sys/vm/vm_page.h

   1 /*
   2  * Copyright (c) 1991, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  * Copyright (c) 2003-2017 The DragonFly Project.  All rights reserved.
   5  *
   6  * This code is derived from software contributed to Berkeley by
   7  * The Mach Operating System project at Carnegie-Mellon University.
   8  *
   9  * This code is derived from software contributed to The DragonFly Project
  10  * by Matthew Dillon <dillon@backplane.com>
  11  *
  12  * Redistribution and use in source and binary forms, with or without
  13  * modification, are permitted provided that the following conditions
  14  * are met:
  15  * 1. Redistributions of source code must retain the above copyright
  16  *    notice, this list of conditions and the following disclaimer.
  17  * 2. Redistributions in binary form must reproduce the above copyright
  18  *    notice, this list of conditions and the following disclaimer in the
  19  *    documentation and/or other materials provided with the distribution.
  20  * 3. Neither the name of the University nor the names of its contributors
  21  *    may be used to endorse or promote products derived from this software
  22  *    without specific prior written permission.
  23  *
  24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  34  * SUCH DAMAGE.
  35  *
  36  *      from: @(#)vm_page.h     8.2 (Berkeley) 12/13/93
  37  *
  38  *
  39  * Copyright (c) 1987, 1990 Carnegie-Mellon University.
  40  * All rights reserved.
  41  *
  42  * Authors: Avadis Tevanian, Jr., Michael Wayne Young
  43  *
  44  * Permission to use, copy, modify and distribute this software and
  45  * its documentation is hereby granted, provided that both the copyright
  46  * notice and this permission notice appear in all copies of the
  47  * software, derivative works or modified versions, and any portions
  48  * thereof, and that both notices appear in supporting documentation.
  49  *
  50  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
  51  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
  52  * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
  53  *
  54  * Carnegie Mellon requests users of this software to return to
  55  *
  56  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
  57  *  School of Computer Science
  58  *  Carnegie Mellon University
  59  *  Pittsburgh PA 15213-3890
  60  *
  61  * any improvements or extensions that they make and grant Carnegie the
  62  * rights to redistribute these changes.
  63  */
  64
  65 /*
  66  * Resident memory system definitions.
  67  */
  68
  69 #ifndef _VM_VM_PAGE_H_
  70 #define _VM_VM_PAGE_H_
  71
  72 #ifndef _SYS_TYPES_H_
  73 #include <sys/types.h>
  74 #endif
  75 #ifndef _SYS_TREE_H_
  76 #include <sys/tree.h>
  77 #endif
  78 #ifndef _MACHINE_PMAP_H_
  79 #include <machine/pmap.h>
  80 #endif
  81 #ifndef _VM_PMAP_H_
  82 #include <vm/pmap.h>
  83 #endif
  84 #include <machine/atomic.h>
  85
  86 #ifdef _KERNEL
  87
  88 #ifndef _SYS_SYSTM_H_
  89 #include <sys/systm.h>
  90 #endif
  91 #ifndef _SYS_SPINLOCK_H_
  92 #include <sys/spinlock.h>
  93 #endif
  94 #ifndef _SYS_THREAD2_H_
  95 #include <sys/thread2.h>
  96 #endif
  97
  98 #ifdef __x86_64__
  99 #include <machine/vmparam.h>
 100 #endif
 101
 102 #endif
 103
 104 /*
 105  * vm_page structure
 106  */
 107 TAILQ_HEAD(pglist, vm_page);
 108
 109 struct vm_object;
 110
 111 int rb_vm_page_compare(struct vm_page *, struct vm_page *);
 112
 113 struct vm_page_rb_tree;
 114 RB_PROTOTYPE2(vm_page_rb_tree, vm_page, rb_entry,
 115               rb_vm_page_compare, vm_pindex_t);
 116
 117 struct vm_page {
 118         TAILQ_ENTRY(vm_page) pageq;     /* vm_page_queues[] list (P)    */
 119         RB_ENTRY(vm_page) rb_entry;     /* Red-Black tree based at object */
 120         struct spinlock spin;
 121         struct vm_object *object;       /* which object am I in (O,P)*/
 122         vm_pindex_t pindex;             /* offset into object (O,P) */
