proc->thread stage 5: BUF/VFS clearance! Remove the ucred argument from
[dragonfly.git] / sys / vm / vm_pager.c
CommitLineData
984263bc
MD
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 $
3b568787 65 * $DragonFly: src/sys/vm/vm_pager.c,v 1.5 2003/06/26 05:55:21 dillon Exp $
984263bc
MD
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
82#include <vm/vm.h>
83#include <vm/vm_param.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>
88
3020e3be
MD
89#include <sys/buf2.h>
90
984263bc
MD
91MALLOC_DEFINE(M_VMPGDATA, "VM pgdata", "XXX: VM pager private data");
92
93extern struct pagerops defaultpagerops;
94extern struct pagerops swappagerops;
95extern struct pagerops vnodepagerops;
96extern struct pagerops devicepagerops;
97extern struct pagerops physpagerops;
98
99int cluster_pbuf_freecnt = -1; /* unlimited to begin with */
100
101static int dead_pager_getpages __P((vm_object_t, vm_page_t *, int, int));
102static vm_object_t dead_pager_alloc __P((void *, vm_ooffset_t, vm_prot_t,
103 vm_ooffset_t));
104static void dead_pager_putpages __P((vm_object_t, vm_page_t *, int, int, int *));
105static boolean_t dead_pager_haspage __P((vm_object_t, vm_pindex_t, int *, int *));
106static void dead_pager_dealloc __P((vm_object_t));
107
108static int
109dead_pager_getpages(obj, ma, count, req)
110 vm_object_t obj;
111 vm_page_t *ma;
112 int count;
113 int req;
114{
115 return VM_PAGER_FAIL;
116}
117
118static vm_object_t
119dead_pager_alloc(handle, size, prot, off)
120 void *handle;
121 vm_ooffset_t size;
122 vm_prot_t prot;
123 vm_ooffset_t off;
124{
125 return NULL;
126}
127
128static void
129dead_pager_putpages(object, m, count, flags, rtvals)
130 vm_object_t object;
131 vm_page_t *m;
132 int count;
133 int flags;
134 int *rtvals;
135{
136 int i;
137
138 for (i = 0; i < count; i++) {
139 rtvals[i] = VM_PAGER_AGAIN;
140 }
141}
142
143static int
144dead_pager_haspage(object, pindex, prev, next)
145 vm_object_t object;
146 vm_pindex_t pindex;
147 int *prev;
148 int *next;
149{
150 if (prev)
151 *prev = 0;
152 if (next)
153 *next = 0;
154 return FALSE;
155}
156
157static void
158dead_pager_dealloc(object)
159 vm_object_t object;
160{
161 return;
162}
163
164static struct pagerops deadpagerops = {
165 NULL,
166 dead_pager_alloc,
167 dead_pager_dealloc,
168 dead_pager_getpages,
169 dead_pager_putpages,
170 dead_pager_haspage,
171 NULL
172};
173
174struct pagerops *pagertab[] = {
175 &defaultpagerops, /* OBJT_DEFAULT */
176 &swappagerops, /* OBJT_SWAP */
177 &vnodepagerops, /* OBJT_VNODE */
178 &devicepagerops, /* OBJT_DEVICE */
179 &physpagerops, /* OBJT_PHYS */
180 &deadpagerops /* OBJT_DEAD */
181};
182
183int npagers = sizeof(pagertab) / sizeof(pagertab[0]);
184
185/*
186 * Kernel address space for mapping pages.
187 * Used by pagers where KVAs are needed for IO.
188 *
189 * XXX needs to be large enough to support the number of pending async
190 * cleaning requests (NPENDINGIO == 64) * the maximum swap cluster size
191 * (MAXPHYS == 64k) if you want to get the most efficiency.
