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 * 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.
19 *
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
30 * SUCH DAMAGE.
31 *
32 *	from: @(#)vm_pager.c	8.6 (Berkeley) 1/12/94
33 *
34 *
35 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
36 * All rights reserved.
37 *
38 * Authors: Avadis Tevanian, Jr., Michael Wayne Young
39 *
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.
45 *
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.
49 *
50 * Carnegie Mellon requests users of this software to return to
51 *
52 *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
53 *  School of Computer Science
54 *  Carnegie Mellon University
55 *  Pittsburgh PA 15213-3890
56 *
57 * any improvements or extensions that they make and grant Carnegie the
58 * rights to redistribute these changes.
59 */
60
61/*
62 *	Paging space routine stubs.  Emulates a matchmaker-like interface
63 *	for builtin pagers.
64 */
65
66#include <sys/cdefs.h>
67__FBSDID("$FreeBSD$");
68
69#include <sys/param.h>
70#include <sys/systm.h>
71#include <sys/kernel.h>
72#include <sys/vnode.h>
73#include <sys/bio.h>
74#include <sys/buf.h>
75#include <sys/ucred.h>
76#include <sys/malloc.h>
77
78#include <vm/vm.h>
79#include <vm/vm_param.h>
80#include <vm/vm_object.h>
81#include <vm/vm_page.h>
82#include <vm/vm_pager.h>
83#include <vm/vm_extern.h>
84
85int cluster_pbuf_freecnt = -1;	/* unlimited to begin with */
86
87static int dead_pager_getpages(vm_object_t, vm_page_t *, int, int);
88static vm_object_t dead_pager_alloc(void *, vm_ooffset_t, vm_prot_t,
89    vm_ooffset_t, struct ucred *);
90static void dead_pager_putpages(vm_object_t, vm_page_t *, int, int, int *);
91static boolean_t dead_pager_haspage(vm_object_t, vm_pindex_t, int *, int *);
92static void dead_pager_dealloc(vm_object_t);
93
94static int
95dead_pager_getpages(obj, ma, count, req)
96	vm_object_t obj;
97	vm_page_t *ma;
98	int count;
99	int req;
100{
101	return VM_PAGER_FAIL;
102}
103
104static vm_object_t
105dead_pager_alloc(void *handle, vm_ooffset_t size, vm_prot_t prot,
106    vm_ooffset_t off, struct ucred *cred)
107{
108	return NULL;
109}
110
111static void
112dead_pager_putpages(object, m, count, flags, rtvals)
113	vm_object_t object;
114	vm_page_t *m;
115	int count;
116	int flags;
117	int *rtvals;
118{
119	int i;
120
121	for (i = 0; i < count; i++) {
122		rtvals[i] = VM_PAGER_AGAIN;
123	}
124}
125
126static int
127dead_pager_haspage(object, pindex, prev, next)
128	vm_object_t object;
129	vm_pindex_t pindex;
130	int *prev;
131	int *next;
132{
133	if (prev)
134		*prev = 0;
135	if (next)
136		*next = 0;
137	return FALSE;
138}
139
140static void
141dead_pager_dealloc(object)
142	vm_object_t object;
143{
144	return;
145}
146
147static struct pagerops deadpagerops = {
148	.pgo_alloc = 	dead_pager_alloc,
149	.pgo_dealloc =	dead_pager_dealloc,
150	.pgo_getpages =	dead_pager_getpages,
151	.pgo_putpages =	dead_pager_putpages,
152	.pgo_haspage =	dead_pager_haspage,
153};
154
155struct pagerops *pagertab[] = {
156	&defaultpagerops,	/* OBJT_DEFAULT */
157	&swappagerops,		/* OBJT_SWAP */
158	&vnodepagerops,		/* OBJT_VNODE */
159	&devicepagerops,	/* OBJT_DEVICE */
160	&physpagerops,		/* OBJT_PHYS */
161	&deadpagerops,		/* OBJT_DEAD */
162	&sgpagerops,		/* OBJT_SG */
163	&mgtdevicepagerops,	/* OBJT_MGTDEVICE */
164};
165
166static const 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 */
176vm_map_t pager_map;
177static int bswneeded;
178static vm_offset_t swapbkva;		/* swap buffers kva */
179struct mtx pbuf_mtx;
180static TAILQ_HEAD(swqueue, buf) bswlist;
181
182void
183vm_pager_init()
184{
185	struct pagerops **pgops;
186
187	TAILQ_INIT(&bswlist);
188	/*
189	 * Initialize known pagers
190	 */
191	for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++)
192		if ((*pgops)->pgo_init != NULL)
193			(*(*pgops)->pgo_init) ();
194}
195
196void
197vm_pager_bufferinit()
198{
199	struct buf *bp;
200	int i;
201
202	mtx_init(&pbuf_mtx, "pbuf mutex", NULL, MTX_DEF);
203	bp = swbuf;
204	/*
205	 * Now set up swap and physical I/O buffer headers.
