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 * $FreeBSD$ 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/ucred.h> 79 #include <sys/malloc.h> 80 81 #include <vm/vm.h> 82 #include <vm/vm_param.h> 83 #include <vm/vm_object.h> 84 #include <vm/vm_page.h> 85 #include <vm/vm_pager.h> 86 #include <vm/vm_extern.h> 87 88 MALLOC_DEFINE(M_VMPGDATA, "VM pgdata", "XXX: VM pager private data"); 89 90 extern struct pagerops defaultpagerops; 91 extern struct pagerops swappagerops; 92 extern struct pagerops vnodepagerops; 93 extern struct pagerops devicepagerops; 94 extern struct pagerops physpagerops; 95 96 int cluster_pbuf_freecnt = -1; /* unlimited to begin with */ 97 98 static int dead_pager_getpages(vm_object_t, vm_page_t *, int, int); 99 static vm_object_t dead_pager_alloc(void *, vm_ooffset_t, vm_prot_t, 100 vm_ooffset_t); 101 static void dead_pager_putpages(vm_object_t, vm_page_t *, int, int, int *); 102 static boolean_t dead_pager_haspage(vm_object_t, vm_pindex_t, int *, int *); 103 static void dead_pager_dealloc(vm_object_t); 104 105 static int 106 dead_pager_getpages(obj, ma, count, req) 107 vm_object_t obj; 108 vm_page_t *ma; 109 int count; 110 int req; 111 { 112 return VM_PAGER_FAIL; 113 } 114 115 static vm_object_t 116 dead_pager_alloc(handle, size, prot, off) 117 void *handle; 118 vm_ooffset_t size; 119 vm_prot_t prot; 120 vm_ooffset_t off; 121 { 122 return NULL; 123 } 124 125 static void 126 dead_pager_putpages(object, m, count, flags, rtvals) 127 vm_object_t object; 128 vm_page_t *m; 129 int count; 130 int flags; 131 int *rtvals; 132 { 133 int i; 134 135 for (i = 0; i < count; i++) { 136 rtvals[i] = VM_PAGER_AGAIN; 137 } 138 } 139 140 static int 141 dead_pager_haspage(object, pindex, prev, next) 142 vm_object_t object; 143 vm_pindex_t pindex; 144 int *prev; 145 int *next; 146 { 147 if (prev) 148 *prev = 0; 149 if (next) 150 *next = 0; 151 return FALSE; 152 } 153 154 static void 155 dead_pager_dealloc(object) 156 vm_object_t object; 157 { 158 return; 159 } 160 161 static struct pagerops deadpagerops = { 162 NULL, 163 dead_pager_alloc, 164 dead_pager_dealloc, 165 dead_pager_getpages, 166 dead_pager_putpages, 167 dead_pager_haspage, 168 NULL 169 }; 170 171 struct pagerops *pagertab[] = { 172 &defaultpagerops, /* OBJT_DEFAULT */ 173 &swappagerops, /* OBJT_SWAP */ 174 &vnodepagerops, /* OBJT_VNODE */ 175 &devicepagerops, /* OBJT_DEVICE */ 176 &physpagerops, /* OBJT_PHYS */ 177 &deadpagerops /* OBJT_DEAD */ 178 }; 179 180 int npagers = sizeof(pagertab) / sizeof(pagertab[0]); 181 182 /* 183 * Kernel address space for mapping pages. 184 * Used by pagers where KVAs are needed for IO. 185 * 186 * XXX needs to be large enough to support the number of pending async 187 * cleaning requests (NPENDINGIO == 64) * the maximum swap cluster size 188 * (MAXPHYS == 64k) if you want to get the most efficiency. 189 */ 190 #define PAGER_MAP_SIZE (8 * 1024 * 1024) 191 192 int pager_map_size = PAGER_MAP_SIZE; 193 vm_map_t pager_map; 194 static int bswneeded; 195 static vm_offset_t swapbkva; /* swap buffers kva */ 196 struct mtx pbuf_mtx; 197 static TAILQ_HEAD(swqueue, buf) bswlist; 198 199 void 200 vm_pager_init() 201 { 202 struct pagerops **pgops; 203 204 TAILQ_INIT(&bswlist); 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 } 212 213 void 214 vm_pager_bufferinit() 215 { 216 struct buf *bp; 217 int i; 218 219 mtx_init(&pbuf_mtx, "pbuf mutex", MTX_DEF); 220 bp = swbuf; 221 /* 222 * Now set up swap and physical I/O buffer headers. 