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/dsched.h>
78 #include <sys/sysctl.h>
79 #include <sys/thread2.h>
82 #include <vm/vm_param.h>
83 #include <vm/vm_kern.h>
84 #include <vm/vm_object.h>
85 #include <vm/vm_page.h>
86 #include <vm/vm_pager.h>
87 #include <vm/vm_extern.h>
91 extern struct pagerops defaultpagerops
;
92 extern struct pagerops swappagerops
;
93 extern struct pagerops vnodepagerops
;
94 extern struct pagerops devicepagerops
;
95 extern struct pagerops physpagerops
;
97 int cluster_pbuf_freecnt
= -1; /* unlimited to begin with */
99 static int dead_pager_getpage (vm_object_t
, vm_page_t
*, int);
100 static void dead_pager_putpages (vm_object_t
, vm_page_t
*, int, int, int *);
101 static boolean_t
dead_pager_haspage (vm_object_t
, vm_pindex_t
);
102 static void dead_pager_dealloc (vm_object_t
);
108 dead_pager_getpage(vm_object_t obj
, vm_page_t
*mpp
, int seqaccess
)
110 return VM_PAGER_FAIL
;
117 dead_pager_putpages(vm_object_t object
, vm_page_t
*m
, int count
, int flags
,
122 for (i
= 0; i
< count
; i
++) {
123 rtvals
[i
] = VM_PAGER_AGAIN
;
131 dead_pager_haspage(vm_object_t object
, vm_pindex_t pindex
)
140 dead_pager_dealloc(vm_object_t object
)
142 KKASSERT(object
->swblock_count
== 0);
146 static struct pagerops deadpagerops
= {
153 struct pagerops
*pagertab
[] = {
154 &defaultpagerops
, /* OBJT_DEFAULT */
155 &swappagerops
, /* OBJT_SWAP */
156 &vnodepagerops
, /* OBJT_VNODE */
157 &devicepagerops
, /* OBJT_DEVICE */
158 &devicepagerops
, /* OBJT_MGTDEVICE */
159 &physpagerops
, /* OBJT_PHYS */
160 &deadpagerops
/* OBJT_DEAD */
163 int npagers
= NELEM(pagertab
);
166 * Kernel address space for mapping pages.
167 * Used by pagers where KVAs are needed for IO.
169 * XXX needs to be large enough to support the number of pending async
170 * cleaning requests (NPENDINGIO == 64) * the maximum swap cluster size
171 * (MAXPHYS == 64k) if you want to get the most efficiency.
173 #define PAGER_MAP_SIZE (8 * 1024 * 1024)
175 TAILQ_HEAD(swqueue
, buf
);
177 int pager_map_size
= PAGER_MAP_SIZE
;
178 struct vm_map pager_map
;
180 static int bswneeded_raw
;
181 static int bswneeded_kva
;
182 static long nswbuf_raw
;
183 static struct buf
*swbuf_raw
;
184 static vm_offset_t swapbkva
; /* swap buffers kva */
185 static struct swqueue bswlist_raw
; /* without kva */
186 static struct swqueue bswlist_kva
; /* with kva */
187 static struct spinlock bswspin
= SPINLOCK_INITIALIZER(&bswspin
, "bswspin");
188 static int pbuf_raw_count
;
189 static int pbuf_kva_count
;
191 SYSCTL_INT(_vfs
, OID_AUTO
, pbuf_raw_count
, CTLFLAG_RD
, &pbuf_raw_count
, 0,
192 "Kernel virtual address space reservations");
193 SYSCTL_INT(_vfs
, OID_AUTO
, pbuf_kva_count
, CTLFLAG_RD
, &pbuf_kva_count
, 0,
194 "Kernel raw address space reservations");
197 * Initialize the swap buffer list.
199 * Called from the low level boot code only.
