fss.c revision 1.44
1 /* $NetBSD: fss.c,v 1.44 2008/03/21 21:54:59 ad Exp $ */ 2 3 /*- 4 * Copyright (c) 2003 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Juergen Hannken-Illjes. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39 /* 40 * File system snapshot disk driver. 41 * 42 * Block/character interface to the snapshot of a mounted file system. 43 */ 44 45 #include <sys/cdefs.h> 46 __KERNEL_RCSID(0, "$NetBSD: fss.c,v 1.44 2008/03/21 21:54:59 ad Exp $"); 47 48 #include "fss.h" 49 50 #include <sys/param.h> 51 #include <sys/systm.h> 52 #include <sys/namei.h> 53 #include <sys/proc.h> 54 #include <sys/errno.h> 55 #include <sys/buf.h> 56 #include <sys/malloc.h> 57 #include <sys/ioctl.h> 58 #include <sys/disklabel.h> 59 #include <sys/device.h> 60 #include <sys/disk.h> 61 #include <sys/stat.h> 62 #include <sys/mount.h> 63 #include <sys/vnode.h> 64 #include <sys/file.h> 65 #include <sys/uio.h> 66 #include <sys/conf.h> 67 #include <sys/kthread.h> 68 #include <sys/fstrans.h> 69 #include <sys/simplelock.h> 70 71 #include <miscfs/specfs/specdev.h> 72 73 #include <dev/fssvar.h> 74 75 #include <machine/stdarg.h> 76 77 #ifdef DEBUG 78 #define FSS_STATISTICS 79 #endif 80 81 #ifdef FSS_STATISTICS 82 struct fss_stat { 83 u_int64_t cow_calls; 84 u_int64_t cow_copied; 85 u_int64_t cow_cache_full; 86 u_int64_t indir_read; 87 u_int64_t indir_write; 88 }; 89 90 static struct fss_stat fss_stat[NFSS]; 91 92 #define FSS_STAT_INC(sc, field) \ 93 do { \ 94 fss_stat[sc->sc_unit].field++; \ 95 } while (0) 96 #define FSS_STAT_SET(sc, field, value) \ 97 do { \ 98 fss_stat[sc->sc_unit].field = value; \ 99 } while (0) 100 #define FSS_STAT_ADD(sc, field, value) \ 101 do { \ 102 fss_stat[sc->sc_unit].field += value; \ 103 } while (0) 104 #define FSS_STAT_VAL(sc, field) fss_stat[sc->sc_unit].field 105 #define FSS_STAT_CLEAR(sc) \ 106 do { \ 107 memset(&fss_stat[sc->sc_unit], 0, \ 108 sizeof(struct fss_stat)); \ 109 } while (0) 110 #else /* FSS_STATISTICS */ 111 #define FSS_STAT_INC(sc, field) 112 #define FSS_STAT_SET(sc, field, value) 113 #define FSS_STAT_ADD(sc, field, value) 114 #define FSS_STAT_CLEAR(sc) 115 #endif /* FSS_STATISTICS */ 116 117 static struct fss_softc fss_softc[NFSS]; 118 119 void fssattach(int); 120 121 dev_type_open(fss_open); 122 dev_type_close(fss_close); 123 dev_type_read(fss_read); 124 dev_type_write(fss_write); 125 dev_type_ioctl(fss_ioctl); 126 dev_type_strategy(fss_strategy); 127 dev_type_dump(fss_dump); 128 dev_type_size(fss_size); 129 130 static int fss_copy_on_write(void *, struct buf *, bool); 131 static inline void fss_error(struct fss_softc *, const char *, ...); 132 static int fss_create_files(struct fss_softc *, struct fss_set *, 133 off_t *, struct lwp *); 134 static int fss_create_snapshot(struct fss_softc *, struct fss_set *, 135 struct lwp *); 136 static int fss_delete_snapshot(struct fss_softc *, struct lwp *); 137 static int fss_softc_alloc(struct fss_softc *); 138 static void fss_softc_free(struct fss_softc *); 139 static void fss_cluster_iodone(struct buf *); 140 static void fss_read_cluster(struct fss_softc *, u_int32_t); 141 static void fss_bs_thread(void *); 142 static int fss_bs_io(struct fss_softc *, fss_io_type, 143 u_int32_t, off_t, int, void *); 144 static u_int32_t *fss_bs_indir(struct fss_softc *, u_int32_t); 145 146 const struct bdevsw fss_bdevsw = { 147 fss_open, fss_close, fss_strategy, fss_ioctl, 148 fss_dump, fss_size, D_DISK 149 }; 150 151 const struct cdevsw fss_cdevsw = { 152 fss_open, fss_close, fss_read, fss_write, fss_ioctl, 153 nostop, notty, nopoll, nommap, nokqfilter, D_DISK 154 }; 155 156 void 157 fssattach(int num) 158 { 159 int i; 160 struct fss_softc *sc; 161 162 for (i = 0; i < NFSS; i++) { 163 sc = &fss_softc[i]; 164 sc->sc_unit = i; 165 sc->sc_bdev = NODEV; 166 simple_lock_init(&sc->sc_slock); 167 mutex_init(&sc->sc_lock, MUTEX_DEFAULT, IPL_NONE); 168 bufq_alloc(&sc->sc_bufq, "fcfs", 0); 169 } 170 } 171 172 int 173 fss_open(dev_t dev, int flags, int mode, struct lwp *l) 174 { 175 int s, mflag; 176 struct fss_softc *sc; 177 178 mflag = (mode == S_IFCHR ? FSS_CDEV_OPEN : FSS_BDEV_OPEN); 179 180 if ((sc = FSS_DEV_TO_SOFTC(dev)) == NULL) 181 return ENODEV; 182 183 FSS_LOCK(sc, s); 184 185 sc->sc_flags |= mflag; 186 187 FSS_UNLOCK(sc, s); 188 189 return 0; 190 } 191 192 int 193 fss_close(dev_t dev, int flags, int mode, struct lwp *l) 194 { 195 int s, mflag, error; 196 struct fss_softc *sc; 197 198 mflag = (mode == S_IFCHR ? FSS_CDEV_OPEN : FSS_BDEV_OPEN); 199 200 if ((sc = FSS_DEV_TO_SOFTC(dev)) == NULL) 201 return ENODEV; 202 203 FSS_LOCK(sc, s); 204 205 if ((sc->sc_flags & (FSS_CDEV_OPEN|FSS_BDEV_OPEN)) == mflag) { 206 if ((sc->sc_uflags & FSS_UNCONFIG_ON_CLOSE) != 0 && 207 (sc->sc_flags & FSS_ACTIVE) != 0) { 208 FSS_UNLOCK(sc, s); 209 error = fss_ioctl(dev, FSSIOCCLR, NULL, FWRITE, l); 210 if (error) 211 return error; 212 FSS_LOCK(sc, s); 213 } 214 sc->sc_uflags &= ~FSS_UNCONFIG_ON_CLOSE; 215 } 216 217 sc->sc_flags &= ~mflag; 218 219 FSS_UNLOCK(sc, s); 220 221 return 0; 222 } 223 224 void 225 fss_strategy(struct buf *bp) 226 { 227 int s; 228 struct fss_softc *sc; 229 230 sc = FSS_DEV_TO_SOFTC(bp->b_dev); 231 232 FSS_LOCK(sc, s); 233 234 if ((bp->b_flags & B_READ) != B_READ || 235 sc == NULL || !FSS_ISVALID(sc)) { 236 237 FSS_UNLOCK(sc, s); 238 239 bp->b_error = (sc == NULL ? ENODEV : EROFS); 240 