192 */
193#define PAGER_MAP_SIZE (8 * 1024 * 1024)
194
195int pager_map_size = PAGER_MAP_SIZE;
196vm_map_t pager_map;
197static int bswneeded;
198static vm_offset_t swapbkva; /* swap buffers kva */
199
200void
201vm_pager_init()
202{
203 struct pagerops **pgops;
204
205 /*
206 * Initialize known pagers
207 */
208 for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++)
209 if (pgops && ((*pgops)->pgo_init != NULL))
210 (*(*pgops)->pgo_init) ();
211}
212
213void
214vm_pager_bufferinit()
215{
216 struct buf *bp;
217 int i;
218
219 bp = swbuf;
220 /*
221 * Now set up swap and physical I/O buffer headers.
222 */
223 for (i = 0; i < nswbuf; i++, bp++) {
224 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist);
225 BUF_LOCKINIT(bp);
226 LIST_INIT(&bp->b_dep);
984263bc
MD
227 bp->b_xflags = 0;
228 }
229
230 cluster_pbuf_freecnt = nswbuf / 2;
231
232 swapbkva = kmem_alloc_pageable(pager_map, nswbuf * MAXPHYS);
233 if (!swapbkva)
234 panic("Not enough pager_map VM space for physical buffers");
235}
236
237/*
238 * Allocate an instance of a pager of the given type.
239 * Size, protection and offset parameters are passed in for pagers that
240 * need to perform page-level validation (e.g. the device pager).
241 */
242vm_object_t
243vm_pager_allocate(objtype_t type, void *handle, vm_ooffset_t size, vm_prot_t prot,
244 vm_ooffset_t off)
245{
246 struct pagerops *ops;
247
248 ops = pagertab[type];
249 if (ops)
250 return ((*ops->pgo_alloc) (handle, size, prot, off));
251 return (NULL);
252}
253
254void
255vm_pager_deallocate(object)
256 vm_object_t object;
257{
258 (*pagertab[object->type]->pgo_dealloc) (object);
259}
260
261/*
262 * vm_pager_strategy:
263 *
264 * called with no specific spl
265 * Execute strategy routine directly to pager.
266 */
267
268void
269vm_pager_strategy(vm_object_t object, struct buf *bp)
270{
271 if (pagertab[object->type]->pgo_strategy) {
272 (*pagertab[object->type]->pgo_strategy)(object, bp);
273 } else {
274 bp->b_flags |= B_ERROR;
275 bp->b_error = ENXIO;
276 biodone(bp);
277 }
278}
279
280/*
281 * vm_pager_get_pages() - inline, see vm/vm_pager.h
282 * vm_pager_put_pages() - inline, see vm/vm_pager.h
283 * vm_pager_has_page() - inline, see vm/vm_pager.h
284 * vm_pager_page_inserted() - inline, see vm/vm_pager.h
285 * vm_pager_page_removed() - inline, see vm/vm_pager.h
286 */
287
288#if 0
289/*
290 * vm_pager_sync:
291 *
292 * Called by pageout daemon before going back to sleep.
293 * Gives pagers a chance to clean up any completed async pageing
294 * operations.