202 vm_pager_init(void *arg __unused
)
204 TAILQ_INIT(&bswlist_raw
);
205 TAILQ_INIT(&bswlist_kva
);
207 SYSINIT(vm_mem
, SI_BOOT1_VM
, SI_ORDER_SECOND
, vm_pager_init
, NULL
)
210 * Called from the low level boot code only.
214 vm_pager_bufferinit(void *dummy __unused
)
220 * Reserve KVM space for pbuf data.
222 swapbkva
= kmem_alloc_pageable(&pager_map
, nswbuf
* MAXPHYS
);
224 panic("Not enough pager_map VM space for physical buffers");
227 * Initial pbuf setup. These pbufs have KVA reservations.
230 for (i
= 0; i
< nswbuf
; ++i
, ++bp
) {
231 bp
->b_kvabase
= (caddr_t
)((intptr_t)i
* MAXPHYS
) + swapbkva
;
232 bp
->b_kvasize
= MAXPHYS
;
235 TAILQ_INSERT_HEAD(&bswlist_kva
, bp
, b_freelist
);
240 * Initial pbuf setup. These pbufs do not have KVA reservations,
241 * so we can have a lot more of them. These are typically used
242 * to massage low level buf/bio requests.
244 * NOTE: We use KM_NOTLBSYNC here to reduce unnecessary IPIs
245 * during startup, which can really slow down emulated
248 nswbuf_raw
= nbuf
* 2;
249 swbuf_raw
= (void *)kmem_alloc3(&kernel_map
,
250 round_page(nswbuf_raw
* sizeof(struct buf
)),
254 for (i
= 0; i
< nswbuf_raw
; ++i
, ++bp
) {
257 TAILQ_INSERT_HEAD(&bswlist_raw
, bp
, b_freelist
);
262 * Allow the clustering code to use half of our pbufs.
264 cluster_pbuf_freecnt
= nswbuf
/ 2;
267 SYSINIT(do_vmpg
, SI_BOOT2_MACHDEP
, SI_ORDER_FIRST
, vm_pager_bufferinit
, NULL
);
273 vm_pager_deallocate(vm_object_t object
)
275 (*pagertab
[object
->type
]->pgo_dealloc
) (object
);
279 * vm_pager_get_pages() - inline, see vm/vm_pager.h
280 * vm_pager_put_pages() - inline, see vm/vm_pager.h
281 * vm_pager_has_page() - inline, see vm/vm_pager.h
282 * vm_pager_page_inserted() - inline, see vm/vm_pager.h
283 * vm_pager_page_removed() - inline, see vm/vm_pager.h
287 * Search the specified pager object list for an object with the
288 * specified handle. If an object with the specified handle is found,
289 * increase its reference count and return it. Otherwise, return NULL.
291 * The pager object list must be locked.
294 vm_pager_object_lookup(struct pagerlst
*pg_list
, void *handle
)
298 TAILQ_FOREACH(object
, pg_list
, pager_object_list
) {
299 if (object
->handle
== handle
) {
300 VM_OBJECT_LOCK(object
);
301 if ((object
->flags
& OBJ_DEAD
) == 0) {
302 vm_object_reference_locked(object
);
303 VM_OBJECT_UNLOCK(object
);
306 VM_OBJECT_UNLOCK(object
);
313 * Initialize a physical buffer.
318 initpbuf(struct buf
*bp
)
320 bp
->b_qindex
= 0; /* BQUEUE_NONE */
321 bp
->b_data
= bp
->b_kvabase
; /* NULL if pbuf sans kva */
322 bp
->b_flags
= B_PAGING
;
323 bp
->b_cmd
= BUF_CMD_DONE
;
326 bp
->b_bufsize
= MAXPHYS
;
328 xio_init(&bp
->b_xio
);
329 BUF_LOCK(bp
, LK_EXCLUSIVE
);
333 * Allocate a physical buffer
335 * There are a limited number (nswbuf) of physical buffers. We need
336 * to make sure that no single subsystem is able to hog all of them,
337 * so each subsystem implements a counter which is typically initialized
338 * to 1/2 nswbuf. getpbuf() decrements this counter in allocation and
339 * increments it on release, and blocks if the counter hits zero. A
340 * subsystem may initialize the counter to -1 to disable the feature,
341 * but it must still be sure to match up all uses of getpbuf() with
342 * relpbuf() using the same variable.
