xref: /dflybsd-src/sys/vm/vm_pager.c (revision 41871674d0079dec70d55eb824f39d07dc7b3310)
1 /*
2  * Copyright (c) 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * The Mach Operating System project at Carnegie-Mellon University.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
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  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	from: @(#)vm_pager.c	8.6 (Berkeley) 1/12/94
37  *
38  *
39  * Copyright (c) 1987, 1990 Carnegie-Mellon University.
40  * All rights reserved.
41  *
42  * Authors: Avadis Tevanian, Jr., Michael Wayne Young
43  *
44  * Permission to use, copy, modify and distribute this software and
45  * its documentation is hereby granted, provided that both the copyright
46  * notice and this permission notice appear in all copies of the
47  * software, derivative works or modified versions, and any portions
48  * thereof, and that both notices appear in supporting documentation.
49  *
50  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
51  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
52  * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
53  *
54  * Carnegie Mellon requests users of this software to return to
55  *
56  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
57  *  School of Computer Science
58  *  Carnegie Mellon University
59  *  Pittsburgh PA 15213-3890
60  *
61  * any improvements or extensions that they make and grant Carnegie the
62  * rights to redistribute these changes.
63  *
64  * $FreeBSD: src/sys/vm/vm_pager.c,v 1.54.2.2 2001/11/18 07:11:00 dillon Exp $
65  * $DragonFly: src/sys/vm/vm_pager.c,v 1.17 2006/03/27 01:54:18 dillon Exp $
66  */
67 
68 /*
69  *	Paging space routine stubs.  Emulates a matchmaker-like interface
70  *	for builtin pagers.
71  */
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/kernel.h>
76 #include <sys/vnode.h>
77 #include <sys/buf.h>
78 #include <sys/ucred.h>
79 #include <sys/malloc.h>
80 #include <sys/proc.h>
81 #include <sys/thread2.h>
82 
83 #include <vm/vm.h>
84 #include <vm/vm_param.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>
89 
90 #include <sys/buf2.h>
91 
92 MALLOC_DEFINE(M_VMPGDATA, "VM pgdata", "XXX: VM pager private data");
93 
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;
99 
100 int cluster_pbuf_freecnt = -1;	/* unlimited to begin with */
101 
102 static int dead_pager_getpages (vm_object_t, vm_page_t *, int, int);
103 static vm_object_t dead_pager_alloc (void *, off_t, vm_prot_t, off_t);
104 static void dead_pager_putpages (vm_object_t, vm_page_t *, int, int, int *);
105 static boolean_t dead_pager_haspage (vm_object_t, vm_pindex_t, int *, int *);
106 static void dead_pager_dealloc (vm_object_t);
107 
108 static int
109 dead_pager_getpages(vm_object_t obj, vm_page_t *ma, int count, int req)
110 {
111 	return VM_PAGER_FAIL;
112 }
113 
114 static vm_object_t
115 dead_pager_alloc(void *handle, off_t size, vm_prot_t prot, off_t off)
116 {
117 	return NULL;
118 }
119 
120 static void
121 dead_pager_putpages(vm_object_t object, vm_page_t *m, int count, int flags,
122     int *rtvals)
123 {
124 	int i;
125 
126 	for (i = 0; i < count; i++) {
127 		rtvals[i] = VM_PAGER_AGAIN;
128 	}
129 }
130 
131 static int
132 dead_pager_haspage(vm_object_t object, vm_pindex_t pindex, int *prev, int *next)
133 {
134 	if (prev)
135 		*prev = 0;
136 	if (next)
137 		*next = 0;
138 	return FALSE;
139 }
140 
141 static void
142 dead_pager_dealloc(vm_object_t object)
143 {
144 	return;
145 }
146 
147 static struct pagerops deadpagerops = {
148 	NULL,
149 	dead_pager_alloc,
150 	dead_pager_dealloc,
151 	dead_pager_getpages,
152 	dead_pager_putpages,
153 	dead_pager_haspage,
154 	NULL
155 };
156 
157 struct pagerops *pagertab[] = {
158 	&defaultpagerops,	/* OBJT_DEFAULT */
159 	&swappagerops,		/* OBJT_SWAP */
160 	&vnodepagerops,		/* OBJT_VNODE */
161 	&devicepagerops,	/* OBJT_DEVICE */
162 	&physpagerops,		/* OBJT_PHYS */
163 	&deadpagerops		/* OBJT_DEAD */
164 };
165 
166 int npagers = sizeof(pagertab) / sizeof(pagertab[0]);
167 
168 /*
169  * Kernel address space for mapping pages.