223 */ 224 for (i = 0; i < nswbuf; i++, bp++) { 225 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist); 226 BUF_LOCKINIT(bp); 227 LIST_INIT(&bp->b_dep); 228 bp->b_rcred = bp->b_wcred = NOCRED; 229 bp->b_xflags = 0; 230 } 231 232 cluster_pbuf_freecnt = nswbuf / 2; 233 234 swapbkva = kmem_alloc_pageable(pager_map, nswbuf * MAXPHYS); 235 if (!swapbkva) 236 panic("Not enough pager_map VM space for physical buffers"); 237 } 238 239 /* 240 * Allocate an instance of a pager of the given type. 241 * Size, protection and offset parameters are passed in for pagers that 242 * need to perform page-level validation (e.g. the device pager). 243 */ 244 vm_object_t 245 vm_pager_allocate(objtype_t type, void *handle, vm_ooffset_t size, 246 vm_prot_t prot, vm_ooffset_t off) 247 { 248 vm_object_t ret; 249 struct pagerops *ops; 250 251 GIANT_REQUIRED; 252 253 ops = pagertab[type]; 254 if (ops) 255 ret = (*ops->pgo_alloc) (handle, size, prot, off); 256 else 257 ret = NULL; 258 return (ret); 259 } 260 261 void 262 vm_pager_deallocate(object) 263 vm_object_t object; 264 { 265 GIANT_REQUIRED; 266 (*pagertab[object->type]->pgo_dealloc) (object); 267 } 268 269 /* 270 * vm_pager_strategy: 271 * 272 * called with no specific spl 273 * Execute strategy routine directly to pager. 274 */ 275 void 276 vm_pager_strategy(vm_object_t object, struct bio *bp) 277 { 278 if (pagertab[object->type]->pgo_strategy) { 279 (*pagertab[object->type]->pgo_strategy)(object, bp); 280 } else { 281 bp->bio_flags |= BIO_ERROR; 282 bp->bio_error = ENXIO; 283 biodone(bp); 284 } 285 } 286 287 /* 288 * vm_pager_get_pages() - inline, see vm/vm_pager.h 289 * vm_pager_put_pages() - inline, see vm/vm_pager.h 290 * vm_pager_has_page() - inline, see vm/vm_pager.h 291 * vm_pager_page_inserted() - inline, see vm/vm_pager.h 292 * vm_pager_page_removed() - inline, see vm/vm_pager.h 293 */ 294 295 #if 0 296 /* 297 * vm_pager_sync: 298 * 299 * Called by pageout daemon before going back to sleep. 300 * Gives pagers a chance to clean up any completed async pageing 301 * operations. 302 */ 303 void 304 vm_pager_sync() 305 { 306 struct pagerops **pgops; 307 308 for (pgops = pagertab; pgops < &pagertab[npagers]; pgops++) 309 if (pgops && ((*pgops)->pgo_sync != NULL)) 310 (*(*pgops)->pgo_sync) (); 311 } 312 #endif 313 314 vm_offset_t 315 vm_pager_map_page(m) 316 vm_page_t m; 317 { 318 vm_offset_t kva; 319 320 kva = kmem_alloc_wait(pager_map, PAGE_SIZE); 321 pmap_qenter(kva, &m, 1); 322 return (kva); 323 } 324 325 void 326 vm_pager_unmap_page(kva) 327 vm_offset_t kva; 328 { 329 330 pmap_qremove(kva, 1); 331 kmem_free_wakeup(pager_map, kva, PAGE_SIZE); 332 } 333 334 vm_object_t 335 vm_pager_object_lookup(pg_list, handle) 336 struct pagerlst *pg_list; 337 void *handle; 338 { 339 vm_object_t object; 340 341 TAILQ_FOREACH(object, pg_list, pager_object_list) 342 if (object->handle == handle) 343 return (object); 344 return (NULL); 345 } 346 347 /* 348 * initialize a physical buffer 349 */ 350 351 /* 352 * XXX This probably belongs in vfs_bio.c 353 */ 354 static void 355 initpbuf(struct buf *bp) 356 { 357 bp->b_rcred = NOCRED; 358 bp->b_wcred = NOCRED; 359 bp->b_qindex = 0; /* On no queue (QUEUE_NONE) */ 360 bp->b_data = (caddr_t) (MAXPHYS * (bp - swbuf)) + swapbkva; 361 bp->b_kvabase = bp->b_data; 362 bp->b_kvasize = MAXPHYS; 363 bp->b_xflags = 0; 364 bp->b_flags = 0; 365 bp->b_ioflags = 0; 366 bp->b_iodone = NULL; 367 bp->b_error = 0; 368 bp->b_magic = B_MAGIC_BIO; 369 bp->b_op = &buf_ops_bio; 370 BUF_LOCK(bp, LK_EXCLUSIVE); 371 } 372 373 /* 374 * allocate a physical buffer 375 * 376 * There are a limited number (nswbuf) of physical buffers. We need 377 * to make sure that no single subsystem is able to hog all of them, 378 * so each subsystem implements a counter which is typically initialized 379 * to 1/2 nswbuf. getpbuf() decrements this counter in allocation and 380 * increments it on release, and blocks if the counter hits zero. A 381 * subsystem may initialize the counter to -1 to disable the feature, 382 * but it must still be sure to match up all uses of getpbuf() with 383 * relpbuf() using the same variable. 384 * 385 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed 386 * relatively soon when the rest of the subsystems get smart about it. XXX 387 */ 388 struct buf * 389 getpbuf(pfreecnt) 390 int *pfreecnt; 391 { 392 int s; 393 struct buf *bp; 394 395 s = splvm(); 396 GIANT_REQUIRED; 397 mtx_lock(&pbuf_mtx); 398 399 for (;;) { 400 if (pfreecnt) { 401 while (*pfreecnt == 0) { 402 msleep(pfreecnt, &pbuf_mtx, PVM, "wswbuf0", 0); 403 } 404 } 405 406 /* get a bp from the swap buffer header pool */ 407 if ((bp = TAILQ_FIRST(&bswlist)) != NULL) 408 break; 409 410 bswneeded = 1; 411 msleep(&bswneeded, &pbuf_mtx, PVM, "wswbuf1", 0); 412 /* loop in case someone else grabbed one */ 413 } 414 TAILQ_REMOVE(&bswlist, bp, b_freelist); 415 if (pfreecnt) 416 --*pfreecnt; 417 mtx_unlock(&pbuf_mtx); 418 splx(s); 419 420 initpbuf(bp); 421 return bp; 422 } 423 424 /* 425 * allocate a physical buffer, if one is available. 426 * 427 * Note that there is no NULL hack here - all subsystems using this 428 * call understand how to use pfreecnt. 429 */ 430 struct buf * 431 trypbuf(pfreecnt) 432 int *pfreecnt; 433 { 434 int s; 435 struct buf *bp; 436 437 s = splvm(); 438 mtx_lock(&pbuf_mtx); 439 if (*pfreecnt == 0 || (bp = TAILQ_FIRST(&bswlist)) == NULL) { 440 mtx_unlock(&pbuf_mtx); 441 splx(s); 442 return NULL; 443 } 444 TAILQ_REMOVE(&bswlist, bp, b_freelist); 445 446 --*pfreecnt; 447 448 mtx_unlock(&pbuf_mtx); 449 splx(s); 450 451 initpbuf(bp); 452 453 return bp; 454 } 455 456 /* 457 * release a physical buffer 458 * 459 * NOTE: pfreecnt can be NULL, but this 'feature' will be removed 460 * relatively soon when the rest of the subsystems get smart about it. XXX 461 */ 462 void 463 relpbuf(bp, pfreecnt) 464 struct buf *bp; 465 int *pfreecnt; 466 { 467 int s; 468 469 s = splvm(); 470 mtx_lock(&pbuf_mtx); 471 472 if (bp->b_rcred != NOCRED) { 473 crfree(bp->b_rcred); 474 bp->b_rcred = NOCRED; 475 } 476 if (bp->b_wcred != NOCRED) { 477 crfree(bp->b_wcred); 478 bp->b_wcred = NOCRED; 479 } 480 481 if (bp->b_vp) 482 pbrelvp(bp); 483 484 BUF_UNLOCK(bp); 485 486 TAILQ_INSERT_HEAD(&bswlist, bp, b_freelist); 487 488 if (bswneeded) { 489 bswneeded = 0; 490 wakeup(&bswneeded); 491 } 492 if (pfreecnt) { 493 if (++*pfreecnt == 1) 494 wakeup(pfreecnt); 495 } 496 mtx_unlock(&pbuf_mtx); 497 splx(s); 498 } 499