bp->b_resid = bp->b_bcount; 241 biodone(bp); 242 return; 243 } 244 245 bp->b_rawblkno = bp->b_blkno; 246 BUFQ_PUT(sc->sc_bufq, bp); 247 wakeup(&sc->sc_bs_lwp); 248 249 FSS_UNLOCK(sc, s); 250 } 251 252 int 253 fss_read(dev_t dev, struct uio *uio, int flags) 254 { 255 return physio(fss_strategy, NULL, dev, B_READ, minphys, uio); 256 } 257 258 int 259 fss_write(dev_t dev, struct uio *uio, int flags) 260 { 261 return physio(fss_strategy, NULL, dev, B_WRITE, minphys, uio); 262 } 263 264 int 265 fss_ioctl(dev_t dev, u_long cmd, void *data, int flag, struct lwp *l) 266 { 267 int error; 268 struct fss_softc *sc; 269 struct fss_set *fss = (struct fss_set *)data; 270 struct fss_get *fsg = (struct fss_get *)data; 271 272 if ((sc = FSS_DEV_TO_SOFTC(dev)) == NULL) 273 return ENODEV; 274 275 switch (cmd) { 276 case FSSIOCSET: 277 mutex_enter(&sc->sc_lock); 278 if ((flag & FWRITE) == 0) 279 error = EPERM; 280 else if ((sc->sc_flags & FSS_ACTIVE) != 0) 281 error = EBUSY; 282 else 283 error = fss_create_snapshot(sc, fss, l); 284 mutex_exit(&sc->sc_lock); 285 break; 286 287 case FSSIOCCLR: 288 mutex_enter(&sc->sc_lock); 289 if ((flag & FWRITE) == 0) 290 error = EPERM; 291 else if ((sc->sc_flags & FSS_ACTIVE) == 0) 292 error = ENXIO; 293 else 294 error = fss_delete_snapshot(sc, l); 295 mutex_exit(&sc->sc_lock); 296 break; 297 298 case FSSIOCGET: 299 mutex_enter(&sc->sc_lock); 300 switch (sc->sc_flags & (FSS_PERSISTENT | FSS_ACTIVE)) { 301 case FSS_ACTIVE: 302 memcpy(fsg->fsg_mount, sc->sc_mntname, MNAMELEN); 303 fsg->fsg_csize = FSS_CLSIZE(sc); 304 fsg->fsg_time = sc->sc_time; 305 fsg->fsg_mount_size = sc->sc_clcount; 306 fsg->fsg_bs_size = sc->sc_clnext; 307 error = 0; 308 break; 309 case FSS_PERSISTENT | FSS_ACTIVE: 310 memcpy(fsg->fsg_mount, sc->sc_mntname, MNAMELEN); 311 fsg->fsg_csize = 0; 312 fsg->fsg_time = sc->sc_time; 313 fsg->fsg_mount_size = 0; 314 fsg->fsg_bs_size = 0; 315 error = 0; 316 break; 317 default: 318 error = ENXIO; 319 break; 320 } 321 mutex_exit(&sc->sc_lock); 322 break; 323 324 case FSSIOFSET: 325 sc->sc_uflags = *(int *)data; 326 error = 0; 327 break; 328 329 case FSSIOFGET: 330 *(int *)data = sc->sc_uflags; 331 error = 0; 332 break; 333 334 default: 335 error = EINVAL; 336 break; 337 } 338 339 return error; 340 } 341 342 int 343 fss_size(dev_t dev) 344 { 345 return -1; 346 } 347 348 int 349 fss_dump(dev_t dev, daddr_t blkno, void *va, 350 size_t size) 351 { 352 return EROFS; 353 } 354 355 /* 356 * An error occurred reading or writing the snapshot or backing store. 357 * If it is the first error log to console. 358 * The caller holds the simplelock. 359 */ 360 static inline void 361 fss_error(struct fss_softc *sc, const char *fmt, ...) 