1/*- 2 * Copyright (c) 1994 John Dyson 3 * Copyright (c) 2001 Matt Dillon 4 * 5 * All Rights Reserved. 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 4. Neither the name of the University nor the names of its contributors 15 * may be used to endorse or promote products derived from this software 16 * without specific prior written permission. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS 19 * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 20 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY 22 * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE 24 * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 25 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 26 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 27 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 28 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 29 * 30 * from: @(#)vm_machdep.c 7.3 (Berkeley) 5/13/91 31 * Utah $Hdr: vm_machdep.c 1.16.1.1 89/06/23$ 32 * from: FreeBSD: .../i386/vm_machdep.c,v 1.165 2001/07/04 23:27:04 dillon 33 */ 34 35#include <sys/cdefs.h> 36__FBSDID("$FreeBSD$"); 37 38#include <opt_sched.h> 39 40#include <sys/param.h> 41#include <sys/systm.h> 42#include <sys/kernel.h> 43#include <sys/proc.h> 44#include <sys/vmmeter.h> 45#include <sys/lock.h> 46#include <sys/mutex.h> 47#include <sys/sched.h> 48#include <sys/sysctl.h> 49#include <sys/kthread.h> 50#include <sys/unistd.h> 51 52#include <vm/vm.h> 53#include <vm/vm_page.h> 54#include <vm/vm_phys.h> 55 56static int idlezero_enable_default = 0; 57TUNABLE_INT("vm.idlezero_enable", &idlezero_enable_default); 58/* Defer setting the enable flag until the kthread is running. */ 59static int idlezero_enable = 0; 60SYSCTL_INT(_vm, OID_AUTO, idlezero_enable, CTLFLAG_RW, &idlezero_enable, 0, 61 "Allow the kernel to use idle cpu cycles to zero-out pages"); 62/* 63 * Implement the pre-zeroed page mechanism. 64 */ 65 66#define ZIDLE_LO(v) ((v) * 2 / 3) 67#define ZIDLE_HI(v) ((v) * 4 / 5) 68 69static boolean_t wakeup_needed = FALSE; 70static int zero_state; 71 72static int 73vm_page_zero_check(void) 74{ 75 76 if (!idlezero_enable) 77 return (0); 78 /* 79 * Attempt to maintain approximately 1/2 of our free pages in a 80 * PG_ZERO'd state. Add some hysteresis to (attempt to) avoid 81 * generally zeroing a page when the system is near steady-state. 82 * Otherwise we might get 'flutter' during disk I/O / IPC or 83 * fast sleeps. We also do not want to be continuously zeroing 84 * pages because doing so may flush our L1 and L2 caches too much. 85 */ 86 if (zero_state && vm_page_zero_count >= ZIDLE_LO(cnt.v_free_count)) 87 return (0); 88 if (vm_page_zero_count >= ZIDLE_HI(cnt.v_free_count)) 89 return (0); 90 return (1); 91} 92 93static void 94vm_page_zero_idle(void) 95{ 96 97 mtx_assert(&vm_page_queue_free_mtx, MA_OWNED); 98 zero_state = 0; 99 if (vm_phys_zero_pages_idle()) { 100 if (vm_page_zero_count >= ZIDLE_HI(cnt.v_free_count)) 101 zero_state = 1; 102 } 103} 104 105/* Called by vm_page_free to hint that a new page is available. */ 106void 107vm_page_zero_idle_wakeup(void) 108{ 109 110 mtx_assert(&vm_page_queue_free_mtx, MA_OWNED); 111 if (wakeup_needed && vm_page_zero_check()) { 112 wakeup_needed = FALSE; 113 wakeup(&zero_state); 114 } 115} 116 117static void 118vm_pagezero(void __unused *arg) 119{ 120 121 idlezero_enable = idlezero_enable_default; 122 123 mtx_lock(&vm_page_queue_free_mtx); 124 for (;;) { 125 if (vm_page_zero_check()) { 126 vm_page_zero_idle(); 127#ifndef PREEMPTION 128 if (sched_runnable()) { 129 thread_lock(curthread); 130 mi_switch(SW_VOL | SWT_IDLE, NULL); 131 thread_unlock(curthread); 132 } 133#endif 134 } else { 135 wakeup_needed = TRUE; 136 msleep(&zero_state, &vm_page_queue_free_mtx, 0, 137 "pgzero", hz * 300); 138 } 139 } 140} 141 142static void 143pagezero_start(void __unused *arg) 144{ 145 int error; 146 struct proc *p; 147 struct thread *td; 148 149 error = kproc_create(vm_pagezero, NULL, &p, RFSTOPPED, 0, "pagezero"); 150 if (error) 151 panic("pagezero_start: error %d\n", error); 152 td = FIRST_THREAD_IN_PROC(p); 153 thread_lock(td); 154 155 /* We're an idle task, don't count us in the load. */ 156 td->td_flags |= TDF_NOLOAD; 157 sched_class(td, PRI_IDLE); 158 sched_prio(td, PRI_MAX_IDLE); 159 sched_add(td, SRQ_BORING); 160 thread_unlock(td); 161} 162SYSINIT(pagezero, SI_SUB_KTHREAD_VM, SI_ORDER_ANY, pagezero_start, NULL); 163