 123         vm_paddr_t phys_addr;           /* physical address of page */
 124         struct md_page md;              /* machine dependant stuff */
 125         uint16_t queue;                 /* page queue index */
 126         uint16_t pc;                    /* page color */
 127         uint8_t act_count;              /* page usage count */
 128         uint8_t pat_mode;               /* hardware page attribute */
 129         uint8_t valid;                  /* map of valid DEV_BSIZE chunks */
 130         uint8_t dirty;                  /* map of dirty DEV_BSIZE chunks */
 131         uint32_t flags;                 /* see below */
 132         uint32_t wire_count;            /* wired down maps refs (P) */
 133         uint32_t busy_count;            /* soft-busy and hard-busy */
 134         int     hold_count;             /* page hold count */
 135         int     ku_pagecnt;             /* kmalloc helper */
 136 #ifdef VM_PAGE_DEBUG
 137         const char *busy_func;
 138         int     busy_line;
 139 #endif
 140 };
 141
 142 #define PBUSY_LOCKED            0x80000000U
 143 #define PBUSY_WANTED            0x40000000U
 144 #define PBUSY_SWAPINPROG        0x20000000U
 145 #define PBUSY_MASK              0x1FFFFFFFU
 146
 147 #ifndef __VM_PAGE_T_DEFINED__
 148 #define __VM_PAGE_T_DEFINED__
 149 typedef struct vm_page *vm_page_t;
 150 #endif
 151
 152 /*
 153  * Page coloring parameters.  We use generous parameters designed to
 154  * statistically spread pages over available cpu cache space.  This has
 155  * become less important over time as cache associativity is higher
 156  * in modern times but we still use the core algorithm to help reduce
 157  * lock contention between cpus.
 158  *
 159  * Page coloring cannot be disabled.
 160  *
 161  * In today's world of many-core systems, we must be able to provide enough VM
 162  * page queues for each logical cpu thread to cover the L1/L2/L3 cache set
 163  * associativity.  If we don't, the cpu caches will not be properly utilized.
 164  * Using 2048 allows 8-way set-assoc with 256 logical cpus.
 165  */
 166 #define PQ_PRIME1 31    /* Prime number somewhat less than PQ_HASH_SIZE */
 167 #define PQ_PRIME2 23    /* Prime number somewhat less than PQ_HASH_SIZE */
 168 #define PQ_L2_SIZE 2048 /* Must be enough for maximal ncpus x hw set-assoc */
 169 #define PQ_L2_MASK      (PQ_L2_SIZE - 1)
 170
 171 #define PQ_NONE         0
 172 #define PQ_FREE         (1 + 0*PQ_L2_SIZE)
 173 #define PQ_INACTIVE     (1 + 1*PQ_L2_SIZE)
 174 #define PQ_ACTIVE       (1 + 2*PQ_L2_SIZE)
 175 #define PQ_CACHE        (1 + 3*PQ_L2_SIZE)
 176 #define PQ_HOLD         (1 + 4*PQ_L2_SIZE)
 177 #define PQ_COUNT        (1 + 5*PQ_L2_SIZE)
 178
 179 /*
 180  * Scan support
 181  */
 182 struct vm_map;
 183
 184 struct rb_vm_page_scan_info {
 185         vm_pindex_t     start_pindex;
 186         vm_pindex_t     end_pindex;
 187         int             limit;
 188         int             desired;
 189         int             error;
 190         int             pagerflags;
 191         int             count;
 192         int             unused01;
 193         vm_offset_t     addr;
 194         vm_pindex_t     backing_offset_index;
 195         struct vm_object *object;
 196         struct vm_object *backing_object;
 197         struct vm_page  *mpte;
 198         struct pmap     *pmap;
 199         struct vm_map   *map;
 200 };
 201
 202 int rb_vm_page_scancmp(struct vm_page *, void *);
 203
 204 struct vpgqueues {
 205         struct spinlock spin;
 206         struct pglist pl;
 207         int     cnt_offset;     /* offset into vmstats structure (int) */
 208         int     lcnt;
 209         int     flipflop;       /* probably not the best place */
 210         int     unused00;
 211         int     unused01;
 212         char    unused[64 - sizeof(struct pglist) -
 213                         sizeof(int *) - sizeof(int) * 4];
 214 };
 215
 216 extern struct vpgqueues vm_page_queues[PQ_COUNT];
 217
 218 /*
 219  * These are the flags defined for vm_page.
