FreeBSD/Linux Kernel Cross Reference
sys/vm/vm_zeroidle.c
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: releng/5.4/sys/vm/vm_zeroidle.c 142302 2005-02-23 15:03:56Z ssouhlal $");
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/resourcevar.h>
45 #include <sys/vmmeter.h>
46 #include <sys/lock.h>
47 #include <sys/mutex.h>
48 #include <sys/sched.h>
49 #include <sys/sysctl.h>
50 #include <sys/kthread.h>
51 #include <sys/unistd.h>
52
53 #include <vm/vm.h>
54 #include <vm/vm_page.h>
55
56 SYSCTL_DECL(_vm_stats_misc);
57
58 static int cnt_prezero;
59 SYSCTL_INT(_vm_stats_misc, OID_AUTO, cnt_prezero, CTLFLAG_RD,
60 &cnt_prezero, 0, "");
61
62 static int idlezero_enable_default = 1;
63 TUNABLE_INT("vm.idlezero_enable", &idlezero_enable_default);
64 /* Defer setting the enable flag until the kthread is running. */
65 static int idlezero_enable = 0;
66 SYSCTL_INT(_vm, OID_AUTO, idlezero_enable, CTLFLAG_RW, &idlezero_enable, 0, "");
67
68 static int idlezero_maxrun = 16;
69 SYSCTL_INT(_vm, OID_AUTO, idlezero_maxrun, CTLFLAG_RW, &idlezero_maxrun, 0, "");
70 TUNABLE_INT("vm.idlezero_maxrun", &idlezero_maxrun);
71
72 /*
73 * Implement the pre-zeroed page mechanism.
74 */
75
76 #define ZIDLE_LO(v) ((v) * 2 / 3)
77 #define ZIDLE_HI(v) ((v) * 4 / 5)
78
79 static boolean_t wakeup_needed = FALSE;
80 static int zero_state;
81
82 static int
83 vm_page_zero_check(void)
84 {
85
86 if (!idlezero_enable)
87 return (0);
88 /*
89 * Attempt to maintain approximately 1/2 of our free pages in a
90 * PG_ZERO'd state. Add some hysteresis to (attempt to) avoid
91 * generally zeroing a page when the system is near steady-state.
92 * Otherwise we might get 'flutter' during disk I/O / IPC or
93 * fast sleeps. We also do not want to be continuously zeroing
94 * pages because doing so may flush our L1 and L2 caches too much.
95 */
96 if (zero_state && vm_page_zero_count >= ZIDLE_LO(cnt.v_free_count))
97 return (0);
98 if (vm_page_zero_count >= ZIDLE_HI(cnt.v_free_count))
99 return (0);
100 return (1);
101 }
102
103 static int
104 vm_page_zero_idle(void)
105 {
106 static int free_rover;
107 vm_page_t m;
108
109 mtx_lock_spin(&vm_page_queue_free_mtx);
110 zero_state = 0;
111 m = vm_pageq_find(PQ_FREE, free_rover, FALSE);
112 if (m != NULL && (m->flags & PG_ZERO) == 0) {
113 vm_pageq_remove_nowakeup(m);
114 mtx_unlock_spin(&vm_page_queue_free_mtx);
115 pmap_zero_page_idle(m);
116 mtx_lock_spin(&vm_page_queue_free_mtx);
117 m->flags |= PG_ZERO;
118 vm_pageq_enqueue(PQ_FREE + m->pc, m);
119 ++vm_page_zero_count;
120 ++cnt_prezero;
121 if (vm_page_zero_count >= ZIDLE_HI(cnt.v_free_count))
122 zero_state = 1;
123 }
124 free_rover = (free_rover + PQ_PRIME2) & PQ_L2_MASK;
125 mtx_unlock_spin(&vm_page_queue_free_mtx);
126 return (1);
127 }
128
129 /* Called by vm_page_free to hint that a new page is available. */
130 void
131 vm_page_zero_idle_wakeup(void)
132 {
133
134 mtx_assert(&vm_page_queue_mtx, MA_OWNED);
135 if (wakeup_needed && vm_page_zero_check()) {
136 wakeup_needed = FALSE;
137 wakeup(&zero_state);
138 }
139 }
140
141 static void
142 vm_pagezero(void __unused *arg)
143 {
144 struct thread *td;
145
146 td = curthread;
147 idlezero_enable = idlezero_enable_default;
148
149 for (;;) {
150 if (vm_page_zero_check()) {
151 vm_page_zero_idle();
152 #ifndef PREEMPTION
153 if (sched_runnable()) {
154 mtx_lock_spin(&sched_lock);
155 mi_switch(SW_VOL, NULL);
156 mtx_unlock_spin(&sched_lock);
157 }
158 #endif
159 } else {
160 vm_page_lock_queues();
161 wakeup_needed = TRUE;
162 msleep(&zero_state, &vm_page_queue_mtx,
163 PDROP | td->td_priority, "pgzero", hz * 300);
164 }
165 }
166 }
167
168 static struct proc *pagezero_proc;
169
170 static void
171 pagezero_start(void __unused *arg)
172 {
173 int error;
174 struct thread *td;
175
176 error = kthread_create(vm_pagezero, NULL, &pagezero_proc, RFSTOPPED, 0,
177 "pagezero");
178 if (error)
179 panic("pagezero_start: error %d\n", error);
180 /*
181 * We're an idle task, don't count us in the load.
182 */
183 PROC_LOCK(pagezero_proc);
184 pagezero_proc->p_flag |= P_NOLOAD;
185 PROC_UNLOCK(pagezero_proc);
186 mtx_lock_spin(&sched_lock);
187 td = FIRST_THREAD_IN_PROC(pagezero_proc);
188 setrunqueue(td, SRQ_BORING);
189 td->td_ksegrp->kg_user_pri = PRI_MAX_IDLE;
190 sched_class(td->td_ksegrp, PRI_IDLE);
191 sched_prio(td, PRI_MAX_IDLE);
192 mtx_unlock_spin(&sched_lock);
193 }
194 SYSINIT(pagezero, SI_SUB_KTHREAD_VM, SI_ORDER_ANY, pagezero_start, NULL)
Cache object: 050eaab2ec0440d3965c757100e2ca7f
|