vm_pager.c revision 69399
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: head/sys/vm/vm_pager.c 69399 2000-11-30 18:52:32Z alfred $ 65 */ 66 67/* 68 * Paging space routine stubs. Emulates a matchmaker-like interface 69 * for builtin pagers. 70 */ 71 72#include <sys/param.h> 73#include <sys/systm.h> 74#include <sys/kernel.h> 75#include <sys/vnode.h> 76#include <sys/bio.h> 77#include <sys/buf.h> 78#include <sys/malloc.h> 79 80#include <vm/vm.h> 81#include <vm/vm_param.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> 86 87MALLOC_DEFINE(M_VMPGDATA, "VM pgdata", "XXX: VM pager private data"); 88 89extern struct pagerops defaultpagerops; 90extern struct pagerops swappagerops; 91extern struct pagerops vnodepagerops; 92extern struct pagerops devicepagerops; 93extern struct pagerops physpagerops; 94 95int cluster_pbuf_freecnt = -1; /* unlimited to begin with */ 96 97static int dead_pager_getpages __P((vm_object_t, vm_page_t *, int, int)); 98static vm_object_t dead_pager_alloc __P((void *, vm_ooffset_t, vm_prot_t, 99 vm_ooffset_t)); 100static void dead_pager_putpages __P((vm_object_t, vm_page_t *, int, int, int *)); 101static boolean_t dead_pager_haspage __P((vm_object_t, vm_pindex_t, int *, int *)); 102static void dead_pager_dealloc __P((vm_object_t)); 103 104static int 105dead_pager_getpages(obj, ma, count, req) 106 vm_object_t obj; 107 vm_page_t *ma; 108 int count; 109 int req; 110{ 111 return VM_PAGER_FAIL; 112} 113 114static vm_object_t 115dead_pager_alloc(handle, size, prot, off) 116 void *handle; 117 vm_ooffset_t size; 118 vm_prot_t prot; 119 vm_ooffset_t off; 120{ 121 return NULL; 122} 123 124static void 125dead_pager_putpages(object, m, count, flags, rtvals) 126 vm_object_t object; 127 vm_page_t *m; 128 int count; 129 int flags; 130 int *rtvals; 131{ 132 int i; 133 134 for (i = 0; i < count; i++) { 135 rtvals[i] = VM_PAGER_AGAIN; 136 } 137} 138 139static int 140dead_pager_haspage(object, pindex, prev, next) 141 vm_object_t object; 142 vm_pindex_t pindex; 143 int *prev; 144 int *next; 145{ 146 if (prev) 147 *prev = 0; 148 if (next) 149 *next = 0; 150 return FALSE; 151} 152 153static void 154dead_pager_dealloc(object) 155 vm_object_t object; 156{ 157 return; 158} 159 160static struct pagerops deadpagerops = { 161 NULL, 162 dead_pager_alloc, 163 dead_pager_dealloc, 164 dead_pager_getpages, 165 dead_pager_putpages, 166 dead_pager_haspage, 167 NULL 168}; 169 170struct pagerops *pagertab[] = { 171 &defaultpagerops, /* OBJT_DEFAULT */ 172 &swappagerops, /* OBJT_SWAP */ 173 &vnodepagerops, /* OBJT_VNODE */ 174 &devicepagerops, /* OBJT_DEVICE */ 175 &physpagerops, /* OBJT_PHYS */ 176 &deadpagerops /* OBJT_DEAD */ 177}; 178 179int npagers = sizeof(pagertab) / sizeof(pagertab[0]); 180 181/* 182 * Kernel address space for mapping pages. 183 * Used by pagers where KVAs are needed for IO. 184 * 185 * XXX needs to be large enough to support the number of pending async 186 * cleaning requests (NPENDINGIO == 64) * the maximum swap cluster size 187 * (MAXPHYS == 64k) if you want to get the most efficiency. 