 220  *
 221  *  PG_UNMANAGED (used by OBJT_PHYS) indicates that the page is
 222  *  not under PV management but otherwise should be treated as a
 223  *  normal page.  Pages not under PV management cannot be paged out
 224  *  via the object/vm_page_t because there is no knowledge of their
 225  *  pte mappings, nor can they be removed from their objects via
 226  *  the object, and such pages are also not on any PQ queue.  The
 227  *  PG_MAPPED and PG_WRITEABLE flags are not applicable.
 228  *
 229  *  PG_MAPPED only applies to managed pages, indicating whether the page
 230  *  is mapped onto one or more pmaps.  A page might still be mapped to
 231  *  special pmaps in an unmanaged fashion, for example when mapped into a
 232  *  buffer cache buffer, without setting PG_MAPPED.
 233  *
 234  *  PG_WRITEABLE indicates that there may be a writeable managed pmap entry
 235  *  somewhere, and that the page can be dirtied by hardware at any time
 236  *  and may have to be tested for that.  The modified bit in unmanaged
 237  *  mappings or in the special clean map is not tested.
 238  *
 239  *  PG_SWAPPED indicates that the page is backed by a swap block.  Any
 240  *  VM object type other than OBJT_DEFAULT can have swap-backed pages now.
 241  */
 242 #define PG_UNUSED0001   0x00000001
 243 #define PG_UNUSED0002   0x00000002
 244 #define PG_WINATCFLS    0x00000004      /* flush dirty page on inactive q */
 245 #define PG_FICTITIOUS   0x00000008      /* physical page doesn't exist (O) */
 246 #define PG_WRITEABLE    0x00000010      /* page is writeable */
 247 #define PG_MAPPED       0x00000020      /* page is mapped (managed) */
 248 #define PG_UNUSED0040   0x00000040
 249 #define PG_REFERENCED   0x00000080      /* page has been referenced */
 250 #define PG_CLEANCHK     0x00000100      /* page will be checked for cleaning */
 251 #define PG_UNUSED0200   0x00000200
 252 #define PG_NOSYNC       0x00000400      /* do not collect for syncer */
 253 #define PG_UNMANAGED    0x00000800      /* No PV management for page */
 254 #define PG_MARKER       0x00001000      /* special queue marker page */
 255 #define PG_RAM          0x00002000      /* read ahead mark */
 256 #define PG_SWAPPED      0x00004000      /* backed by swap */
 257 #define PG_NOTMETA      0x00008000      /* do not back with swap */
 258 #define PG_UNUSED10000  0x00010000
 259 #define PG_UNUSED20000  0x00020000
 260 #define PG_NEED_COMMIT  0x00040000      /* clean page requires commit */
 261
 262 #define PG_KEEP_NEWPAGE_MASK    (0)
 263
 264 /*
 265  * Misc constants.
 266  */
 267
 268 #define ACT_DECLINE             1
 269 #define ACT_ADVANCE             3
 270 #define ACT_INIT                5
 271 #define ACT_MAX                 64
 272
 273 #ifdef VM_PAGE_DEBUG
 274 #define VM_PAGE_DEBUG_EXT(name) name ## _debug
 275 #define VM_PAGE_DEBUG_ARGS      , const char *func, int lineno
 276 #else
 277 #define VM_PAGE_DEBUG_EXT(name) name
 278 #define VM_PAGE_DEBUG_ARGS
 279 #endif
 280
 281 #ifdef _KERNEL
 282 /*
 283  * Each pageable resident page falls into one of four lists:
 284  *
 285  *      free
 286  *              Available for allocation now.
 287  *
 288  * The following are all LRU sorted:
 289  *
 290  *      cache
 291  *              Almost available for allocation. Still in an
 292  *              object, but clean and immediately freeable at
 293  *              non-interrupt times.
 294  *
 295  *      inactive
 296  *              Low activity, candidates for reclamation.
 297  *              This is the list of pages that should be
 298  *              paged out next.
 299  *
 300  *      active
 301  *              Pages that are "active" i.e. they have been
 302  *              recently referenced.
