4 * Copyright (c) 1991, 1993
5 * The Regents of the University of California. All rights reserved.
7 * This code is derived from software contributed to Berkeley by
8 * The Mach Operating System project at Carnegie-Mellon University.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * from: @(#)vm_pager.c 8.6 (Berkeley) 1/12/94
37 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
38 * All rights reserved.
40 * Authors: Avadis Tevanian, Jr., Michael Wayne Young
42 * Permission to use, copy, modify and distribute this software and
43 * its documentation is hereby granted, provided that both the copyright
44 * notice and this permission notice appear in all copies of the
45 * software, derivative works or modified versions, and any portions
46 * thereof, and that both notices appear in supporting documentation.
48 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
49 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
50 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
52 * Carnegie Mellon requests users of this software to return to
54 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
55 * School of Computer Science
56 * Carnegie Mellon University
57 * Pittsburgh PA 15213-3890
59 * any improvements or extensions that they make and grant Carnegie the
60 * rights to redistribute these changes.
62 * $FreeBSD: src/sys/vm/vm_pager.c,v 1.54.2.2 2001/11/18 07:11:00 dillon Exp $
66 * Paging space routine stubs. Emulates a matchmaker-like interface
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/kernel.h>
73 #include <sys/vnode.h>
75 #include <sys/ucred.h>
76 #include <sys/malloc.h>
77 #include <sys/dsched.h>
79 #include <sys/sysctl.h>
80 #include <sys/thread2.h>
83 #include <vm/vm_param.h>
84 #include <vm/vm_kern.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/vm_pager.h>
88 #include <vm/vm_extern.h>
92 MALLOC_DEFINE(M_VMPGDATA, "VM pgdata", "XXX: VM pager private data");
94 extern struct pagerops defaultpagerops;
95 extern struct pagerops swappagerops;
96 extern struct pagerops vnodepagerops;
97 extern struct pagerops devicepagerops;
98 extern struct pagerops physpagerops;
100 int cluster_pbuf_freecnt = -1; /* unlimited to begin with */
102 static int dead_pager_getpage (vm_object_t, vm_page_t *, int);
103 static void dead_pager_putpages (vm_object_t, vm_page_t *, int, int, int *);
104 static boolean_t dead_pager_haspage (vm_object_t, vm_pindex_t);
105 static void dead_pager_dealloc (vm_object_t);
111 dead_pager_getpage(vm_object_t obj, vm_page_t *mpp, int seqaccess)
113 return VM_PAGER_FAIL;
120 dead_pager_putpages(vm_object_t object, vm_page_t *m, int count, int flags,
125 for (i = 0; i < count; i++) {
126 rtvals[i] = VM_PAGER_AGAIN;
134 dead_pager_haspage(vm_object_t object, vm_pindex_t pindex)
143 dead_pager_dealloc(vm_object_t object)
145 KKASSERT(object->swblock_count == 0);
149 static struct pagerops deadpagerops = {
156 struct pagerops *pagertab[] = {
157 &defaultpagerops, /* OBJT_DEFAULT */
158 &swappagerops, /* OBJT_SWAP */
159 &vnodepagerops, /* OBJT_VNODE */
160 &devicepagerops, /* OBJT_DEVICE */
161 &physpagerops, /* OBJT_PHYS */
162 &deadpagerops /* OBJT_DEAD */
165 int npagers = NELEM(pagertab);
168 * Kernel address space for mapping pages.
169 * Used by pagers where KVAs are needed for IO.
171 * XXX needs to be large enough to support the number of pending async
172 * cleaning requests (NPENDINGIO == 64) * the maximum swap cluster size
173 * (MAXPHYS == 64k) if you want to get the most efficiency.
175 #define PAGER_MAP_SIZE (8 * 1024 * 1024)
177 TAILQ_HEAD(swqueue, buf);
179 int pager_map_size = PAGER_MAP_SIZE;
180 struct vm_map pager_map;
182 static int bswneeded_raw;
183 static int bswneeded_kva;
184 static long nswbuf_raw;
185 static struct buf *swbuf_raw;
186 static vm_offset_t swapbkva; /* swap buffers kva */
187 static struct swqueue bswlist_raw; /* without kva */
188 static struct swqueue bswlist_kva; /* with kva */
189 static struct spinlock bswspin = SPINLOCK_INITIALIZER(&bswspin);
190 static int pbuf_raw_count;
191 static int pbuf_kva_count;
193 SYSCTL_INT(_vfs, OID_AUTO, pbuf_raw_count, CTLFLAG_RD, &pbuf_raw_count, 0,
194 "Kernel virtual address space reservations");
195 SYSCTL_INT(_vfs, OID_AUTO, pbuf_kva_count, CTLFLAG_RD, &pbuf_kva_count, 0,
196 "Kernel raw address space reservations");
199 * Initialize the swap buffer list.
201 * Called from the low level boot code only.
204 vm_pager_init(void *arg __unused)
206 TAILQ_INIT(&bswlist_raw);
207 TAILQ_INIT(&bswlist_kva);
209 SYSINIT(vm_mem, SI_BOOT1_VM, SI_ORDER_SECOND, vm_pager_init, NULL)
212 * Called from the low level boot code only.
215 vm_pager_bufferinit(void)
221 * Reserve KVM space for pbuf data.
223 swapbkva = kmem_alloc_pageable(&pager_map, nswbuf * MAXPHYS);
225 panic("Not enough pager_map VM space for physical buffers");
228 * Initial pbuf setup. These pbufs have KVA reservations.