188 */ 189#define PAGER_MAP_SIZE (8 * 1024 * 1024) 190 191int pager_map_size = PAGER_MAP_SIZE; 192vm_map_t pager_map; 193static int bswneeded; 194static vm_offset_t swapbkva; /* swap buffers kva */ 195 196void 197vm_pager_init() 198{ 199 struct pagerops **pgops; 200 201 /* 202 * Initialize known pagers 203 */ 204 for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++) 205 if (pgops && ((*pgops)->pgo_init != NULL)) 206 (*(*pgops)->pgo_init) (); 207} 208 209void 210vm_pager_bufferinit() 211{ 212 struct buf *bp; 213 int i; 214 215 bp = swbuf; 216 /* 217 * Now set up swap and physical I/O buffer headers. 218 */ 219 for (i = 0; i < nswbuf; i++, bp++) { 220 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist); 221 BUF_LOCKINIT(bp); 222 LIST_INIT(&bp->b_dep); 223 bp->b_rcred = bp->b_wcred = NOCRED; 224 bp->b_xflags = 0; 225 } 226 227 cluster_pbuf_freecnt = nswbuf / 2; 228 229 swapbkva = kmem_alloc_pageable(pager_map, nswbuf * MAXPHYS); 230 if (!swapbkva) 231 panic("Not enough pager_map VM space for physical buffers"); 232} 233 234/* 235 * Allocate an instance of a pager of the given type. 236 * Size, protection and offset parameters are passed in for pagers that 237 * need to perform page-level validation (e.g. the device pager). 238 */ 239vm_object_t 240vm_pager_allocate(objtype_t type, void *handle, vm_ooffset_t size, vm_prot_t prot, 241 vm_ooffset_t off) 242{ 243 struct pagerops *ops; 244 245 ops = pagertab[type]; 246 if (ops) 247 return ((*ops->pgo_alloc) (handle, size, prot, off)); 248 return (NULL); 249} 250 251void 252vm_pager_deallocate(object) 253 vm_object_t object; 254{ 255 (*pagertab[object->type]->pgo_dealloc) (object); 256} 257 258/* 259 * vm_pager_strategy: 260 * 261 * called with no specific spl 262 * Execute strategy routine directly to pager. 263 */ 264 265void 266vm_pager_strategy(vm_object_t object, struct bio *bp) 267{ 268 if (pagertab[object->type]->pgo_strategy) { 269 (*pagertab[object->type]->pgo_strategy)(object, bp); 270 } else { 271 bp->bio_flags |= BIO_ERROR; 272 bp->bio_error = ENXIO; 273 biodone(bp); 274 } 275} 276 277/* 278 * vm_pager_get_pages() - inline, see vm/vm_pager.h 279 * vm_pager_put_pages() - inline, see vm/vm_pager.h 280 * vm_pager_has_page() - inline, see vm/vm_pager.h 281 * vm_pager_page_inserted() - inline, see vm/vm_pager.h 282 * vm_pager_page_removed() - inline, see vm/vm_pager.h 283 */ 284 285#if 0 286/* 287 * vm_pager_sync: 288 * 289 * Called by pageout daemon before going back to sleep. 290 * Gives pagers a chance to clean up any completed async pageing 291 * operations. 292 */ 293void 294vm_pager_sync() 295{ 296 struct pagerops **pgops; 297 298 for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++) 299 if (pgops && ((*pgops)->pgo_sync != NULL)) 300 (*(*pgops)->pgo_sync) (); 301} 302 303#endif 304 305vm_offset_t 306vm_pager_map_page(m) 307 vm_page_t m; 308{ 309 vm_offset_t kva; 310 311 kva = kmem_alloc_wait(pager_map, PAGE_SIZE); 312 pmap_kenter(kva, VM_PAGE_TO_PHYS(m)); 313 return (kva); 314} 315 316void 317vm_pager_unmap_page(kva) 318 vm_offset_t kva; 319{ 320 pmap_kremove(kva); 321 kmem_free_wakeup(pager_map, kva, PAGE_SIZE); 322} 323 324vm_object_t 325vm_pager_object_lookup(pg_list, handle) 326 register struct pagerlst *pg_list; 327 void *handle; 328{ 329 register vm_object_t object; 330 331 for (object = TAILQ_FIRST(pg_list); object != NULL; object = TAILQ_NEXT(object,pager_object_list)) 