2 * Copyright (c) 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * The Mach Operating System project at Carnegie-Mellon University.
8 * Redistribution and use in source and binary forms, with or without
9 * 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 the
15 * documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * from: @(#)vm_pager.c 8.6 (Berkeley) 1/12/94
35 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
36 * All rights reserved.
38 * Authors: Avadis Tevanian, Jr., Michael Wayne Young
40 * Permission to use, copy, modify and distribute this software and
41 * its documentation is hereby granted, provided that both the copyright
42 * notice and this permission notice appear in all copies of the
43 * software, derivative works or modified versions, and any portions
44 * thereof, and that both notices appear in supporting documentation.
46 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
47 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
48 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
50 * Carnegie Mellon requests users of this software to return to
52 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
53 * School of Computer Science
54 * Carnegie Mellon University
55 * Pittsburgh PA 15213-3890
57 * any improvements or extensions that they make and grant Carnegie the
58 * rights to redistribute these changes.
62 * Paging space routine stubs. Emulates a matchmaker-like interface
66 #include <sys/cdefs.h>
67 __FBSDID("$FreeBSD$");
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/kernel.h>
72 #include <sys/vnode.h>
75 #include <sys/ucred.h>
76 #include <sys/malloc.h>
77 #include <sys/rwlock.h>
80 #include <vm/vm_param.h>
81 #include <vm/vm_kern.h>
82 #include <vm/vm_object.h>
83 #include <vm/vm_page.h>
84 #include <vm/vm_pager.h>
85 #include <vm/vm_extern.h>
87 extern vm_page_t bogus_page;
89 int cluster_pbuf_freecnt = -1; /* unlimited to begin with */
93 static int dead_pager_getpages(vm_object_t, vm_page_t *, int, int *, int *);
94 static vm_object_t dead_pager_alloc(void *, vm_ooffset_t, vm_prot_t,
95 vm_ooffset_t, struct ucred *);
96 static void dead_pager_putpages(vm_object_t, vm_page_t *, int, int, int *);
97 static boolean_t dead_pager_haspage(vm_object_t, vm_pindex_t, int *, int *);
98 static void dead_pager_dealloc(vm_object_t);
101 dead_pager_getpages(vm_object_t obj, vm_page_t *ma, int count, int *rbehind,
105 return (VM_PAGER_FAIL);
109 dead_pager_alloc(void *handle, vm_ooffset_t size, vm_prot_t prot,
110 vm_ooffset_t off, struct ucred *cred)
116 dead_pager_putpages(object, m, count, flags, rtvals)
125 for (i = 0; i < count; i++) {
126 rtvals[i] = VM_PAGER_AGAIN;
131 dead_pager_haspage(object, pindex, prev, next)
145 dead_pager_dealloc(object)
151 static struct pagerops deadpagerops = {
152 .pgo_alloc = dead_pager_alloc,
153 .pgo_dealloc = dead_pager_dealloc,
154 .pgo_getpages = dead_pager_getpages,
155 .pgo_putpages = dead_pager_putpages,
156 .pgo_haspage = dead_pager_haspage,
159 struct pagerops *pagertab[] = {
160 &defaultpagerops, /* OBJT_DEFAULT */
161 &swappagerops, /* OBJT_SWAP */
162 &vnodepagerops, /* OBJT_VNODE */
163 &devicepagerops, /* OBJT_DEVICE */
164 &physpagerops, /* OBJT_PHYS */
165 &deadpagerops, /* OBJT_DEAD */
166 &sgpagerops, /* OBJT_SG */
167 &mgtdevicepagerops, /* OBJT_MGTDEVICE */
171 * Kernel address space for mapping pages.
172 * Used by pagers where KVAs are needed for IO.
174 * XXX needs to be large enough to support the number of pending async
175 * cleaning requests (NPENDINGIO == 64) * the maximum swap cluster size
176 * (MAXPHYS == 64k) if you want to get the most efficiency.
178 struct mtx_padalign pbuf_mtx;
179 static TAILQ_HEAD(swqueue, buf) bswlist;
180 static int bswneeded;
181 vm_offset_t swapbkva; /* swap buffers kva */
186 struct pagerops **pgops;
188 TAILQ_INIT(&bswlist);
190 * Initialize known pagers
192 for (pgops = pagertab; pgops < &pagertab[nitems(pagertab)]; pgops++)
193 if ((*pgops)->pgo_init != NULL)
194 (*(*pgops)->pgo_init) ();
198 vm_pager_bufferinit()
203 mtx_init(&pbuf_mtx, "pbuf mutex", NULL, MTX_DEF);
206 * Now set up swap and physical I/O buffer headers.