344 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed
345 * relatively soon when the rest of the subsystems get smart about it. XXX
347 * Physical buffers can be with or without KVA space reserved. There
348 * are severe limitations on the ones with KVA reserved, and fewer
349 * limitations on the ones without. getpbuf() gets one without,
350 * getpbuf_kva() gets one with.
355 getpbuf(int *pfreecnt
)
363 while (*pfreecnt
== 0)
364 ssleep(pfreecnt
, &bswspin
, 0, "wswbuf0", 0);
367 /* get a bp from the swap buffer header pool */
368 if ((bp
= TAILQ_FIRST(&bswlist_raw
)) != NULL
)
371 ssleep(&bswneeded_raw
, &bswspin
, 0, "wswbuf1", 0);
372 /* loop in case someone else grabbed one */
374 TAILQ_REMOVE(&bswlist_raw
, bp
, b_freelist
);
379 spin_unlock(&bswspin
);
382 KKASSERT(dsched_is_clear_buf_priv(bp
));
388 getpbuf_kva(int *pfreecnt
)
396 while (*pfreecnt
== 0)
397 ssleep(pfreecnt
, &bswspin
, 0, "wswbuf0", 0);
400 /* get a bp from the swap buffer header pool */
401 if ((bp
= TAILQ_FIRST(&bswlist_kva
)) != NULL
)
404 ssleep(&bswneeded_kva
, &bswspin
, 0, "wswbuf1", 0);
405 /* loop in case someone else grabbed one */
407 TAILQ_REMOVE(&bswlist_kva
, bp
, b_freelist
);
412 spin_unlock(&bswspin
);
415 KKASSERT(dsched_is_clear_buf_priv(bp
));
421 * Allocate a physical buffer, if one is available.
423 * Note that there is no NULL hack here - all subsystems using this
424 * call understand how to use pfreecnt.
429 trypbuf(int *pfreecnt
)
435 if (*pfreecnt
== 0 || (bp
= TAILQ_FIRST(&bswlist_raw
)) == NULL
) {
436 spin_unlock(&bswspin
);
439 TAILQ_REMOVE(&bswlist_raw
, bp
, b_freelist
);
443 spin_unlock(&bswspin
);
451 trypbuf_kva(int *pfreecnt
)
457 if (*pfreecnt
== 0 || (bp
= TAILQ_FIRST(&bswlist_kva
)) == NULL
) {
458 spin_unlock(&bswspin
);
461 TAILQ_REMOVE(&bswlist_kva
, bp
, b_freelist
);
465 spin_unlock(&bswspin
);
473 * Release a physical buffer
475 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed
476 * relatively soon when the rest of the subsystems get smart about it. XXX
481 relpbuf(struct buf
*bp
, int *pfreecnt
)
483 int wake_bsw_kva
= 0;
484 int wake_bsw_raw
= 0;
485 int wake_freecnt
= 0;
487 KKASSERT(bp
->b_flags
& B_PAGING
);
494 TAILQ_INSERT_HEAD(&bswlist_kva
, bp
, b_freelist
);
497 TAILQ_INSERT_HEAD(&bswlist_raw
, bp
, b_freelist
);
509 if (++*pfreecnt
== 1)
512 spin_unlock(&bswspin
);
515 wakeup(&bswneeded_kva
);
517 wakeup(&bswneeded_raw
);
523 pbuf_adjcount(int *pfreecnt
, int n
)
528 spin_unlock(&bswspin
);