295 */
296void
297vm_pager_sync()
298{
299 struct pagerops **pgops;
300
301 for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++)
302 if (pgops && ((*pgops)->pgo_sync != NULL))
303 (*(*pgops)->pgo_sync) ();
304}
305
306#endif
307
308vm_offset_t
309vm_pager_map_page(m)
310 vm_page_t m;
311{
312 vm_offset_t kva;
313
314 kva = kmem_alloc_wait(pager_map, PAGE_SIZE);
315 pmap_kenter(kva, VM_PAGE_TO_PHYS(m));
316 return (kva);
317}
318
319void
320vm_pager_unmap_page(kva)
321 vm_offset_t kva;
322{
323 pmap_kremove(kva);
324 kmem_free_wakeup(pager_map, kva, PAGE_SIZE);
325}
326
327vm_object_t
328vm_pager_object_lookup(pg_list, handle)
329 register struct pagerlst *pg_list;
330 void *handle;
331{
332 register vm_object_t object;
333
334 for (object = TAILQ_FIRST(pg_list); object != NULL; object = TAILQ_NEXT(object,pager_object_list))
335 if (object->handle == handle)
336 return (object);
337 return (NULL);
338}
339
340/*
341 * initialize a physical buffer
342 */
343
344static void
345initpbuf(struct buf *bp)
346{
984263bc
MD
347 bp->b_qindex = QUEUE_NONE;
348 bp->b_data = (caddr_t) (MAXPHYS * (bp - swbuf)) + swapbkva;
349 bp->b_kvabase = bp->b_data;
350 bp->b_kvasize = MAXPHYS;
351 bp->b_xflags = 0;
352 bp->b_flags = 0;
353 bp->b_error = 0;
354 BUF_LOCK(bp, LK_EXCLUSIVE);
355}
356
357/*
358 * allocate a physical buffer
359 *
360 * There are a limited number (nswbuf) of physical buffers. We need
361 * to make sure that no single subsystem is able to hog all of them,
362 * so each subsystem implements a counter which is typically initialized
363 * to 1/2 nswbuf. getpbuf() decrements this counter in allocation and
364 * increments it on release, and blocks if the counter hits zero. A
365 * subsystem may initialize the counter to -1 to disable the feature,
366 * but it must still be sure to match up all uses of getpbuf() with
367 * relpbuf() using the same variable.
368 *
369 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed
370 * relatively soon when the rest of the subsystems get smart about it. XXX
371 */
372struct buf *
373getpbuf(pfreecnt)
374 int *pfreecnt;
375{
376 int s;
377 struct buf *bp;
378
379 s = splvm();
380
381 for (;;) {
382 if (pfreecnt) {
383 while (*pfreecnt == 0) {
384 tsleep(pfreecnt, PVM, "wswbuf0", 0);
385 }
386 }
387
388 /* get a bp from the swap buffer header pool */
389 if ((bp = TAILQ_FIRST(&bswlist)) != NULL)
390 break;
391
392 bswneeded = 1;
393 tsleep(&bswneeded, PVM, "wswbuf1", 0);
394 /* loop in case someone else grabbed one */
395 }
396 TAILQ_REMOVE(&bswlist, bp, b_freelist);
397 if (pfreecnt)
398 --*pfreecnt;
399 splx(s);
400
401 initpbuf(bp);
402 return bp;
403}
404
405/*
406 * allocate a physical buffer, if one is available.
407 *
408 * Note that there is no NULL hack here - all subsystems using this
409 * call understand how to use pfreecnt.
410 */
411struct buf *
412trypbuf(pfreecnt)
413 int *pfreecnt;
414{
415 int s;
416 struct buf *bp;
417
418 s = splvm();
419 if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist)) == NULL) {
420 splx(s);
421 return NULL;
422 }
423 TAILQ_REMOVE(&bswlist, bp, b_freelist);
424
425 --*pfreecnt;
426
427 splx(s);
428
429 initpbuf(bp);
430
431 return bp;
432}
433
434/*
435 * release a physical buffer
436 *
437 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed
438 * relatively soon when the rest of the subsystems get smart about it. XXX
439 */
440void
441relpbuf(bp, pfreecnt)
442 struct buf *bp;
443 int *pfreecnt;
444{
445 int s;
446
447 s = splvm();
448
984263bc
MD
449 if (bp->b_vp)
450 pbrelvp(bp);
451
452 BUF_UNLOCK(bp);
453
454 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist);
455
456 if (bswneeded) {
457 bswneeded = 0;
458 wakeup(&bswneeded);
459 }
460 if (pfreecnt) {
461 if (++*pfreecnt == 1)
462 wakeup(pfreecnt);
463 }
464 splx(s);
465}
466
467/********************************************************
468 * CHAINING FUNCTIONS *
469 ********************************************************
470 *
471 * These functions support recursion of I/O operations
472 * on bp's, typically by chaining one or more 'child' bp's
473 * to the parent. Synchronous, asynchronous, and semi-synchronous
474 * chaining is possible.
