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 1.16 2006/02/17 19:18:08 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 *, vm_ooffset_t, vm_prot_t, 104 vm_ooffset_t); 105 static void dead_pager_putpages (vm_object_t, vm_page_t *, int, int, int *); 106 static boolean_t dead_pager_haspage (vm_object_t, vm_pindex_t, int *, int *); 107 static void dead_pager_dealloc (vm_object_t); 108 109 static int 110 dead_pager_getpages(vm_object_t obj, vm_page_t *ma, int count, int req) 111 { 112 return VM_PAGER_FAIL; 113 } 114 115 static vm_object_t 116 dead_pager_alloc(void *handle, vm_ooffset_t size, vm_prot_t prot, 117 vm_ooffset_t off) 118 { 119 return NULL; 120 } 121 122 static void 123 dead_pager_putpages(vm_object_t object, vm_page_t *m, int count, int flags, 124 int *rtvals) 125 { 126 int i; 127 128 for (i = 0; i < count; i++) { 129 rtvals[i] = VM_PAGER_AGAIN; 130 } 131 } 132 133 static int 134 dead_pager_haspage(vm_object_t object, vm_pindex_t pindex, int *prev, int *next) 135 { 136 if (prev) 137 *prev = 0; 138 if (next) 139 *next = 0; 140 return FALSE; 141 } 142 143 static void 144 dead_pager_dealloc(vm_object_t object) 145 { 146 return; 147 } 148 149 static struct pagerops deadpagerops = { 150 NULL, 151 dead_pager_alloc, 152 dead_pager_dealloc, 153 dead_pager_getpages, 154 dead_pager_putpages, 155 dead_pager_haspage, 156 NULL 157 }; 158 159 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 }; 167 168 int npagers = sizeof(pagertab) / sizeof(pagertab[0]); 169 170 /* 171 * Kernel address space for mapping pages. 172 * Used by pagers where KVAs are needed for IO. 173 * 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. 177 */ 178 #define PAGER_MAP_SIZE (8 * 1024 * 1024) 179 180 int pager_map_size = PAGER_MAP_SIZE; 181 vm_map_t pager_map; 182 static int bswneeded; 183 static vm_offset_t swapbkva; /* swap buffers kva */ 184 static TAILQ_HEAD(swqueue, buf) bswlist; 185 186 void 187 vm_pager_init(void) 188 { 189 struct pagerops **pgops; 190 191 /* 192 * Initialize the swap buffer list. 193 */ 194 TAILQ_INIT(&bswlist); 195 196 /* 197 * Initialize known pagers 198 */ 199 for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++) 200 if (pgops && ((*pgops)->pgo_init != NULL)) 201 (*(*pgops)->pgo_init) (); 202 } 203 204 void 205 vm_pager_bufferinit(void) 206 { 207 struct buf *bp; 208 int i; 209 210 bp = swbuf; 211 /* 212 * Now set up swap and physical I/O buffer headers. 213 */ 214 for (i = 0; i < nswbuf; i++, bp++) { 215 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist); 216 BUF_LOCKINIT(bp); 217 LIST_INIT(&bp->b_dep); 218 bp->b_xflags = 0; 219 } 220 221 cluster_pbuf_freecnt = nswbuf / 2; 222 223 swapbkva = kmem_alloc_pageable(pager_map, nswbuf * MAXPHYS); 224 if (!swapbkva) 225 panic("Not enough pager_map VM space for physical buffers"); 226 } 227 228 /* 229 * Allocate an instance of a pager of the given type. 230 * Size, protection and offset parameters are passed in for pagers that 231 * need to perform page-level validation (e.g. the device pager). 232 */ 233 vm_object_t 234 vm_pager_allocate(objtype_t type, void *handle, vm_ooffset_t size, vm_prot_t prot, 235 vm_ooffset_t off) 236 { 237 struct pagerops *ops; 238 239 ops = pagertab[type]; 240 if (ops) 241 return ((*ops->pgo_alloc) (handle, size, prot, off)); 242 return (NULL); 243 } 244 245 void 246 vm_pager_deallocate(vm_object_t object) 247 { 248 (*pagertab[object->type]->pgo_dealloc) (object); 249 } 250 251 /* 252 * vm_pager_strategy: 253 * 254 * called with no specific spl 255 * Execute strategy routine directly to pager. 