 303  *
 304  *      zero
 305  *              Pages that are really free and have been pre-zeroed
 306  *
 307  */
 308
 309 extern struct vm_page *vm_page_array;   /* First resident page in table */
 310 extern vm_pindex_t vm_page_array_size;  /* number of vm_page_t's */
 311 extern vm_pindex_t first_page;          /* first physical page number */
 312
 313 #define VM_PAGE_TO_PHYS(entry)  \
 314                 ((entry)->phys_addr)
 315
 316 #define PHYS_TO_VM_PAGE(pa)     \
 317                 (&vm_page_array[atop(pa) - first_page])
 318
 319
 320 #if PAGE_SIZE == 4096
 321 #define VM_PAGE_BITS_ALL 0xff
 322 #endif
 323
 324 /*
 325  * Note: the code will always use nominally free pages from the free list
 326  * before trying other flag-specified sources.
 327  *
 328  * At least one of VM_ALLOC_NORMAL|VM_ALLOC_SYSTEM|VM_ALLOC_INTERRUPT
 329  * must be specified.  VM_ALLOC_RETRY may only be specified if VM_ALLOC_NORMAL
 330  * is also specified.
 331  */
 332 #define VM_ALLOC_NORMAL         0x0001  /* ok to use cache pages */
 333 #define VM_ALLOC_SYSTEM         0x0002  /* ok to exhaust most of free list */
 334 #define VM_ALLOC_INTERRUPT      0x0004  /* ok to exhaust entire free list */
 335 #define VM_ALLOC_ZERO           0x0008  /* req pre-zero'd memory if avail */
 336 #define VM_ALLOC_QUICK          0x0010  /* like NORMAL but do not use cache */
 337 #define VM_ALLOC_FORCE_ZERO     0x0020  /* zero page even if already valid */
 338 #define VM_ALLOC_NULL_OK        0x0040  /* ok to return NULL on collision */
 339 #define VM_ALLOC_RETRY          0x0080  /* indefinite block (vm_page_grab()) */
 340 #define VM_ALLOC_USE_GD         0x0100  /* use per-gd cache */
 341 #define VM_ALLOC_CPU_SPEC       0x0200
 342
 343 #define VM_ALLOC_CPU_SHIFT      16
 344 #define VM_ALLOC_CPU(n)         (((n) << VM_ALLOC_CPU_SHIFT) | \
 345                                  VM_ALLOC_CPU_SPEC)
 346 #define VM_ALLOC_GETCPU(flags)  ((flags) >> VM_ALLOC_CPU_SHIFT)
 347
 348 void vm_page_queue_spin_lock(vm_page_t);
 349 void vm_page_queues_spin_lock(u_short);
 350 void vm_page_and_queue_spin_lock(vm_page_t);
 351
 352 void vm_page_queue_spin_unlock(vm_page_t);
 353 void vm_page_queues_spin_unlock(u_short);
 354 void vm_page_and_queue_spin_unlock(vm_page_t m);
 355
 356 void vm_page_init(vm_page_t m);
 357 void vm_page_io_finish(vm_page_t m);
 358 void vm_page_io_start(vm_page_t m);
 359 void vm_page_need_commit(vm_page_t m);
 360 void vm_page_clear_commit(vm_page_t m);
 361 void vm_page_wakeup(vm_page_t m);
 362 void vm_page_hold(vm_page_t);
 363 void vm_page_unhold(vm_page_t);
 364 void vm_page_activate (vm_page_t);
 365
 366 vm_size_t vm_contig_avail_pages(void);
 367 vm_page_t vm_page_alloc (struct vm_object *, vm_pindex_t, int);
 368 vm_page_t vm_page_alloc_contig(vm_paddr_t low, vm_paddr_t high,
 369                      unsigned long alignment, unsigned long boundary,
 370                      unsigned long size, vm_memattr_t memattr);
 371
 372 vm_page_t vm_page_grab (struct vm_object *, vm_pindex_t, int);
 373 void vm_page_cache (vm_page_t);
 374 int vm_page_try_to_cache (vm_page_t);
 375 int vm_page_try_to_free (vm_page_t);
 376 void vm_page_dontneed (vm_page_t);
 377 void vm_page_deactivate (vm_page_t);
 378 void vm_page_deactivate_locked (vm_page_t);