231 for (i = 0; i < nswbuf; ++i, ++bp) {
232 bp->b_kvabase = (caddr_t)((intptr_t)i * MAXPHYS) + swapbkva;
233 bp->b_kvasize = MAXPHYS;
236 TAILQ_INSERT_HEAD(&bswlist_kva, bp, b_freelist);
241 * Initial pbuf setup. These pbufs do not have KVA reservations,
242 * so we can have a lot more of them. These are typically used
243 * to massage low level buf/bio requests.
245 nswbuf_raw = nbuf * 2;
246 swbuf_raw = (void *)kmem_alloc(&kernel_map,
247 round_page(nswbuf_raw * sizeof(struct buf)));
249 for (i = 0; i < nswbuf_raw; ++i, ++bp) {
252 TAILQ_INSERT_HEAD(&bswlist_raw, bp, b_freelist);
257 * Allow the clustering code to use half of our pbufs.
259 cluster_pbuf_freecnt = nswbuf / 2;
266 vm_pager_deallocate(vm_object_t object)
268 (*pagertab[object->type]->pgo_dealloc) (object);
272 * vm_pager_get_pages() - inline, see vm/vm_pager.h
273 * vm_pager_put_pages() - inline, see vm/vm_pager.h
274 * vm_pager_has_page() - inline, see vm/vm_pager.h
275 * vm_pager_page_inserted() - inline, see vm/vm_pager.h
276 * vm_pager_page_removed() - inline, see vm/vm_pager.h
283 * Called by pageout daemon before going back to sleep.
284 * Gives pagers a chance to clean up any completed async pageing
290 struct pagerops **pgops;
292 for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++)
293 if (pgops && ((*pgops)->pgo_sync != NULL))
294 (*(*pgops)->pgo_sync) ();
300 * Initialize a physical buffer.
305 initpbuf(struct buf *bp)
307 bp->b_qindex = 0; /* BQUEUE_NONE */
308 bp->b_data = bp->b_kvabase; /* NULL if pbuf sans kva */
309 bp->b_flags = B_PAGING;
310 bp->b_cmd = BUF_CMD_DONE;
313 bp->b_bufsize = MAXPHYS;
315 xio_init(&bp->b_xio);
316 BUF_LOCK(bp, LK_EXCLUSIVE);
320 * Allocate a physical buffer
322 * There are a limited number (nswbuf) of physical buffers. We need
323 * to make sure that no single subsystem is able to hog all of them,
324 * so each subsystem implements a counter which is typically initialized
325 * to 1/2 nswbuf. getpbuf() decrements this counter in allocation and
326 * increments it on release, and blocks if the counter hits zero. A
327 * subsystem may initialize the counter to -1 to disable the feature,
328 * but it must still be sure to match up all uses of getpbuf() with
329 * relpbuf() using the same variable.
331 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed
332 * relatively soon when the rest of the subsystems get smart about it. XXX
334 * Physical buffers can be with or without KVA space reserved. There
335 * are severe limitations on the ones with KVA reserved, and fewer
336 * limitations on the ones without. getpbuf() gets one without,
337 * getpbuf_kva() gets one with.
342 getpbuf(int *pfreecnt)
350 while (*pfreecnt == 0)
351 ssleep(pfreecnt, &bswspin, 0, "wswbuf0", 0);
354 /* get a bp from the swap buffer header pool */
355 if ((bp = TAILQ_FIRST(&bswlist_raw)) != NULL)
358 ssleep(&bswneeded_raw, &bswspin, 0, "wswbuf1", 0);
359 /* loop in case someone else grabbed one */
361 TAILQ_REMOVE(&bswlist_raw, bp, b_freelist);
366 spin_unlock(&bswspin);
369 KKASSERT(dsched_is_clear_buf_priv(bp));
375 getpbuf_kva(int *pfreecnt)
383 while (*pfreecnt == 0)
384 ssleep(pfreecnt, &bswspin, 0, "wswbuf0", 0);
387 /* get a bp from the swap buffer header pool */
388 if ((bp = TAILQ_FIRST(&bswlist_kva)) != NULL)
391 ssleep(&bswneeded_kva, &bswspin, 0, "wswbuf1", 0);
392 /* loop in case someone else grabbed one */
394 TAILQ_REMOVE(&bswlist_kva, bp, b_freelist);
399 spin_unlock(&bswspin);
402 KKASSERT(dsched_is_clear_buf_priv(bp));
408 * Allocate a physical buffer, if one is available.
410 * Note that there is no NULL hack here - all subsystems using this
411 * call understand how to use pfreecnt.
416 trypbuf(int *pfreecnt)
422 if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist_raw)) == NULL) {
423 spin_unlock(&bswspin);
426 TAILQ_REMOVE(&bswlist_raw, bp, b_freelist);
430 spin_unlock(&bswspin);
438 trypbuf_kva(int *pfreecnt)
444 if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist_kva)) == NULL) {
445 spin_unlock(&bswspin);
448 TAILQ_REMOVE(&bswlist_kva, bp, b_freelist);
452 spin_unlock(&bswspin);
460 * Release a physical buffer
462 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed
463 * relatively soon when the rest of the subsystems get smart about it. XXX
468 relpbuf(struct buf *bp, int *pfreecnt)
470 int wake_bsw_kva = 0;
471 int wake_bsw_raw = 0;
472 int wake_freecnt = 0;
474 KKASSERT(bp->b_flags & B_PAGING);
481 TAILQ_INSERT_HEAD(&bswlist_kva, bp, b_freelist);
484 TAILQ_INSERT_HEAD(&bswlist_raw, bp, b_freelist);
496 if (++*pfreecnt == 1)
499 spin_unlock(&bswspin);
502 wakeup(&bswneeded_kva);
504 wakeup(&bswneeded_raw);
510 pbuf_adjcount(int *pfreecnt, int n)
515 spin_unlock(&bswspin);