206	 */
207	for (i = 0; i < nswbuf; i++, bp++) {
208		TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist);
209		BUF_LOCKINIT(bp);
210		LIST_INIT(&bp->b_dep);
211		bp->b_rcred = bp->b_wcred = NOCRED;
212		bp->b_xflags = 0;
213	}
214
215	cluster_pbuf_freecnt = nswbuf / 2;
216	vnode_pbuf_freecnt = nswbuf / 2 + 1;
217
218	swapbkva = kmem_alloc_nofault(pager_map, nswbuf * MAXPHYS);
219	if (!swapbkva)
220		panic("Not enough pager_map VM space for physical buffers");
221}
222
223/*
224 * Allocate an instance of a pager of the given type.
225 * Size, protection and offset parameters are passed in for pagers that
226 * need to perform page-level validation (e.g. the device pager).
227 */
228vm_object_t
229vm_pager_allocate(objtype_t type, void *handle, vm_ooffset_t size,
230    vm_prot_t prot, vm_ooffset_t off, struct ucred *cred)
231{
232	vm_object_t ret;
233	struct pagerops *ops;
234
235	ops = pagertab[type];
236	if (ops)
237		ret = (*ops->pgo_alloc) (handle, size, prot, off, cred);
238	else
239		ret = NULL;
240	return (ret);
241}
242
243/*
244 *	The object must be locked.
245 */
246void
247vm_pager_deallocate(object)
248	vm_object_t object;
249{
250
251	VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
252	(*pagertab[object->type]->pgo_dealloc) (object);
253}
254
255/*
256 * vm_pager_get_pages() - inline, see vm/vm_pager.h
257 * vm_pager_put_pages() - inline, see vm/vm_pager.h
258 * vm_pager_has_page() - inline, see vm/vm_pager.h
259 */
260
261/*
262 * Search the specified pager object list for an object with the
263 * specified handle.  If an object with the specified handle is found,
264 * increase its reference count and return it.  Otherwise, return NULL.
265 *
266 * The pager object list must be locked.
267 */
268vm_object_t
269vm_pager_object_lookup(struct pagerlst *pg_list, void *handle)
270{
271	vm_object_t object;
272
273	TAILQ_FOREACH(object, pg_list, pager_object_list) {
274		if (object->handle == handle) {
275			VM_OBJECT_LOCK(object);
276			if ((object->flags & OBJ_DEAD) == 0) {
277				vm_object_reference_locked(object);
278				VM_OBJECT_UNLOCK(object);
279				break;
280			}
281			VM_OBJECT_UNLOCK(object);
282		}
283	}
284	return (object);
285}
286
287/*
288 * initialize a physical buffer
289 */
290
291/*
292 * XXX This probably belongs in vfs_bio.c
293 */
294static void
295initpbuf(struct buf *bp)
296{
297	KASSERT(bp->b_bufobj == NULL, ("initpbuf with bufobj"));
298	KASSERT(bp->b_vp == NULL, ("initpbuf with vp"));
299	bp->b_rcred = NOCRED;
300	bp->b_wcred = NOCRED;
301	bp->b_qindex = 0;	/* On no queue (QUEUE_NONE) */
302	bp->b_saveaddr = (caddr_t) (MAXPHYS * (bp - swbuf)) + swapbkva;
303	bp->b_data = bp->b_saveaddr;
304	bp->b_kvabase = bp->b_saveaddr;
305	bp->b_kvasize = MAXPHYS;
306	bp->b_xflags = 0;
307	bp->b_flags = 0;
308	bp->b_ioflags = 0;
309	bp->b_iodone = NULL;
310	bp->b_error = 0;
311	BUF_LOCK(bp, LK_EXCLUSIVE, NULL);
312}
313
314/*
315 * allocate a physical buffer
316 *
317 *	There are a limited number (nswbuf) of physical buffers.  We need
318 *	to make sure that no single subsystem is able to hog all of them,
319 *	so each subsystem implements a counter which is typically initialized
320 *	to 1/2 nswbuf.  getpbuf() decrements this counter in allocation and
321 *	increments it on release, and blocks if the counter hits zero.  A
322 *	subsystem may initialize the counter to -1 to disable the feature,
323 *	but it must still be sure to match up all uses of getpbuf() with
324 *	relpbuf() using the same variable.
325 *
326 *	NOTE: pfreecnt can be NULL, but this 'feature' will be removed
327 *	relatively soon when the rest of the subsystems get smart about it. XXX
328 */
329struct buf *
330getpbuf(int *pfreecnt)
331{
332	struct buf *bp;
333
334	mtx_lock(&pbuf_mtx);
335
336	for (;;) {
337		if (pfreecnt) {
338			while (*pfreecnt == 0) {
339				msleep(pfreecnt, &pbuf_mtx, PVM, "wswbuf0", 0);
340			}
341		}
342
343		/* get a bp from the swap buffer header pool */
344		if ((bp = TAILQ_FIRST(&bswlist)) != NULL)
345			break;
346
347		bswneeded = 1;
348		msleep(&bswneeded, &pbuf_mtx, PVM, "wswbuf1", 0);
349		/* loop in case someone else grabbed one */
350	}
351	TAILQ_REMOVE(&bswlist, bp, b_freelist);
352	if (pfreecnt)
353		--*pfreecnt;
354	mtx_unlock(&pbuf_mtx);
355
356	initpbuf(bp);
357	return bp;
358}
359
360/*
361 * allocate a physical buffer, if one is available.
