
 350 /*
 351  * Note: the code will always use nominally free pages from the free list
 352  * before trying other flag-specified sources.
 353  *
 354  * At least one of VM_ALLOC_NORMAL|VM_ALLOC_SYSTEM|VM_ALLOC_INTERRUPT
 355  * must be specified.  VM_ALLOC_RETRY may only be specified if VM_ALLOC_NORMAL
 356  * is also specified.
 357  */
 358 #define VM_ALLOC_NORMAL         0x0001  /* ok to use cache pages */
 359 #define VM_ALLOC_SYSTEM         0x0002  /* ok to exhaust most of free list */
 360 #define VM_ALLOC_INTERRUPT      0x0004  /* ok to exhaust entire free list */
 361 #define VM_ALLOC_ZERO           0x0008  /* req pre-zero'd memory if avail */
 362 #define VM_ALLOC_QUICK          0x0010  /* like NORMAL but do not use cache */
 363 #define VM_ALLOC_FORCE_ZERO     0x0020  /* zero page even if already valid */
 364 #define VM_ALLOC_NULL_OK        0x0040  /* ok to return NULL on collision */
 365 #define VM_ALLOC_RETRY          0x0080  /* indefinite block (vm_page_grab()) */
 366 #define VM_ALLOC_USE_GD         0x0100  /* use per-gd cache */
 367 #define VM_ALLOC_CPU_SPEC       0x0200
 368
 369 #define VM_ALLOC_CPU_SHIFT      16
 370 #define VM_ALLOC_CPU(n)         (((n) << VM_ALLOC_CPU_SHIFT) | \
 371                                  VM_ALLOC_CPU_SPEC)
 372 #define VM_ALLOC_GETCPU(flags)  ((flags) >> VM_ALLOC_CPU_SHIFT)
 373
 374 void vm_page_queue_spin_lock(vm_page_t);
 375 void vm_page_queues_spin_lock(u_short);
 376 void vm_page_and_queue_spin_lock(vm_page_t);
 377
 378 void vm_page_queue_spin_unlock(vm_page_t);
 379 void vm_page_queues_spin_unlock(u_short);
 380 void vm_page_and_queue_spin_unlock(vm_page_t m);
 381
 382 void vm_page_init(vm_page_t m);
 383 void vm_page_io_finish(vm_page_t m);
 384 void vm_page_io_start(vm_page_t m);
 385 void vm_page_need_commit(vm_page_t m);
 386 void vm_page_clear_commit(vm_page_t m);
 387 void vm_page_wakeup(vm_page_t m);
 388 void vm_page_hold(vm_page_t);
 389 void vm_page_unhold(vm_page_t);
 390 void vm_page_activate (vm_page_t);
 391
 392 vm_size_t vm_contig_avail_pages(void);
 393 vm_page_t vm_page_alloc (struct vm_object *, vm_pindex_t, int);
 394 vm_page_t vm_page_alloc_contig(vm_paddr_t low, vm_paddr_t high,
 395                      unsigned long alignment, unsigned long boundary,
 396                      unsigned long size, vm_memattr_t memattr);
 397
 398 vm_page_t vm_page_grab (struct vm_object *, vm_pindex_t, int);
 399 void vm_page_cache (vm_page_t);
 400 int vm_page_try_to_cache (vm_page_t);
 401 int vm_page_try_to_free (vm_page_t);
 402 void vm_page_dontneed (vm_page_t);
 403 void vm_page_deactivate (vm_page_t);
 404 void vm_page_deactivate_locked (vm_page_t);
 405 void vm_page_initfake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr);
 406 int vm_page_insert (vm_page_t, struct vm_object *, vm_pindex_t);
 407 vm_page_t vm_page_lookup (struct vm_object *, vm_pindex_t);
 408 vm_page_t VM_PAGE_DEBUG_EXT(vm_page_lookup_busy_wait)(
 409                 struct vm_object *, vm_pindex_t, int, const char *
 410                 VM_PAGE_DEBUG_ARGS);
 411 vm_page_t VM_PAGE_DEBUG_EXT(vm_page_lookup_busy_try)(
 412                 struct vm_object *, vm_pindex_t, int, int *
 413                 VM_PAGE_DEBUG_ARGS);
 414 vm_page_t vm_page_repurpose(struct vm_object *, vm_pindex_t, int, int *,
 415                 vm_page_t, int *, int *);
 416 void vm_page_remove (vm_page_t);
 417 void vm_page_rename (vm_page_t, struct vm_object *, vm_pindex_t);
 418 void vm_page_startup (void);
 419 void vm_numa_organize(vm_paddr_t ran_beg, vm_paddr_t bytes, int physid);
 420 void vm_page_unmanage (vm_page_t);
 421 void vm_page_unwire (vm_page_t, int);
 422 void vm_page_wire (vm_page_t);
 423 void vm_page_unqueue (vm_page_t);
 424 void vm_page_unqueue_nowakeup (vm_page_t);
 425 vm_page_t vm_page_next (vm_page_t);
 426 void vm_page_set_validclean (vm_page_t, int, int);
 427 void vm_page_set_validdirty (vm_page_t, int, int);
 428 void vm_page_set_valid (vm_page_t, int, int);
 429 void vm_page_set_dirty (vm_page_t, int, int);
 430 void vm_page_clear_dirty (vm_page_t, int, int);
 431 void vm_page_set_invalid (vm_page_t, int, int);
 432 int vm_page_is_valid (vm_page_t, int, int);
 433 void vm_page_test_dirty (vm_page_t);
 434 int vm_page_bits (int, int);
 435 vm_page_t vm_page_list_find(int basequeue, int index);
 436 void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid);
 437 void vm_page_free_toq(vm_page_t m);
 438 void vm_page_free_contig(vm_page_t m, unsigned long size);
 439 vm_page_t vm_page_free_fromq_fast(void);
 440 void vm_page_dirty(vm_page_t m);
 441 void vm_page_sleep_busy(vm_page_t m, int also_m_busy, const char *msg);
 442 void VM_PAGE_DEBUG_EXT(vm_page_busy_wait)(vm_page_t m,
 443                         int also_m_busy, const char *wmsg VM_PAGE_DEBUG_ARGS);
 444 int VM_PAGE_DEBUG_EXT(vm_page_busy_try)(vm_page_t m,
 445                         int also_m_busy VM_PAGE_DEBUG_ARGS);
 446 u_short vm_get_pg_color(int cpuid, vm_object_t object, vm_pindex_t pindex);
 447
 448 #ifdef VM_PAGE_DEBUG
 449
 450 #define vm_page_lookup_busy_wait(object, pindex, alsob, msg)            \
 451         vm_page_lookup_busy_wait_debug(object, pindex, alsob, msg,      \
 452                                         __func__, __LINE__)
 453
 454 #define vm_page_lookup_busy_try(object, pindex, alsob, errorp)          \
 455         vm_page_lookup_busy_try_debug(object, pindex, alsob, errorp,    \
 456                                         __func__, __LINE__)
 457
 458 #define vm_page_busy_wait(m, alsob, msg)                                \
 459         vm_page_busy_wait_debug(m, alsob, msg, __func__, __LINE__)
 460
 461 #define vm_page_busy_try(m, alsob)                                      \
 462         vm_page_busy_try_debug(m, alsob, __func__, __LINE__)
 463
 464 #endif
 465
 466 #endif                          /* _KERNEL */
 467 #endif                          /* !_VM_VM_PAGE_H_ */