362 { 363 va_list ap; 364 365 if ((sc->sc_flags & (FSS_ACTIVE|FSS_ERROR)) == FSS_ACTIVE) { 366 va_start(ap, fmt); 367 printf("fss%d: snapshot invalid: ", sc->sc_unit); 368 vprintf(fmt, ap); 369 printf("\n"); 370 va_end(ap); 371 } 372 if ((sc->sc_flags & FSS_ACTIVE) == FSS_ACTIVE) 373 sc->sc_flags |= FSS_ERROR; 374 } 375 376 /* 377 * Allocate the variable sized parts of the softc and 378 * fork the kernel thread. 379 * 380 * The fields sc_clcount, sc_clshift, sc_cache_size and sc_indir_size 381 * must be initialized. 382 */ 383 static int 384 fss_softc_alloc(struct fss_softc *sc) 385 { 386 int i, len, error; 387 388 len = (sc->sc_clcount+NBBY-1)/NBBY; 389 sc->sc_copied = malloc(len, M_TEMP, M_ZERO|M_WAITOK|M_CANFAIL); 390 if (sc->sc_copied == NULL) 391 return(ENOMEM); 392 393 len = sc->sc_cache_size*sizeof(struct fss_cache); 394 sc->sc_cache = malloc(len, M_TEMP, M_ZERO|M_WAITOK|M_CANFAIL); 395 if (sc->sc_cache == NULL) 396 return(ENOMEM); 397 398 len = FSS_CLSIZE(sc); 399 for (i = 0; i < sc->sc_cache_size; i++) { 400 sc->sc_cache[i].fc_type = FSS_CACHE_FREE; 401 sc->sc_cache[i].fc_softc = sc; 402 sc->sc_cache[i].fc_xfercount = 0; 403 sc->sc_cache[i].fc_data = malloc(len, M_TEMP, 404 M_WAITOK|M_CANFAIL); 405 if (sc->sc_cache[i].fc_data == NULL) 406 return(ENOMEM); 407 } 408 409 len = (sc->sc_indir_size+NBBY-1)/NBBY; 410 sc->sc_indir_valid = malloc(len, M_TEMP, M_ZERO|M_WAITOK|M_CANFAIL); 411 if (sc->sc_indir_valid == NULL) 412 return(ENOMEM); 413 414 len = FSS_CLSIZE(sc); 415 sc->sc_indir_data = malloc(len, M_TEMP, M_ZERO|M_WAITOK|M_CANFAIL); 416 if (sc->sc_indir_data == NULL) 417 return(ENOMEM); 418 419 if ((error = kthread_create(PRI_BIO, 0, NULL, fss_bs_thread, sc, 420 &sc->sc_bs_lwp, "fssbs%d", sc->sc_unit)) != 0) 421 return error; 422 423 sc->sc_flags |= FSS_BS_THREAD; 424 return 0; 425 } 426 427 /* 428 * Free the variable sized parts of the softc. 429 */ 430 static void 431 fss_softc_free(struct fss_softc *sc) 432 { 433 int s, i; 434 435 if ((sc->sc_flags & FSS_BS_THREAD) != 0) { 436 FSS_LOCK(sc, s); 437 sc->sc_flags &= ~FSS_BS_THREAD; 438 wakeup(&sc->sc_bs_lwp); 439 while (sc->sc_bs_lwp != NULL) 440 ltsleep(&sc->sc_bs_lwp, PRIBIO, "fssthread", 0, 441 &sc->sc_slock); 442 FSS_UNLOCK(sc, s); 443 } 444 445 if (sc->sc_copied != NULL) 446 free(sc->sc_copied, M_TEMP); 447 sc->sc_copied = NULL; 448 449 if (sc->sc_cache != NULL) { 450 for (i = 0; i < sc->sc_cache_size; i++) 451 if (sc->sc_cache[i].fc_data != NULL) 452 free(sc->sc_cache[i].fc_data, M_TEMP); 453 free(sc->sc_cache, M_TEMP); 454 } 455 sc->sc_cache = NULL; 456 457 if (sc->sc_indir_valid != NULL) 458 free(sc->sc_indir_valid, M_TEMP); 459 sc->sc_indir_valid = NULL; 460 461 if (sc->sc_indir_data != NULL) 462 free(sc->sc_indir_data, M_TEMP); 463 sc->sc_indir_data = NULL; 464 } 465 466 /* 467 * Check if an unmount is ok. If forced, set this snapshot into ERROR state. 