362 *
363 *	Note that there is no NULL hack here - all subsystems using this
364 *	call understand how to use pfreecnt.
365 */
366struct buf *
367trypbuf(int *pfreecnt)
368{
369	struct buf *bp;
370
371	mtx_lock(&pbuf_mtx);
372	if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist)) == NULL) {
373		mtx_unlock(&pbuf_mtx);
374		return NULL;
375	}
376	TAILQ_REMOVE(&bswlist, bp, b_freelist);
377
378	--*pfreecnt;
379
380	mtx_unlock(&pbuf_mtx);
381
382	initpbuf(bp);
383
384	return bp;
385}
386
387/*
388 * release a physical buffer
389 *
390 *	NOTE: pfreecnt can be NULL, but this 'feature' will be removed
391 *	relatively soon when the rest of the subsystems get smart about it. XXX
392 */
393void
394relpbuf(struct buf *bp, int *pfreecnt)
395{
396
397	if (bp->b_rcred != NOCRED) {
398		crfree(bp->b_rcred);
399		bp->b_rcred = NOCRED;
400	}
401	if (bp->b_wcred != NOCRED) {
402		crfree(bp->b_wcred);
403		bp->b_wcred = NOCRED;
404	}
405
406	KASSERT(bp->b_vp == NULL, ("relpbuf with vp"));
407	KASSERT(bp->b_bufobj == NULL, ("relpbuf with bufobj"));
408
409	BUF_UNLOCK(bp);
410
411	mtx_lock(&pbuf_mtx);
412	TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist);
413
414	if (bswneeded) {
415		bswneeded = 0;
416		wakeup(&bswneeded);
417	}
418	if (pfreecnt) {
419		if (++*pfreecnt == 1)
420			wakeup(pfreecnt);
421	}
422	mtx_unlock(&pbuf_mtx);
423}
424
425/*
426 * Associate a p-buffer with a vnode.
427 *
428 * Also sets B_PAGING flag to indicate that vnode is not fully associated
429 * with the buffer.  i.e. the bp has not been linked into the vnode or
430 * ref-counted.
431 */
432void
433pbgetvp(struct vnode *vp, struct buf *bp)
434{
435
436	KASSERT(bp->b_vp == NULL, ("pbgetvp: not free"));
437	KASSERT(bp->b_bufobj == NULL, ("pbgetvp: not free (bufobj)"));
438
439	bp->b_vp = vp;
440	bp->b_flags |= B_PAGING;
441	bp->b_bufobj = &vp->v_bufobj;
442}
443
444/*
445 * Associate a p-buffer with a vnode.
446 *
447 * Also sets B_PAGING flag to indicate that vnode is not fully associated
448 * with the buffer.  i.e. the bp has not been linked into the vnode or
449 * ref-counted.
450 */
451void
452pbgetbo(struct bufobj *bo, struct buf *bp)
453{
454
455	KASSERT(bp->b_vp == NULL, ("pbgetbo: not free (vnode)"));
456	KASSERT(bp->b_bufobj == NULL, ("pbgetbo: not free (bufobj)"));
457
458	bp->b_flags |= B_PAGING;
459	bp->b_bufobj = bo;
460}
461
462/*
463 * Disassociate a p-buffer from a vnode.
464 */
465void
466pbrelvp(struct buf *bp)
467{
468
469	KASSERT(bp->b_vp != NULL, ("pbrelvp: NULL"));
470	KASSERT(bp->b_bufobj != NULL, ("pbrelvp: NULL bufobj"));
471
472	/* XXX REMOVE ME */
473	BO_LOCK(bp->b_bufobj);
474	if (TAILQ_NEXT(bp, b_bobufs) != NULL) {
475		panic(
476		    "relpbuf(): b_vp was probably reassignbuf()d %p %x",
477		    bp,
478		    (int)bp->b_flags
479		);
480	}
481	BO_UNLOCK(bp->b_bufobj);
482	bp->b_vp = NULL;
483	bp->b_bufobj = NULL;
484	bp->b_flags &= ~B_PAGING;
485}
486
487/*
488 * Disassociate a p-buffer from a bufobj.
489 */
490void
491pbrelbo(struct buf *bp)
492{
493
494	KASSERT(bp->b_vp == NULL, ("pbrelbo: vnode"));
495	KASSERT(bp->b_bufobj != NULL, ("pbrelbo: NULL bufobj"));
496
497	/* XXX REMOVE ME */
498	BO_LOCK(bp->b_bufobj);
499	if (TAILQ_NEXT(bp, b_bobufs) != NULL) {
500		panic(
501		    "relpbuf(): b_vp was probably reassignbuf()d %p %x",
502		    bp,
503		    (int)bp->b_flags
504		);
505	}
506	BO_UNLOCK(bp->b_bufobj);
507	bp->b_bufobj = NULL;
508	bp->b_flags &= ~B_PAGING;
509}
510