2 * Copyright (c) 2010 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * Implement the swapcache daemon. When enabled swap is assumed to be
37 * configured on a fast storage device such as a SSD. Swap is assigned
38 * to clean vnode-backed pages in the inactive queue, clustered by object
39 * if possible, and written out. The swap assignment sticks around even
40 * after the underlying pages have been recycled.
42 * The daemon manages write bandwidth based on sysctl settings to control
45 * The vnode strategy code will check for the swap assignments and divert
46 * reads to the swap device.
48 * This operates on both regular files and the block device vnodes used by
49 * filesystems to manage meta-data.
53 #include <sys/param.h>
54 #include <sys/systm.h>
55 #include <sys/kernel.h>
57 #include <sys/kthread.h>
58 #include <sys/resourcevar.h>
59 #include <sys/signalvar.h>
60 #include <sys/vnode.h>
61 #include <sys/vmmeter.h>
62 #include <sys/sysctl.h>
65 #include <vm/vm_param.h>
67 #include <vm/vm_object.h>
68 #include <vm/vm_page.h>
69 #include <vm/vm_map.h>
70 #include <vm/vm_pageout.h>
71 #include <vm/vm_pager.h>
72 #include <vm/swap_pager.h>
73 #include <vm/vm_extern.h>
75 #include <sys/thread2.h>
76 #include <vm/vm_page2.h>
78 #define INACTIVE_LIST (&vm_page_queues[PQ_INACTIVE].pl)
80 /* the kernel process "vm_pageout"*/
81 static void vm_swapcached (void);
82 static void vm_swapcached_flush (vm_page_t m);
83 struct thread *swapcached_thread;
85 static struct kproc_desc swpc_kp = {
90 SYSINIT(swapcached, SI_SUB_KTHREAD_PAGE, SI_ORDER_SECOND, kproc_start, &swpc_kp)
92 SYSCTL_NODE(_vm, OID_AUTO, swapcache, CTLFLAG_RW, NULL, NULL);
94 static int vm_swapcache_sleep;
95 static int vm_swapcache_maxlaunder = 64;
96 static int vm_swapcache_data_enable = 0;
97 static int vm_swapcache_meta_enable = 0;
98 static int64_t vm_swapcache_write_count;
100 SYSCTL_INT(_vm_swapcache, OID_AUTO, maxlaunder,
101 CTLFLAG_RW, &vm_swapcache_maxlaunder, 0, "");
102 SYSCTL_INT(_vm_swapcache, OID_AUTO, data_enable,
103 CTLFLAG_RW, &vm_swapcache_data_enable, 0, "");
104 SYSCTL_INT(_vm_swapcache, OID_AUTO, meta_enable,
105 CTLFLAG_RW, &vm_swapcache_meta_enable, 0, "");
106 SYSCTL_QUAD(_vm_swapcache, OID_AUTO, write_count,
107 CTLFLAG_RW, &vm_swapcache_write_count, 0, "");
110 * vm_swapcached is the high level pageout daemon.
115 struct vm_page marker;
123 curthread->td_flags |= TDF_SYSTHREAD;
126 * Initialize our marker
128 bzero(&marker, sizeof(marker));
129 marker.flags = PG_BUSY | PG_FICTITIOUS | PG_MARKER;
130 marker.queue = PQ_INACTIVE;
131 marker.wire_count = 1;
134 TAILQ_INSERT_HEAD(INACTIVE_LIST, &marker, pageq);
138 * Loop once a second or so looking for work when enabled.
140 if (vm_swapcache_data_enable == 0 &&
141 vm_swapcache_meta_enable == 0) {
142 tsleep(&vm_swapcache_sleep, 0, "csleep", hz * 5);
145 tsleep(&vm_swapcache_sleep, 0, "csleep", hz);
148 * Calculate the number of pages to test. We don't want
149 * to get into a cpu-bound loop.
151 count = vmstats.v_inactive_count;
152 if (count > vm_swapcache_maxlaunder)
153 count = vm_swapcache_maxlaunder;
156 * Scan the inactive queue from our marker to locate
157 * suitable pages to push to the swap cache.
159 * We are looking for clean vnode-backed pages.
162 while ((m = TAILQ_NEXT(m, pageq)) != NULL && count--) {
163 if (m->flags & PG_MARKER) {
167 if (m->flags & (PG_SWAPPED | PG_BUSY | PG_UNMANAGED))
169 if (m->busy || m->hold_count || m->wire_count)
171 if (m->valid != VM_PAGE_BITS_ALL)
173 if (m->dirty & m->valid)
175 if ((object = m->object) == NULL)
177 if (object->type != OBJT_VNODE)
179 vm_page_test_dirty(m);
180 if (m->dirty & m->valid)
184 * Ok, move the marker and soft-busy the page.
186 TAILQ_REMOVE(INACTIVE_LIST, &marker, pageq);
187 TAILQ_INSERT_AFTER(INACTIVE_LIST, m, &marker, pageq);
190 * Assign swap and initiate I/O
192 vm_swapcached_flush(m);
195 * Setup for next loop using marker.
199 TAILQ_REMOVE(INACTIVE_LIST, &marker, pageq);
201 TAILQ_INSERT_BEFORE(m, &marker, pageq);
203 TAILQ_INSERT_HEAD(INACTIVE_LIST, &marker, pageq);
206 TAILQ_REMOVE(INACTIVE_LIST, &marker, pageq);
211 * Flush the specified page using the swap_pager.
215 vm_swapcached_flush(vm_page_t m)
221 vm_page_protect(m, VM_PROT_READ);
224 vm_object_pip_add(object, 1);
225 swap_pager_putpages(object, &m, 1, FALSE, &rtvals);
227 if (rtvals != VM_PAGER_PEND) {
228 vm_object_pip_wakeup(object);
229 vm_page_io_finish(m);