468 */ 469 int 470 fss_umount_hook(struct mount *mp, int forced) 471 { 472 int i, s; 473 474 for (i = 0; i < NFSS; i++) { 475 FSS_LOCK(&fss_softc[i], s); 476 if ((fss_softc[i].sc_flags & FSS_ACTIVE) != 0 && 477 fss_softc[i].sc_mount == mp) { 478 if (forced) 479 fss_error(&fss_softc[i], "forced unmount"); 480 else { 481 FSS_UNLOCK(&fss_softc[i], s); 482 return EBUSY; 483 } 484 } 485 FSS_UNLOCK(&fss_softc[i], s); 486 } 487 488 return 0; 489 } 490 491 /* 492 * A buffer is written to the snapshotted block device. Copy to 493 * backing store if needed. 494 */ 495 static int 496 fss_copy_on_write(void *v, struct buf *bp, bool data_valid) 497 { 498 int s; 499 u_int32_t cl, ch, c; 500 struct fss_softc *sc = v; 501 502 FSS_LOCK(sc, s); 503 if (!FSS_ISVALID(sc)) { 504 FSS_UNLOCK(sc, s); 505 return 0; 506 } 507 508 FSS_UNLOCK(sc, s); 509 510 FSS_STAT_INC(sc, cow_calls); 511 512 cl = FSS_BTOCL(sc, dbtob(bp->b_blkno)); 513 ch = FSS_BTOCL(sc, dbtob(bp->b_blkno)+bp->b_bcount-1); 514 515 for (c = cl; c <= ch; c++) 516 fss_read_cluster(sc, c); 517 518 return 0; 519 } 520 521 /* 522 * Lookup and open needed files. 523 * 524 * For file system internal snapshot initializes sc_mntname, sc_mount, 525 * sc_bs_vp and sc_time. 526 * 527 * Otherwise returns dev and size of the underlying block device. 528 * Initializes sc_mntname, sc_mount, sc_bdev, sc_bs_vp and sc_mount 529 */ 530 static int 531 fss_create_files(struct fss_softc *sc, struct fss_set *fss, 532 off_t *bsize, struct lwp *l) 533 { 534 int error, bits, fsbsize; 535 struct timespec ts; 536 struct partinfo dpart; 537 struct vattr va; 538 struct nameidata nd; 539 540 /* 541 * Get the mounted file system. 542 */ 543 544 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, fss->fss_mount); 545 if ((error = namei(&nd)) != 0) 546 return error; 547 548 if ((nd.ni_vp->v_vflag & VV_ROOT) != VV_ROOT) { 549 vrele(nd.ni_vp); 550 return EINVAL; 551 } 552 553 sc->sc_mount = nd.ni_vp->v_mount; 554 memcpy(sc->sc_mntname, sc->sc_mount->mnt_stat.f_mntonname, MNAMELEN); 555 556 vrele(nd.ni_vp); 557 558 /* 559 * Check for file system internal snapshot. 