vm_page.h revision 18779
15455Sdg/* 21541Srgrimes * Copyright (c) 1991, 1993 31541Srgrimes * The Regents of the University of California. All rights reserved. 41541Srgrimes * 51541Srgrimes * This code is derived from software contributed to Berkeley by 61541Srgrimes * The Mach Operating System project at Carnegie-Mellon University. 71541Srgrimes * 81541Srgrimes * Redistribution and use in source and binary forms, with or without 91541Srgrimes * modification, are permitted provided that the following conditions 101541Srgrimes * are met: 111541Srgrimes * 1. Redistributions of source code must retain the above copyright 121541Srgrimes * notice, this list of conditions and the following disclaimer. 131541Srgrimes * 2. Redistributions in binary form must reproduce the above copyright 141541Srgrimes * notice, this list of conditions and the following disclaimer in the 151541Srgrimes * documentation and/or other materials provided with the distribution. 161541Srgrimes * 3. All advertising materials mentioning features or use of this software 171541Srgrimes * must display the following acknowledgement: 181541Srgrimes * This product includes software developed by the University of 191541Srgrimes * California, Berkeley and its contributors. 201541Srgrimes * 4. Neither the name of the University nor the names of its contributors 211541Srgrimes * may be used to endorse or promote products derived from this software 221541Srgrimes * without specific prior written permission. 231541Srgrimes * 241541Srgrimes * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 251541Srgrimes * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 261541Srgrimes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 271541Srgrimes * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 281541Srgrimes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 291541Srgrimes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 301541Srgrimes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 311541Srgrimes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 321541Srgrimes * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 331541Srgrimes * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 341541Srgrimes * SUCH DAMAGE. 351541Srgrimes * 361817Sdg * from: @(#)vm_page.h 8.2 (Berkeley) 12/13/93 371541Srgrimes * 381541Srgrimes * 391541Srgrimes * Copyright (c) 1987, 1990 Carnegie-Mellon University. 401541Srgrimes * All rights reserved. 411541Srgrimes * 421541Srgrimes * Authors: Avadis Tevanian, Jr., Michael Wayne Young 435455Sdg * 441541Srgrimes * Permission to use, copy, modify and distribute this software and 451541Srgrimes * its documentation is hereby granted, provided that both the copyright 461541Srgrimes * notice and this permission notice appear in all copies of the 471541Srgrimes * software, derivative works or modified versions, and any portions 481541Srgrimes * thereof, and that both notices appear in supporting documentation. 495455Sdg * 505455Sdg * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 515455Sdg * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND 521541Srgrimes * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 535455Sdg * 541541Srgrimes * Carnegie Mellon requests users of this software to return to 551541Srgrimes * 561541Srgrimes * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 571541Srgrimes * School of Computer Science 581541Srgrimes * Carnegie Mellon University 591541Srgrimes * Pittsburgh PA 15213-3890 601541Srgrimes * 611541Srgrimes * any improvements or extensions that they make and grant Carnegie the 621541Srgrimes * rights to redistribute these changes. 