475 */
476
477/*
478 * vm_pager_chain_iodone:
479 *
480 * io completion routine for child bp. Currently we fudge a bit
481 * on dealing with b_resid. Since users of these routines may issue
482 * multiple children simultaniously, sequencing of the error can be lost.
483 */
484
485static void
486vm_pager_chain_iodone(struct buf *nbp)
487{
488 struct buf *bp;
489
490 if ((bp = nbp->b_chain.parent) != NULL) {
491 if (nbp->b_flags & B_ERROR) {
492 bp->b_flags |= B_ERROR;
493 bp->b_error = nbp->b_error;
494 } else if (nbp->b_resid != 0) {
495 bp->b_flags |= B_ERROR;
496 bp->b_error = EINVAL;
497 } else {
498 bp->b_resid -= nbp->b_bcount;
499 }
500 nbp->b_chain.parent = NULL;
501 --bp->b_chain.count;
502 if (bp->b_flags & B_WANT) {
503 bp->b_flags &= ~B_WANT;
504 wakeup(bp);
505 }
506 if (!bp->b_chain.count && (bp->b_xflags & BX_AUTOCHAINDONE)) {
507 bp->b_xflags &= ~BX_AUTOCHAINDONE;
508 if (bp->b_resid != 0 && !(bp->b_flags & B_ERROR)) {
509 bp->b_flags |= B_ERROR;
510 bp->b_error = EINVAL;
511 }
512 biodone(bp);
513 }
514 }
515 nbp->b_flags |= B_DONE;
516 nbp->b_flags &= ~B_ASYNC;
517 relpbuf(nbp, NULL);
518}
519
520/*
521 * getchainbuf:
522 *
523 * Obtain a physical buffer and chain it to its parent buffer. When
524 * I/O completes, the parent buffer will be B_SIGNAL'd. Errors are
525 * automatically propogated to the parent
526 *
527 * Since these are brand new buffers, we do not have to clear B_INVAL
528 * and B_ERROR because they are already clear.
529 */
530
531struct buf *
532getchainbuf(struct buf *bp, struct vnode *vp, int flags)
533{
534 struct buf *nbp = getpbuf(NULL);
535
536 nbp->b_chain.parent = bp;
537 ++bp->b_chain.count;
538
539 if (bp->b_chain.count > 4)
540 waitchainbuf(bp, 4, 0);
541
542 nbp->b_flags = B_CALL | (bp->b_flags & B_ORDERED) | flags;
984263bc
MD
543 nbp->b_iodone = vm_pager_chain_iodone;
544
984263bc
MD
545 if (vp)
546 pbgetvp(vp, nbp);
547 return(nbp);
548}
549
550void
551flushchainbuf(struct buf *nbp)
552{
553 if (nbp->b_bcount) {
554 nbp->b_bufsize = nbp->b_bcount;
555 if ((nbp->b_flags & B_READ) == 0)
556 nbp->b_dirtyend = nbp->b_bcount;
557 BUF_KERNPROC(nbp);
558 VOP_STRATEGY(nbp->b_vp, nbp);
559 } else {
560 biodone(nbp);
561 }
562}
563
564void
565waitchainbuf(struct buf *bp, int count, int done)
566{
567 int s;
568
569 s = splbio();
570 while (bp->b_chain.count > count) {
571 bp->b_flags |= B_WANT;
572 tsleep(bp, PRIBIO + 4, "bpchain", 0);
573 }
574 if (done) {
575 if (bp->b_resid != 0 && !(bp->b_flags & B_ERROR)) {
576 bp->b_flags |= B_ERROR;
577 bp->b_error = EINVAL;
578 }
579 biodone(bp);
580 }
581 splx(s);
582}
583
584void
585autochaindone(struct buf *bp)
586{
587 int s;
588
589 s = splbio();
590 if (bp->b_chain.count == 0)
591 biodone(bp);
592 else
593 bp->b_xflags |= BX_AUTOCHAINDONE;
594 splx(s);
595}
596