332 if (object->handle == handle) 333 return (object); 334 return (NULL); 335} 336 337/* 338 * initialize a physical buffer 339 */ 340 341static void 342initpbuf(struct buf *bp) 343{ 344 bp->b_rcred = NOCRED; 345 bp->b_wcred = NOCRED; 346 bp->b_qindex = QUEUE_NONE; 347 bp->b_data = (caddr_t) (MAXPHYS * (bp - swbuf)) + swapbkva; 348 bp->b_kvabase = bp->b_data; 349 bp->b_kvasize = MAXPHYS; 350 bp->b_xflags = 0; 351 bp->b_flags = 0; 352 bp->b_ioflags = 0; 353 bp->b_iodone = NULL; 354 bp->b_error = 0; 355 BUF_LOCK(bp, LK_EXCLUSIVE); 356} 357 358/* 359 * allocate a physical buffer 360 * 361 * There are a limited number (nswbuf) of physical buffers. We need 362 * to make sure that no single subsystem is able to hog all of them, 363 * so each subsystem implements a counter which is typically initialized 364 * to 1/2 nswbuf. getpbuf() decrements this counter in allocation and 365 * increments it on release, and blocks if the counter hits zero. A 366 * subsystem may initialize the counter to -1 to disable the feature, 367 * but it must still be sure to match up all uses of getpbuf() with 368 * relpbuf() using the same variable. 369 * 370 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed 371 * relatively soon when the rest of the subsystems get smart about it. XXX 372 */ 373struct buf * 374getpbuf(pfreecnt) 375 int *pfreecnt; 376{ 377 int s; 378 struct buf *bp; 379 380 s = splvm(); 381 382 for (;;) { 383 if (pfreecnt) { 384 while (*pfreecnt == 0) { 385 tsleep(pfreecnt, PVM, "wswbuf0", 0); 386 } 387 } 388 389 /* get a bp from the swap buffer header pool */ 390 if ((bp = TAILQ_FIRST(&bswlist)) != NULL) 391 break; 392 393 bswneeded = 1; 394 tsleep(&bswneeded, PVM, "wswbuf1", 0); 395 /* loop in case someone else grabbed one */ 396 } 397 TAILQ_REMOVE(&bswlist, bp, b_freelist); 398 if (pfreecnt) 399 --*pfreecnt; 400 splx(s); 401 402 initpbuf(bp); 403 return bp; 404} 405 406/* 407 * allocate a physical buffer, if one is available. 408 * 409 * Note that there is no NULL hack here - all subsystems using this 410 * call understand how to use pfreecnt. 411 */ 412struct buf * 413trypbuf(pfreecnt) 414 int *pfreecnt; 415{ 416 int s; 417 struct buf *bp; 418 419 s = splvm(); 420 if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist)) == NULL) { 421 splx(s); 422 return NULL; 423 } 424 TAILQ_REMOVE(&bswlist, bp, b_freelist); 425 426 --*pfreecnt; 427 428 splx(s); 429 430 initpbuf(bp); 431 432 return bp; 433} 434 435/* 436 * release a physical buffer 437 * 438 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed 439 * relatively soon when the rest of the subsystems get smart about it. XXX 440 */ 441void 442relpbuf(bp, pfreecnt) 443 struct buf *bp; 444 int *pfreecnt; 445{ 446 int s; 447 448 s = splvm(); 449 450 if (bp->b_rcred != NOCRED) { 451 crfree(bp->b_rcred); 452 bp->b_rcred = NOCRED; 453 } 454 if (bp->b_wcred != NOCRED) { 455 crfree(bp->b_wcred); 456 bp->b_wcred = NOCRED; 457 } 458 459 if (bp->b_vp) 460 pbrelvp(bp); 461 462 BUF_UNLOCK(bp); 463 464 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist); 465 466 if (bswneeded) { 467 bswneeded = 0; 468 wakeup(&bswneeded); 469 } 470 if (pfreecnt) { 471 if (++*pfreecnt == 1) 472 wakeup(pfreecnt); 473 } 474 splx(s); 475} 476