256 */ 257 258 void 259 vm_pager_strategy(vm_object_t object, struct bio *bio) 260 { 261 struct buf *bp; 262 263 if (pagertab[object->type]->pgo_strategy) { 264 (*pagertab[object->type]->pgo_strategy)(object, bio); 265 } else { 266 bp = bio->bio_buf; 267 bp->b_flags |= B_ERROR; 268 bp->b_error = ENXIO; 269 biodone(bio); 270 } 271 } 272 273 /* 274 * vm_pager_get_pages() - inline, see vm/vm_pager.h 275 * vm_pager_put_pages() - inline, see vm/vm_pager.h 276 * vm_pager_has_page() - inline, see vm/vm_pager.h 277 * vm_pager_page_inserted() - inline, see vm/vm_pager.h 278 * vm_pager_page_removed() - inline, see vm/vm_pager.h 279 */ 280 281 #if 0 282 /* 283 * vm_pager_sync: 284 * 285 * Called by pageout daemon before going back to sleep. 286 * Gives pagers a chance to clean up any completed async pageing 287 * operations. 288 */ 289 void 290 vm_pager_sync(void) 291 { 292 struct pagerops **pgops; 293 294 for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++) 295 if (pgops && ((*pgops)->pgo_sync != NULL)) 296 (*(*pgops)->pgo_sync) (); 297 } 298 299 #endif 300 301 vm_object_t 302 vm_pager_object_lookup(struct pagerlst *pg_list, void *handle) 303 { 304 vm_object_t object; 305 306 for (object = TAILQ_FIRST(pg_list); object != NULL; object = TAILQ_NEXT(object,pager_object_list)) 307 if (object->handle == handle) 308 return (object); 309 return (NULL); 310 } 311 312 /* 313 * initialize a physical buffer 314 */ 315 316 static void 317 initpbuf(struct buf *bp) 318 { 319 bp->b_qindex = 0; /* BQUEUE_NONE */ 320 bp->b_data = (caddr_t) (MAXPHYS * (bp - swbuf)) + swapbkva; 321 bp->b_kvabase = bp->b_data; 322 bp->b_kvasize = MAXPHYS; 323 bp->b_xflags = 0; 324 bp->b_flags = 0; 325 bp->b_error = 0; 326 initbufbio(bp); 327 xio_init(&bp->b_xio); 328 BUF_LOCK(bp, LK_EXCLUSIVE); 329 } 330 331 /* 332 * allocate a physical buffer 333 * 334 * There are a limited number (nswbuf) of physical buffers. We need 335 * to make sure that no single subsystem is able to hog all of them, 336 * so each subsystem implements a counter which is typically initialized 337 * to 1/2 nswbuf. getpbuf() decrements this counter in allocation and 338 * increments it on release, and blocks if the counter hits zero. A 339 * subsystem may initialize the counter to -1 to disable the feature, 340 * but it must still be sure to match up all uses of getpbuf() with 341 * relpbuf() using the same variable. 342 * 343 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed 344 * relatively soon when the rest of the subsystems get smart about it. XXX 345 */ 346 struct buf * 347 getpbuf(int *pfreecnt) 348 { 349 struct buf *bp; 350 351 crit_enter(); 352 353 for (;;) { 354 if (pfreecnt) { 355 while (*pfreecnt == 0) { 356 tsleep(pfreecnt, 0, "wswbuf0", 0); 357 } 358 } 359 360 /* get a bp from the swap buffer header pool */ 361 if ((bp = TAILQ_FIRST(&bswlist)) != NULL) 362 break; 363 364 bswneeded = 1; 365 tsleep(&bswneeded, 0, "wswbuf1", 0); 366 /* loop in case someone else grabbed one */ 367 } 368 TAILQ_REMOVE(&bswlist, bp, b_freelist); 369 if (pfreecnt) 370 --*pfreecnt; 371 crit_exit(); 372 373 initpbuf(bp); 374 return bp; 375 } 376 377 /* 378 * allocate a physical buffer, if one is available. 379 * 380 * Note that there is no NULL hack here - all subsystems using this 381 * call understand how to use pfreecnt. 382 */ 383 struct buf * 384 trypbuf(int *pfreecnt) 385 { 386 struct buf *bp; 387 388 crit_enter(); 389 if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist)) == NULL) { 390 crit_exit(); 391 return NULL; 392 } 393 TAILQ_REMOVE(&bswlist, bp, b_freelist); 394 395 --*pfreecnt; 396 397 crit_exit(); 398 399 initpbuf(bp); 400 401 return bp; 402 } 403 404 /* 405 * release a physical buffer 406 * 407 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed 408 * relatively soon when the rest of the subsystems get smart about it. XXX 409 */ 410 void 411 relpbuf(struct buf *bp, int *pfreecnt) 412 { 413 crit_enter(); 414 415 if (bp->b_vp) 416 pbrelvp(bp); 417 418 BUF_UNLOCK(bp); 419 420 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist); 421 422 if (bswneeded) { 423 bswneeded = 0; 424 wakeup(&bswneeded); 425 } 426 if (pfreecnt) { 427 if (++*pfreecnt == 1) 428 wakeup(pfreecnt); 429 } 430 crit_exit(); 431 } 432 433