560 */ 561 562 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, fss->fss_bstore); 563 if ((error = namei(&nd)) != 0) 564 return error; 565 566 if (nd.ni_vp->v_type == VREG && nd.ni_vp->v_mount == sc->sc_mount) { 567 vrele(nd.ni_vp); 568 sc->sc_flags |= FSS_PERSISTENT; 569 570 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, fss->fss_bstore); 571 if ((error = vn_open(&nd, FREAD, 0)) != 0) 572 return error; 573 sc->sc_bs_vp = nd.ni_vp; 574 575 fsbsize = sc->sc_bs_vp->v_mount->mnt_stat.f_iosize; 576 bits = sizeof(sc->sc_bs_bshift)*NBBY; 577 for (sc->sc_bs_bshift = 1; sc->sc_bs_bshift < bits; 578 sc->sc_bs_bshift++) 579 if (FSS_FSBSIZE(sc) == fsbsize) 580 break; 581 if (sc->sc_bs_bshift >= bits) { 582 VOP_UNLOCK(sc->sc_bs_vp, 0); 583 return EINVAL; 584 } 585 586 sc->sc_bs_bmask = FSS_FSBSIZE(sc)-1; 587 sc->sc_clshift = 0; 588 589 error = VFS_SNAPSHOT(sc->sc_mount, sc->sc_bs_vp, &ts); 590 TIMESPEC_TO_TIMEVAL(&sc->sc_time, &ts); 591 592 VOP_UNLOCK(sc->sc_bs_vp, 0); 593 594 return error; 595 } 596 vrele(nd.ni_vp); 597 598 /* 599 * Get the block device it is mounted on. 600 */ 601 602 NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, 603 sc->sc_mount->mnt_stat.f_mntfromname); 604 if ((error = namei(&nd)) != 0) 605 return error; 606 607 if (nd.ni_vp->v_type != VBLK) { 608 vrele(nd.ni_vp); 609 return EINVAL; 610 } 611 612 error = VOP_IOCTL(nd.ni_vp, DIOCGPART, &dpart, FREAD, l->l_cred); 613 if (error) { 614 vrele(nd.ni_vp); 615 return error; 616 } 617 618 sc->sc_bdev = nd.ni_vp->v_rdev; 619 *bsize = (off_t)dpart.disklab->d_secsize*dpart.part->p_size; 620 vrele(nd.ni_vp); 621 622 /* 623 * Get the backing store 624 */ 625 626 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, fss->fss_bstore); 627 if ((error = vn_open(&nd, FREAD|FWRITE, 0)) != 0) 628 return error; 629 VOP_UNLOCK(nd.ni_vp, 0); 630 631 sc->sc_bs_vp = nd.ni_vp; 632 633 if (nd.ni_vp->v_type != VREG && nd.ni_vp->v_type != VCHR) 634 return EINVAL; 635 636 if (sc->sc_bs_vp->v_type == VREG) { 637 error = VOP_GETATTR(sc->sc_bs_vp, &va, l->l_cred); 638 if (error != 0) 639 return error; 640 sc->sc_bs_size = va.va_size; 641 fsbsize = sc->sc_bs_vp->v_mount->mnt_stat.f_iosize; 642 if (fsbsize & (fsbsize-1)) /* No power of two */ 643 return EINVAL; 644 for (sc->sc_bs_bshift = 1; sc->sc_bs_bshift < 32; 645 sc->sc_bs_bshift++) 646 if (FSS_FSBSIZE(sc) == fsbsize) 647 break; 648 if (sc->sc_bs_bshift >= 32) 649 return EINVAL; 650 sc->sc_bs_bmask = FSS_FSBSIZE(sc)-1; 651 } else { 652 sc->sc_bs_bshift = DEV_BSHIFT; 653 sc->sc_bs_bmask = FSS_FSBSIZE(sc)-1; 654 } 655 656 /* 657 * As all IO to from/to the backing store goes through 658 * VOP_STRATEGY() clean the buffer cache to prevent 659 * cache incoherencies. 660 */ 661 if ((error = vinvalbuf(sc->sc_bs_vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 662 return error; 663 664 return 0; 665 } 666 667 /* 668 * Create a snapshot. 669 */ 670 static int 671 fss_create_snapshot(struct fss_softc *sc, struct fss_set *fss, struct lwp *l) 672 { 673 int len, error; 674 u_int32_t csize; 675 off_t bsize; 676 677 bsize = 0; /* XXX gcc */ 678 679 /* 680 * Open needed files. 