 379 void vm_page_initfake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr);
 380 int vm_page_insert (vm_page_t, struct vm_object *, vm_pindex_t);
 381 vm_page_t vm_page_lookup (struct vm_object *, vm_pindex_t);
 382 vm_page_t vm_page_lookup_sbusy_try(struct vm_object *object,
 383                 vm_pindex_t pindex, int pgoff, int pgbytes);
 384 vm_page_t VM_PAGE_DEBUG_EXT(vm_page_lookup_busy_wait)(
 385                 struct vm_object *, vm_pindex_t, int, const char *
 386                 VM_PAGE_DEBUG_ARGS);
 387 vm_page_t VM_PAGE_DEBUG_EXT(vm_page_lookup_busy_try)(
 388                 struct vm_object *, vm_pindex_t, int, int *
 389                 VM_PAGE_DEBUG_ARGS);
 390 void vm_page_remove (vm_page_t);
 391 void vm_page_rename (vm_page_t, struct vm_object *, vm_pindex_t);
 392 void vm_page_startup (void);
 393 void vm_numa_organize(vm_paddr_t ran_beg, vm_paddr_t bytes, int physid);
 394 void vm_numa_organize_finalize(void);
 395 void vm_page_unmanage (vm_page_t);
 396 void vm_page_unwire (vm_page_t, int);
 397 void vm_page_wire (vm_page_t);
 398 void vm_page_unqueue (vm_page_t);
 399 void vm_page_unqueue_nowakeup (vm_page_t);
 400 vm_page_t vm_page_next (vm_page_t);
 401 void vm_page_set_validclean (vm_page_t, int, int);
 402 void vm_page_set_validdirty (vm_page_t, int, int);
 403 void vm_page_set_valid (vm_page_t, int, int);
 404 void vm_page_set_dirty (vm_page_t, int, int);
 405 void vm_page_clear_dirty (vm_page_t, int, int);
 406 void vm_page_set_invalid (vm_page_t, int, int);
 407 int vm_page_is_valid (vm_page_t, int, int);
 408 void vm_page_test_dirty (vm_page_t);
 409 int vm_page_bits (int, int);
 410 vm_page_t vm_page_list_find(int basequeue, int index);
 411 void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid);
 412 void vm_page_free_toq(vm_page_t m);
 413 void vm_page_free_contig(vm_page_t m, unsigned long size);
 414 vm_page_t vm_page_free_fromq_fast(void);
 415 void vm_page_dirty(vm_page_t m);
 416 void vm_page_sleep_busy(vm_page_t m, int also_m_busy, const char *msg);
 417 int vm_page_sbusy_try(vm_page_t m);
 418 void VM_PAGE_DEBUG_EXT(vm_page_busy_wait)(vm_page_t m,
 419                         int also_m_busy, const char *wmsg VM_PAGE_DEBUG_ARGS);
 420 int VM_PAGE_DEBUG_EXT(vm_page_busy_try)(vm_page_t m,
 421                         int also_m_busy VM_PAGE_DEBUG_ARGS);
 422 u_short vm_get_pg_color(int cpuid, vm_object_t object, vm_pindex_t pindex);
 423
 424 #ifdef VM_PAGE_DEBUG
 425
 426 #define vm_page_lookup_busy_wait(object, pindex, alsob, msg)            \
 427         vm_page_lookup_busy_wait_debug(object, pindex, alsob, msg,      \
 428                                         __func__, __LINE__)
 429
 430 #define vm_page_lookup_busy_try(object, pindex, alsob, errorp)          \
 431         vm_page_lookup_busy_try_debug(object, pindex, alsob, errorp,    \
 432                                         __func__, __LINE__)
 433
 434 #define vm_page_busy_wait(m, alsob, msg)                                \
 435         vm_page_busy_wait_debug(m, alsob, msg, __func__, __LINE__)
 436
 437 #define vm_page_busy_try(m, alsob)                                      \
 438         vm_page_busy_try_debug(m, alsob, __func__, __LINE__)
 439
 440 #endif
 441
 442 #endif                          /* _KERNEL */
 443 #endif                          /* !_VM_VM_PAGE_H_ */