170  * Used by pagers where KVAs are needed for IO.
171  *
172  * XXX needs to be large enough to support the number of pending async
173  * cleaning requests (NPENDINGIO == 64) * the maximum swap cluster size
174  * (MAXPHYS == 64k) if you want to get the most efficiency.
175  */
176 #define PAGER_MAP_SIZE	(8 * 1024 * 1024)
177 
178 int pager_map_size = PAGER_MAP_SIZE;
179 vm_map_t pager_map;
180 static int bswneeded;
181 static vm_offset_t swapbkva;		/* swap buffers kva */
182 static TAILQ_HEAD(swqueue, buf) bswlist;
183 
184 void
185 vm_pager_init(void)
186 {
187 	struct pagerops **pgops;
188 
189 	/*
190 	 * Initialize the swap buffer list.
191 	 */
192 	TAILQ_INIT(&bswlist);
193 
194 	/*
195 	 * Initialize known pagers
196 	 */
197 	for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++)
198 		if (pgops && ((*pgops)->pgo_init != NULL))
199 			(*(*pgops)->pgo_init) ();
200 }
201 
202 void
203 vm_pager_bufferinit(void)
204 {
205 	struct buf *bp;
206 	int i;
207 
208 	bp = swbuf;
209 	/*
210 	 * Now set up swap and physical I/O buffer headers.
211 	 */
212 	for (i = 0; i < nswbuf; i++, bp++) {
213 		TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist);
214 		BUF_LOCKINIT(bp);
215 		LIST_INIT(&bp->b_dep);
216 		bp->b_xflags = 0;
217 	}
218 
219 	cluster_pbuf_freecnt = nswbuf / 2;
220 
221 	swapbkva = kmem_alloc_pageable(pager_map, nswbuf * MAXPHYS);
222 	if (!swapbkva)
223 		panic("Not enough pager_map VM space for physical buffers");
224 }
225 
226 /*
227  * Allocate an instance of a pager of the given type.
228  * Size, protection and offset parameters are passed in for pagers that
229  * need to perform page-level validation (e.g. the device pager).
230  */
231 vm_object_t
232 vm_pager_allocate(objtype_t type, void *handle, vm_ooffset_t size,
233 		  vm_prot_t prot, off_t off)
234 {
235 	struct pagerops *ops;
236 
237 	ops = pagertab[type];
238 	if (ops)
239 		return ((*ops->pgo_alloc) (handle, size, prot, off));
240 	return (NULL);
241 }
242 
243 void
244 vm_pager_deallocate(vm_object_t object)
245 {
246 	(*pagertab[object->type]->pgo_dealloc) (object);
247 }
248 
249 /*
250  *      vm_pager_strategy:
251  *
252  *      called with no specific spl
253  *      Execute strategy routine directly to pager.
254  */
255 
256 void
257 vm_pager_strategy(vm_object_t object, struct bio *bio)
258 {
259 	struct buf *bp;
260 
261 	if (pagertab[object->type]->pgo_strategy) {
262 	    (*pagertab[object->type]->pgo_strategy)(object, bio);
263 	} else {
264 		bp = bio->bio_buf;
265 		bp->b_flags |= B_ERROR;
266 		bp->b_error = ENXIO;
267 		biodone(bio);
268 	}
269 }
270 
271 /*
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
277  */
278 
279 #if 0
280 /*
281  *	vm_pager_sync:
282  *
283  *	Called by pageout daemon before going back to sleep.
284  *	Gives pagers a chance to clean up any completed async pageing
285  *	operations.