208 for (i = 0; i < nswbuf; i++, bp++) {
209 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist);
211 LIST_INIT(&bp->b_dep);
212 bp->b_rcred = bp->b_wcred = NOCRED;
216 cluster_pbuf_freecnt = nswbuf / 2;
217 vnode_pbuf_freecnt = nswbuf / 2 + 1;
218 vnode_async_pbuf_freecnt = nswbuf / 2;
222 * Allocate an instance of a pager of the given type.
223 * Size, protection and offset parameters are passed in for pagers that
224 * need to perform page-level validation (e.g. the device pager).
227 vm_pager_allocate(objtype_t type, void *handle, vm_ooffset_t size,
228 vm_prot_t prot, vm_ooffset_t off, struct ucred *cred)
231 struct pagerops *ops;
233 ops = pagertab[type];
235 ret = (*ops->pgo_alloc) (handle, size, prot, off, cred);
242 * The object must be locked.
245 vm_pager_deallocate(object)
249 VM_OBJECT_ASSERT_WLOCKED(object);
250 (*pagertab[object->type]->pgo_dealloc) (object);
254 vm_pager_assert_in(vm_object_t object, vm_page_t *m, int count)
258 VM_OBJECT_ASSERT_WLOCKED(object);
259 KASSERT(count > 0, ("%s: 0 count", __func__));
261 * All pages must be busied, not mapped, not fully valid,
262 * not dirty and belong to the proper object.
264 for (int i = 0 ; i < count; i++) {
265 if (m[i] == bogus_page)
267 vm_page_assert_xbusied(m[i]);
268 KASSERT(!pmap_page_is_mapped(m[i]),
269 ("%s: page %p is mapped", __func__, m[i]));
270 KASSERT(m[i]->valid != VM_PAGE_BITS_ALL,
271 ("%s: request for a valid page %p", __func__, m[i]));
272 KASSERT(m[i]->dirty == 0,
273 ("%s: page %p is dirty", __func__, m[i]));
274 KASSERT(m[i]->object == object,
275 ("%s: wrong object %p/%p", __func__, object, m[i]->object));
281 * Page in the pages for the object using its associated pager.
282 * The requested page must be fully valid on successful return.
285 vm_pager_get_pages(vm_object_t object, vm_page_t *m, int count, int *rbehind,
289 vm_pindex_t pindex = m[0]->pindex;
293 vm_pager_assert_in(object, m, count);
295 r = (*pagertab[object->type]->pgo_getpages)(object, m, count, rbehind,
297 if (r != VM_PAGER_OK)
300 for (int i = 0; i < count; i++) {
302 * If pager has replaced a page, assert that it had
305 KASSERT(m[i] == vm_page_lookup(object, pindex++),
306 ("%s: mismatch page %p pindex %ju", __func__,
307 m[i], (uintmax_t )pindex - 1));
309 * Zero out partially filled data.
311 if (m[i]->valid != VM_PAGE_BITS_ALL)
312 vm_page_zero_invalid(m[i], TRUE);
314 return (VM_PAGER_OK);
318 vm_pager_get_pages_async(vm_object_t object, vm_page_t *m, int count,
319 int *rbehind, int *rahead, pgo_getpages_iodone_t iodone, void *arg)
322 vm_pager_assert_in(object, m, count);
324 return ((*pagertab[object->type]->pgo_getpages_async)(object, m,
325 count, rbehind, rahead, iodone, arg));
329 * vm_pager_put_pages() - inline, see vm/vm_pager.h
330 * vm_pager_has_page() - inline, see vm/vm_pager.h
334 * Search the specified pager object list for an object with the
335 * specified handle. If an object with the specified handle is found,
336 * increase its reference count and return it. Otherwise, return NULL.
338 * The pager object list must be locked.
341 vm_pager_object_lookup(struct pagerlst *pg_list, void *handle)
345 TAILQ_FOREACH(object, pg_list, pager_object_list) {
346 if (object->handle == handle) {
347 VM_OBJECT_WLOCK(object);
348 if ((object->flags & OBJ_DEAD) == 0) {
349 vm_object_reference_locked(object);
350 VM_OBJECT_WUNLOCK(object);
353 VM_OBJECT_WUNLOCK(object);
360 * initialize a physical buffer
364 * XXX This probably belongs in vfs_bio.c
367 initpbuf(struct buf *bp)
369 KASSERT(bp->b_bufobj == NULL, ("initpbuf with bufobj"));
370 KASSERT(bp->b_vp == NULL, ("initpbuf with vp"));
371 bp->b_rcred = NOCRED;
372 bp->b_wcred = NOCRED;
373 bp->b_qindex = 0; /* On no queue (QUEUE_NONE) */
374 bp->b_kvabase = (caddr_t) (MAXPHYS * (bp - swbuf)) + swapbkva;
375 bp->b_data = bp->b_kvabase;
376 bp->b_kvasize = MAXPHYS;
382 BUF_LOCK(bp, LK_EXCLUSIVE, NULL);
383 buf_track(bp, __func__);
387 * allocate a physical buffer
389 * There are a limited number (nswbuf) of physical buffers. We need
390 * to make sure that no single subsystem is able to hog all of them,
391 * so each subsystem implements a counter which is typically initialized
392 * to 1/2 nswbuf. getpbuf() decrements this counter in allocation and
393 * increments it on release, and blocks if the counter hits zero. A
394 * subsystem may initialize the counter to -1 to disable the feature,
395 * but it must still be sure to match up all uses of getpbuf() with
396 * relpbuf() using the same variable.