681 */ 682 if ((error = fss_create_files(sc, fss, &bsize, l)) != 0) 683 goto bad; 684 685 if (sc->sc_flags & FSS_PERSISTENT) { 686 fss_softc_alloc(sc); 687 sc->sc_flags |= FSS_ACTIVE; 688 return 0; 689 } 690 691 /* 692 * Set cluster size. Must be a power of two and 693 * a multiple of backing store block size. 694 */ 695 if (fss->fss_csize <= 0) 696 csize = MAXPHYS; 697 else 698 csize = fss->fss_csize; 699 if (bsize/csize > FSS_CLUSTER_MAX) 700 csize = bsize/FSS_CLUSTER_MAX+1; 701 702 for (sc->sc_clshift = sc->sc_bs_bshift; sc->sc_clshift < 32; 703 sc->sc_clshift++) 704 if (FSS_CLSIZE(sc) >= csize) 705 break; 706 if (sc->sc_clshift >= 32) { 707 error = EINVAL; 708 goto bad; 709 } 710 sc->sc_clmask = FSS_CLSIZE(sc)-1; 711 712 /* 713 * Set number of cache slots. 714 */ 715 if (FSS_CLSIZE(sc) <= 8192) 716 sc->sc_cache_size = 32; 717 else if (FSS_CLSIZE(sc) <= 65536) 718 sc->sc_cache_size = 8; 719 else 720 sc->sc_cache_size = 4; 721 722 /* 723 * Set number of clusters and size of last cluster. 724 */ 725 sc->sc_clcount = FSS_BTOCL(sc, bsize-1)+1; 726 sc->sc_clresid = FSS_CLOFF(sc, bsize-1)+1; 727 728 /* 729 * Set size of indirect table. 730 */ 731 len = sc->sc_clcount*sizeof(u_int32_t); 732 sc->sc_indir_size = FSS_BTOCL(sc, len)+1; 733 sc->sc_clnext = sc->sc_indir_size; 734 sc->sc_indir_cur = 0; 735 736 if ((error = fss_softc_alloc(sc)) != 0) 737 goto bad; 738 739 /* 740 * Activate the snapshot. 741 */ 742 743 if ((error = vfs_suspend(sc->sc_mount, 0)) != 0) 744 goto bad; 745 746 microtime(&sc->sc_time); 747 748 if (error == 0) 749 error = fscow_establish(sc->sc_mount, 750 fss_copy_on_write, sc); 751 if (error == 0) 752 sc->sc_flags |= FSS_ACTIVE; 753 754 vfs_resume(sc->sc_mount); 755 756 if (error != 0) 757 goto bad; 758 759 #ifdef DEBUG 760 printf("fss%d: %s snapshot active\n", sc->sc_unit, sc->sc_mntname); 761 printf("fss%d: %u clusters of %u, %u cache slots, %u indir clusters\n", 762 sc->sc_unit, sc->sc_clcount, FSS_CLSIZE(sc), 763 sc->sc_cache_size, sc->sc_indir_size); 764 #endif 765 766 return 0; 767 768 bad: 769 fss_softc_free(sc); 770 if (sc->sc_bs_vp != NULL) { 771 if (sc->sc_flags & FSS_PERSISTENT) 772 vn_close(sc->sc_bs_vp, FREAD, l->l_cred); 773 else 774 vn_close(sc->sc_bs_vp, FREAD|FWRITE, l->l_cred); 775 } 776 sc->sc_bs_vp = NULL; 777 778 return error; 779 } 780 781 /* 782 * Delete a snapshot. 