631817Sdg * 6418779Sdyson * $Id: vm_page.h,v 1.32 1996/09/08 20:44:46 dyson Exp $ 651541Srgrimes */ 661541Srgrimes 671541Srgrimes/* 681541Srgrimes * Resident memory system definitions. 691541Srgrimes */ 701541Srgrimes 711541Srgrimes#ifndef _VM_PAGE_ 721541Srgrimes#define _VM_PAGE_ 731541Srgrimes 746816Sdg#include <vm/pmap.h> 751541Srgrimes/* 761541Srgrimes * Management of resident (logical) pages. 771541Srgrimes * 781541Srgrimes * A small structure is kept for each resident 791541Srgrimes * page, indexed by page number. Each structure 801541Srgrimes * is an element of several lists: 811541Srgrimes * 821541Srgrimes * A hash table bucket used to quickly 831541Srgrimes * perform object/offset lookups 841541Srgrimes * 851541Srgrimes * A list of all pages for a given object, 861541Srgrimes * so they can be quickly deactivated at 871541Srgrimes * time of deallocation. 881541Srgrimes * 891541Srgrimes * An ordered list of pages due for pageout. 901541Srgrimes * 911541Srgrimes * In addition, the structure contains the object 921541Srgrimes * and offset to which this page belongs (for pageout), 931541Srgrimes * and sundry status bits. 941541Srgrimes * 951541Srgrimes * Fields in this structure are locked either by the lock on the 961541Srgrimes * object that the page belongs to (O) or by the lock on the page 971541Srgrimes * queues (P). 981541Srgrimes */ 991541Srgrimes 1001541SrgrimesTAILQ_HEAD(pglist, vm_page); 1011541Srgrimes 1021541Srgrimesstruct vm_page { 1035455Sdg TAILQ_ENTRY(vm_page) pageq; /* queue info for FIFO queue or free list (P) */ 1045455Sdg TAILQ_ENTRY(vm_page) hashq; /* hash table links (O) */ 1055455Sdg TAILQ_ENTRY(vm_page) listq; /* pages in same object (O) */ 1061541Srgrimes 1075455Sdg vm_object_t object; /* which object am I in (O,P) */ 10812767Sdyson vm_pindex_t pindex; /* offset into object (O,P) */ 1095455Sdg vm_offset_t phys_addr; /* physical address of page */ 11018169Sdyson u_short queue; /* page queue index */ 11118169Sdyson u_short flags, /* see below */ 11218169Sdyson pc; /* page color */ 1135455Sdg u_short wire_count; /* wired down maps refs (P) */ 1145455Sdg short hold_count; /* page hold count */ 11513490Sdyson u_char act_count; /* page usage count */ 11613490Sdyson u_char busy; /* page busy count */ 11713490Sdyson /* NOTE that these must support one bit per DEV_BSIZE in a page!!! */ 11813490Sdyson /* so, on normal X86 kernels, they must be at least 8 bits wide */ 11913490Sdyson u_char valid; /* map of valid DEV_BSIZE chunks */ 12013490Sdyson u_char dirty; /* map of dirty DEV_BSIZE chunks */ 1211541Srgrimes}; 1221541Srgrimes 12318169Sdyson/* 12418169Sdyson * Page coloring parameters 12518169Sdyson */ 12618169Sdyson/* Each of PQ_FREE, PQ_ZERO and PQ_CACHE have PQ_HASH_SIZE entries */ 12718169Sdyson 12818169Sdyson/* Define one of the following */ 12918169Sdyson#if defined(PQ_LARGECACHE) 13018169Sdyson#define PQ_PRIME1 31 /* Prime number somewhat less than PQ_HASH_SIZE */ 13118169Sdyson#define PQ_PRIME2 23 /* Prime number somewhat less than PQ_HASH_SIZE */ 13218169Sdyson#define PQ_PRIME3 17 /* Prime number somewhat less than PQ_HASH_SIZE */ 13318169Sdyson#define PQ_L2_SIZE 128 /* A number of colors opt for 512K cache */ 13418169Sdyson#define PQ_L1_SIZE 2 /* Two