398 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed
399 * relatively soon when the rest of the subsystems get smart about it. XXX
402 getpbuf(int *pfreecnt)
410 while (*pfreecnt == 0) {
411 msleep(pfreecnt, &pbuf_mtx, PVM, "wswbuf0", 0);
415 /* get a bp from the swap buffer header pool */
416 if ((bp = TAILQ_FIRST(&bswlist)) != NULL)
420 msleep(&bswneeded, &pbuf_mtx, PVM, "wswbuf1", 0);
421 /* loop in case someone else grabbed one */
423 TAILQ_REMOVE(&bswlist, bp, b_freelist);
426 mtx_unlock(&pbuf_mtx);
433 * allocate a physical buffer, if one is available.
435 * Note that there is no NULL hack here - all subsystems using this
436 * call understand how to use pfreecnt.
439 trypbuf(int *pfreecnt)
444 if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist)) == NULL) {
445 mtx_unlock(&pbuf_mtx);
448 TAILQ_REMOVE(&bswlist, bp, b_freelist);
452 mtx_unlock(&pbuf_mtx);
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
466 relpbuf(struct buf *bp, int *pfreecnt)
469 if (bp->b_rcred != NOCRED) {
471 bp->b_rcred = NOCRED;
473 if (bp->b_wcred != NOCRED) {
475 bp->b_wcred = NOCRED;
478 KASSERT(bp->b_vp == NULL, ("relpbuf with vp"));
479 KASSERT(bp->b_bufobj == NULL, ("relpbuf with bufobj"));
481 buf_track(bp, __func__);
485 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist);
492 if (++*pfreecnt == 1)
495 mtx_unlock(&pbuf_mtx);
499 * Associate a p-buffer with a vnode.
501 * Also sets B_PAGING flag to indicate that vnode is not fully associated
502 * with the buffer. i.e. the bp has not been linked into the vnode or
506 pbgetvp(struct vnode *vp, struct buf *bp)
509 KASSERT(bp->b_vp == NULL, ("pbgetvp: not free"));
510 KASSERT(bp->b_bufobj == NULL, ("pbgetvp: not free (bufobj)"));
513 bp->b_flags |= B_PAGING;
514 bp->b_bufobj = &vp->v_bufobj;
518 * Associate a p-buffer with a vnode.
520 * Also sets B_PAGING flag to indicate that vnode is not fully associated
521 * with the buffer. i.e. the bp has not been linked into the vnode or
525 pbgetbo(struct bufobj *bo, struct buf *bp)
528 KASSERT(bp->b_vp == NULL, ("pbgetbo: not free (vnode)"));
529 KASSERT(bp->b_bufobj == NULL, ("pbgetbo: not free (bufobj)"));
531 bp->b_flags |= B_PAGING;
536 * Disassociate a p-buffer from a vnode.
539 pbrelvp(struct buf *bp)
542 KASSERT(bp->b_vp != NULL, ("pbrelvp: NULL"));
543 KASSERT(bp->b_bufobj != NULL, ("pbrelvp: NULL bufobj"));
544 KASSERT((bp->b_xflags & (BX_VNDIRTY | BX_VNCLEAN)) == 0,
545 ("pbrelvp: pager buf on vnode list."));
549 bp->b_flags &= ~B_PAGING;
553 * Disassociate a p-buffer from a bufobj.
556 pbrelbo(struct buf *bp)
559 KASSERT(bp->b_vp == NULL, ("pbrelbo: vnode"));
560 KASSERT(bp->b_bufobj != NULL, ("pbrelbo: NULL bufobj"));
561 KASSERT((bp->b_xflags & (BX_VNDIRTY | BX_VNCLEAN)) == 0,
562 ("pbrelbo: pager buf on vnode list."));
565 bp->b_flags &= ~B_PAGING;