286  */
287 void
288 vm_pager_sync(void)
289 {
290 	struct pagerops **pgops;
291 
292 	for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++)
293 		if (pgops && ((*pgops)->pgo_sync != NULL))
294 			(*(*pgops)->pgo_sync) ();
295 }
296 
297 #endif
298 
299 vm_object_t
300 vm_pager_object_lookup(struct pagerlst *pg_list, void *handle)
301 {
302 	vm_object_t object;
303 
304 	for (object = TAILQ_FIRST(pg_list); object != NULL; object = TAILQ_NEXT(object,pager_object_list))
305 		if (object->handle == handle)
306 			return (object);
307 	return (NULL);
308 }
309 
310 /*
311  * initialize a physical buffer
312  */
313 
314 static void
315 initpbuf(struct buf *bp)
316 {
317 	bp->b_qindex = 0; /* BQUEUE_NONE */
318 	bp->b_data = (caddr_t) (MAXPHYS * (bp - swbuf)) + swapbkva;
319 	bp->b_kvabase = bp->b_data;
320 	bp->b_kvasize = MAXPHYS;
321 	bp->b_xflags = 0;
322 	bp->b_flags = 0;
323 	bp->b_error = 0;
324 	initbufbio(bp);
325 	xio_init(&bp->b_xio);
326 	BUF_LOCK(bp, LK_EXCLUSIVE);
327 }
328 
329 /*
330  * allocate a physical buffer
331  *
332  *	There are a limited number (nswbuf) of physical buffers.  We need
333  *	to make sure that no single subsystem is able to hog all of them,
334  *	so each subsystem implements a counter which is typically initialized
335  *	to 1/2 nswbuf.  getpbuf() decrements this counter in allocation and
336  *	increments it on release, and blocks if the counter hits zero.  A
337  *	subsystem may initialize the counter to -1 to disable the feature,
338  *	but it must still be sure to match up all uses of getpbuf() with
339  *	relpbuf() using the same variable.
340  *
341  *	NOTE: pfreecnt can be NULL, but this 'feature' will be removed
342  *	relatively soon when the rest of the subsystems get smart about it. XXX
343  */
344 struct buf *
345 getpbuf(int *pfreecnt)
346 {
347 	struct buf *bp;
348 
349 	crit_enter();
350 
351 	for (;;) {
352 		if (pfreecnt) {
353 			while (*pfreecnt == 0) {
354 				tsleep(pfreecnt, 0, "wswbuf0", 0);
355 			}
356 		}
357 
358 		/* get a bp from the swap buffer header pool */
359 		if ((bp = TAILQ_FIRST(&bswlist)) != NULL)
360 			break;
361 
362 		bswneeded = 1;
363 		tsleep(&bswneeded, 0, "wswbuf1", 0);
364 		/* loop in case someone else grabbed one */
365 	}
366 	TAILQ_REMOVE(&bswlist, bp, b_freelist);
367 	if (pfreecnt)
368 		--*pfreecnt;
369 	crit_exit();
370 
371 	initpbuf(bp);
372 	return bp;
373 }
374 
375 /*
376  * allocate a physical buffer, if one is available.
377  *
378  *	Note that there is no NULL hack here - all subsystems using this
379  *	call understand how to use pfreecnt.
380  */
381 struct buf *
382 trypbuf(int *pfreecnt)
383 {
384 	struct buf *bp;
385 
386 	crit_enter();
387 	if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist)) == NULL) {
388 		crit_exit();
389 		return NULL;
390 	}
391 	TAILQ_REMOVE(&bswlist, bp, b_freelist);
392 
393 	--*pfreecnt;
394 
395 	crit_exit();
396 
397 	initpbuf(bp);
398 
399 	return bp;
400 }
401 
402 /*
403  * release a physical buffer
404  *
405  *	NOTE: pfreecnt can be NULL, but this 'feature' will be removed
406  *	relatively soon when the rest of the subsystems get smart about it. XXX
407  */
408 void
409 relpbuf(struct buf *bp, int *pfreecnt)
410 {
411 	crit_enter();
412 
413 	if (bp->b_vp)
414 		pbrelvp(bp);
415 
416 	BUF_UNLOCK(bp);
417 
418 	TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist);
419 
420 	if (bswneeded) {
421 		bswneeded = 0;
422 		wakeup(&bswneeded);
423 	}
424 	if (pfreecnt) {
425 		if (++*pfreecnt == 1)
426 			wakeup(pfreecnt);
427 	}
428 	crit_exit();
429 }
430 
431