page L1 cache */ 13518169Sdyson#endif 13618169Sdyson 13718169Sdyson 13818169Sdyson/* 13918169Sdyson * Use 'options PQ_NOOPT' to disable page coloring 14018169Sdyson */ 14118169Sdyson#if defined(PQ_NOOPT) 14218169Sdyson#define PQ_PRIME1 1 14318169Sdyson#define PQ_PRIME2 1 14418169Sdyson#define PQ_PRIME3 1 14518169Sdyson#define PQ_L2_SIZE 1 14618169Sdyson#define PQ_L1_SIZE 1 14718169Sdyson#endif 14818169Sdyson 14918779Sdyson#if defined(PQ_NORMALCACHE) 15018169Sdyson#define PQ_PRIME1 5 /* Prime number somewhat less than PQ_HASH_SIZE */ 15118169Sdyson#define PQ_PRIME2 3 /* Prime number somewhat less than PQ_HASH_SIZE */ 15218169Sdyson#define PQ_PRIME3 11 /* Prime number somewhat less than PQ_HASH_SIZE */ 15318169Sdyson#define PQ_L2_SIZE 16 /* A reasonable number of colors (opt for 64K cache) */ 15418169Sdyson#define PQ_L1_SIZE 2 /* Two page L1 cache */ 15518169Sdyson#endif 15618169Sdyson 15718779Sdyson#if defined(PQ_MEDIUMCACHE) || !defined(PQ_L2_SIZE) 15818779Sdyson#define PQ_PRIME1 13 /* Prime number somewhat less than PQ_HASH_SIZE */ 15918779Sdyson#define PQ_PRIME2 7 /* Prime number somewhat less than PQ_HASH_SIZE */ 16018779Sdyson#define PQ_PRIME3 5 /* Prime number somewhat less than PQ_HASH_SIZE */ 16118779Sdyson#define PQ_L2_SIZE 64 /* A number of colors opt for 256K cache */ 16218779Sdyson#define PQ_L1_SIZE 2 /* Two page L1 cache */ 16318779Sdyson#endif 16418779Sdyson 16518169Sdyson#define PQ_L2_MASK (PQ_L2_SIZE - 1) 16618169Sdyson 16713490Sdyson#define PQ_NONE 0 16813490Sdyson#define PQ_FREE 1 16918169Sdyson#define PQ_ZERO (1 + PQ_L2_SIZE) 17018169Sdyson#define PQ_INACTIVE (1 + 2*PQ_L2_SIZE) 17118169Sdyson#define PQ_ACTIVE (2 + 2*PQ_L2_SIZE) 17218169Sdyson#define PQ_CACHE (3 + 2*PQ_L2_SIZE) 17318169Sdyson#define PQ_COUNT (3 + 3*PQ_L2_SIZE) 17413490Sdyson 17518169Sdysonextern struct vpgqueues { 17618169Sdyson struct pglist *pl; 17718169Sdyson int *cnt; 17818169Sdyson int *lcnt; 17918169Sdyson} vm_page_queues[PQ_COUNT]; 18018169Sdyson 1811541Srgrimes/* 1821541Srgrimes * These are the flags defined for vm_page. 1831541Srgrimes * 1841541Srgrimes * Note: PG_FILLED and PG_DIRTY are added for the filesystems. 1851541Srgrimes */ 18613490Sdyson#define PG_BUSY 0x01 /* page is in transit (O) */ 18713490Sdyson#define PG_WANTED 0x02 /* someone is waiting for page (O) */ 18813490Sdyson#define PG_TABLED 0x04 /* page is in VP table (O) */ 18913490Sdyson#define PG_FICTITIOUS 0x08 /* physical page doesn't exist (O) */ 19013490Sdyson#define PG_WRITEABLE 0x10 /* page is mapped writeable */ 19113490Sdyson#define PG_MAPPED 0x20 /* page is mapped */ 19213490Sdyson#define PG_ZERO 0x40 /* page is zeroed */ 19313490Sdyson#define PG_REFERENCED 0x80 /* page has been referenced */ 19413490Sdyson#define PG_CLEANCHK 0x100 /* page has been checked for cleaning */ 1951541Srgrimes 1969507Sdg/* 1979507Sdg * Misc constants. 1989507Sdg */ 1991541Srgrimes 2009507Sdg#define ACT_DECLINE 1 2019507Sdg#define ACT_ADVANCE 3 20216750Sdyson#define ACT_INIT 5 20318169Sdyson#define ACT_MAX 64 2049507Sdg#define PFCLUSTER_BEHIND 3 2059507Sdg#define PFCLUSTER_AHEAD 3 2069507Sdg 2071541Srgrimes#ifdef KERNEL 2081541Srgrimes/* 2099507Sdg * Each pageable resident page falls into one of four lists: 2101541Srgrimes * 2115455Sdg * free 2121541Srgrimes * Available for allocation now. 2139507Sdg * 2149507Sdg * The following are all LRU sorted: 2159507Sdg * 2169507Sdg * cache 2179507Sdg * Almost available for allocation. Still in an 2189507Sdg * object, but clean and immediately freeable at 2199507Sdg * non-interrupt times. 2209507Sdg * 2211541Srgrimes * inactive 22213765Smpp * Low activity, candidates for reclamation. 2231541Srgrimes * This is the list of pages that should be 2241541Srgrimes * paged out next. 2259507Sdg * 2261541Srgrimes * active 2279507Sdg * Pages that are "active" i.e. they have been 2289507Sdg * recently referenced. 22910544Sdyson * 23010544Sdyson * zero 23110544Sdyson * Pages that are really free and have been pre-zeroed 23210544Sdyson * 2331541Srgrimes */ 2341541Srgrimes 23518169Sdysonextern struct pglist vm_page_queue_free[PQ_L2_SIZE];/* memory free queue */ 23618169Sdysonextern struct pglist vm_page_queue_zero[PQ_L2_SIZE];/* zeroed memory free queue */ 2375455Sdgextern struct pglist vm_page_queue_active; /* active memory queue */ 2385455Sdgextern struct pglist vm_page_queue_inactive; /* inactive memory queue */ 23918169Sdysonextern struct pglist vm_page_queue_cache[PQ_L2_SIZE];/* cache memory queue */ 2401541Srgrimes 24112767Sdysonextern int vm_page_zero_count; 24212767Sdyson 2435455Sdgextern vm_page_t vm_page_array; /* First resident page in table */ 2445455Sdgextern long first_page; /* first physical page number */ 2451541Srgrimes 2465455Sdg /* ... represented in vm_page_array */ 2475455Sdgextern long last_page; /* last physical page number */ 2485455Sdg 2495455Sdg /* ... represented in vm_page_array */ 2505455Sdg /* [INCLUSIVE] */ 2515455Sdgextern vm_offset_t first_phys_addr; /* physical address for first_page */ 2525455Sdgextern vm_offset_t last_phys_addr; /* physical address for last_page */ 2535455Sdg 2541541Srgrimes#define VM_PAGE_TO_PHYS(entry) ((entry)->phys_addr) 2551541Srgrimes 2561541Srgrimes#define IS_VM_PHYSADDR(pa) \ 2571541Srgrimes ((pa) >= first_phys_addr && (pa) <= last_phys_addr) 2581541Srgrimes 2591541Srgrimes#define PHYS_TO_VM_PAGE(pa) \ 2601541Srgrimes (&vm_page_array[atop(pa) - first_page ]) 2611541Srgrimes 2621541Srgrimes/* 2631541Srgrimes * Functions implemented as macros 2641541Srgrimes */ 2651541Srgrimes 2661541Srgrimes#define PAGE_ASSERT_WAIT(m, interruptible) { \ 2671541Srgrimes (m)->flags |= PG_WANTED; \ 2681541Srgrimes assert_wait((int) (m), (interruptible)); \ 2691541Srgrimes } 2701541Srgrimes 2711541Srgrimes#define PAGE_WAKEUP(m) { \ 2721541Srgrimes (m)->flags &= ~PG_BUSY; \ 2731541Srgrimes if ((m)->flags & PG_WANTED) { \ 2741541Srgrimes (m)->flags &= ~PG_WANTED; \ 2751549Srgrimes wakeup((caddr_t) (m)); \ 2761541Srgrimes } \ 2771541Srgrimes } 2781541Srgrimes 2795455Sdg#if PAGE_SIZE == 4096 2805455Sdg#define VM_PAGE_BITS_ALL 0xff 2815455Sdg#endif 2821541Srgrimes 2835455Sdg#if PAGE_SIZE == 8192 2845455Sdg#define VM_PAGE_BITS_ALL 0xffff 2855455Sdg#endif 2861549Srgrimes 2875841Sdg#define VM_ALLOC_NORMAL 0 2885841Sdg#define VM_ALLOC_INTERRUPT 1 2895841Sdg#define VM_ALLOC_SYSTEM 2 29013490Sdyson#define VM_ALLOC_ZERO 3 2915455Sdg 2925455Sdgvoid vm_page_activate __P((vm_page_t)); 29312767Sdysonvm_page_t vm_page_alloc __P((vm_object_t, vm_pindex_t, int)); 2946357Sphkvoid vm_page_cache __P((register vm_page_t)); 29515811Sdysonstatic __inline void vm_page_copy __P((vm_page_t, vm_page_t)); 2965455Sdgvoid vm_page_deactivate __P((vm_page_t)); 2975455Sdgvoid vm_page_free __P((vm_page_t)); 29816122Sdysonvoid vm_page_free_zero __P((vm_page_t)); 29912767Sdysonvoid vm_page_insert __P((vm_page_t, vm_object_t, vm_pindex_t)); 30012767Sdysonvm_page_t vm_page_lookup __P((vm_object_t, vm_pindex_t)); 3015455Sdgvoid vm_page_remove __P((vm_page_t)); 30212767Sdysonvoid vm_page_rename __P((vm_page_t, vm_object_t, vm_pindex_t)); 3035455Sdgvm_offset_t vm_page_startup __P((vm_offset_t, vm_offset_t, vm_offset_t)); 3045455Sdgvoid vm_page_unwire __P((vm_page_t)); 3055455Sdgvoid vm_page_wire __P((vm_page_t)); 30617334Sdysonvoid vm_page_unqueue __P((vm_page_t)); 30717334Sdysonvoid vm_page_unqueue_nowakeup __P((vm_page_t)); 30810544Sdysonvoid vm_page_set_validclean __P((vm_page_t, int, int)); 3095455Sdgvoid vm_page_set_invalid __P((vm_page_t, int, int)); 31015811Sdysonstatic __inline boolean_t vm_page_zero_fill __P((vm_page_t)); 3115455Sdgint vm_page_is_valid __P((vm_page_t, int, int)); 3125455Sdgvoid vm_page_test_dirty __P((vm_page_t)); 3136357Sphkint vm_page_bits __P((int, int)); 31418169Sdysonvm_page_t vm_page_list_find __P((int, int)); 31518169Sdysonint vm_page_queue_index __P((vm_offset_t, int)); 31618169Sdysonvm_page_t vm_page_select __P((vm_object_t, vm_pindex_t, int)); 3175455Sdg 3181549Srgrimes/* 3191549Srgrimes * Keep page from being freed by the page daemon 3201549Srgrimes * much of the same effect as wiring, except much lower 3211549Srgrimes * overhead and should be used only for *very* temporary 3221549Srgrimes * holding ("wiring"). 3231549Srgrimes */ 3248010Sbdestatic __inline void 3254461Sbdevm_page_hold(vm_page_t mem) 3261549Srgrimes{ 3271549Srgrimes mem->hold_count++; 3281549Srgrimes} 3291549Srgrimes 3303745Swollman#ifdef DIAGNOSTIC 3313745Swollman#include <sys/systm.h> /* make GCC shut up */ 3323745Swollman#endif 3333745Swollman 3348010Sbdestatic __inline void 3354461Sbdevm_page_unhold(vm_page_t mem) 3361549Srgrimes{ 3373660Sdg#ifdef DIAGNOSTIC 3385455Sdg if (--mem->hold_count < 0) 3391549Srgrimes panic("vm_page_unhold: hold count < 0!!!"); 3403660Sdg#else 3413660Sdg --mem->hold_count; 3423660Sdg#endif 3431549Srgrimes} 3441549Srgrimes 3458010Sbdestatic __inline void 3466816Sdgvm_page_protect(vm_page_t mem, int prot) 3476816Sdg{ 3486816Sdg if (prot == VM_PROT_NONE) { 3496816Sdg if (mem->flags & (PG_WRITEABLE|PG_MAPPED)) { 35017334Sdyson pmap_page_protect(VM_PAGE_TO_PHYS(mem), prot); 3516816Sdg mem->flags &= ~(PG_WRITEABLE|PG_MAPPED); 3526816Sdg } 3536816Sdg } else if ((prot == VM_PROT_READ) && (mem->flags & PG_WRITEABLE)) { 35417334Sdyson pmap_page_protect(VM_PAGE_TO_PHYS(mem), prot); 3556816Sdg mem->flags &= ~PG_WRITEABLE; 3566816Sdg } 3576816Sdg} 3586816Sdg 35915811Sdyson/* 36015811Sdyson * vm_page_zero_fill: 36115811Sdyson * 36215811Sdyson * Zero-fill the specified page. 36315811Sdyson * Written as a standard pagein routine, to 36415811Sdyson * be used by the zero-fill object. 36515811Sdyson */ 36615811Sdysonstatic __inline boolean_t 36715811Sdysonvm_page_zero_fill(m) 36815811Sdyson vm_page_t m; 36915811Sdyson{ 37015811Sdyson pmap_zero_page(VM_PAGE_TO_PHYS(m)); 37115811Sdyson return (TRUE); 37215811Sdyson} 3736816Sdg 37415811Sdyson/* 37515811Sdyson * vm_page_copy: 37615811Sdyson * 37715811Sdyson * Copy one page to another 37815811Sdyson */ 37915811Sdysonstatic __inline void 38015811Sdysonvm_page_copy(src_m, dest_m) 38115811Sdyson vm_page_t src_m; 38215811Sdyson vm_page_t dest_m; 38315811Sdyson{ 38415811Sdyson pmap_copy_page(VM_PAGE_TO_PHYS(src_m), VM_PAGE_TO_PHYS(dest_m)); 38515811Sdyson dest_m->valid = VM_PAGE_BITS_ALL; 38615811Sdyson} 38715811Sdyson 3885455Sdg#endif /* KERNEL */ 3895455Sdg#endif /* !_VM_PAGE_ */ 390