783 */ 784 static int 785 fss_delete_snapshot(struct fss_softc *sc, struct lwp *l) 786 { 787 int s; 788 789 if ((sc->sc_flags & FSS_PERSISTENT) == 0) 790 fscow_disestablish(sc->sc_mount, fss_copy_on_write, sc); 791 792 FSS_LOCK(sc, s); 793 sc->sc_flags &= ~(FSS_ACTIVE|FSS_ERROR); 794 sc->sc_mount = NULL; 795 sc->sc_bdev = NODEV; 796 FSS_UNLOCK(sc, s); 797 798 fss_softc_free(sc); 799 if (sc->sc_flags & FSS_PERSISTENT) 800 vn_close(sc->sc_bs_vp, FREAD, l->l_cred); 801 else 802 vn_close(sc->sc_bs_vp, FREAD|FWRITE, l->l_cred); 803 sc->sc_bs_vp = NULL; 804 sc->sc_flags &= ~FSS_PERSISTENT; 805 806 FSS_STAT_CLEAR(sc); 807 808 return 0; 809 } 810 811 /* 812 * A read from the snapshotted block device has completed. 813 */ 814 static void 815 fss_cluster_iodone(struct buf *bp) 816 { 817 int s; 818 struct fss_cache *scp = bp->b_private; 819 820 KASSERT(bp->b_vp == NULL); 821 822 FSS_LOCK(scp->fc_softc, s); 823 824 if (bp->b_error != 0) 825 fss_error(scp->fc_softc, "fs read error %d", bp->b_error); 826 827 if (--scp->fc_xfercount == 0) 828 wakeup(&scp->fc_data); 829 830 FSS_UNLOCK(scp->fc_softc, s); 831 832 putiobuf(bp); 833 } 834 835 /* 836 * Read a cluster from the snapshotted block device to the cache. 837 */ 838 static void 839 fss_read_cluster(struct fss_softc *sc, u_int32_t cl) 840 { 841 int s, todo, len; 842 char *addr; 843 daddr_t dblk; 844 struct buf *bp; 845 struct fss_cache *scp, *scl; 846 847 /* 848 * Get a free cache slot. 849 */ 850 scl = sc->sc_cache+sc->sc_cache_size; 851 852 FSS_LOCK(sc, s); 853 854 restart: 855 if (isset(sc->sc_copied, cl) || !FSS_ISVALID(sc)) { 856 FSS_UNLOCK(sc, s); 857 return; 858 } 859 860 for (scp = sc->sc_cache; scp < scl; scp++) 861 if (scp->fc_type != FSS_CACHE_FREE && 862 scp->fc_cluster == cl) { 863 ltsleep(&scp->fc_type, PRIBIO, "cowwait2", 0, 864 &sc->sc_slock); 865 goto restart; 866 } 867 868 for (scp = sc->sc_cache; scp < scl; scp++) 869 if (scp->fc_type == FSS_CACHE_FREE) { 870 scp->fc_type = FSS_CACHE_BUSY; 871 scp->fc_cluster = cl; 872 break; 873 } 874 if (scp >= scl) { 875 FSS_STAT_INC(sc, cow_cache_full); 876 ltsleep(&sc->sc_cache, PRIBIO, "cowwait3", 0, &sc->sc_slock); 877 goto restart; 878 } 879 880 FSS_UNLOCK(sc, s); 881 882 /* 883 * Start the read. 884 */ 885 FSS_STAT_INC(sc, cow_copied); 886 887 dblk = btodb(FSS_CLTOB(sc, cl)); 888 addr = scp->fc_data; 889 if (cl == sc->sc_clcount-1) { 890 todo = sc->sc_clresid; 891 memset((char *)addr + todo, 0, FSS_CLSIZE(sc) - todo); 892 } else 893 todo = FSS_CLSIZE(sc); 894 while (todo > 0) { 895 len = todo; 896 if (len > MAXPHYS) 897 len = MAXPHYS; 898 899 bp = getiobuf(NULL, true); 900 bp->b_flags = B_READ; 901 bp->b_bcount = len; 902 bp->b_bufsize = bp->b_bcount; 903 bp->b_error = 0; 904 bp->b_data = addr; 905 bp->b_blkno = dblk; 906 bp->b_proc = NULL; 907 bp->b_dev = sc->sc_bdev; 908 bp->b_private = scp; 909 bp->b_iodone = fss_cluster_iodone; 910 911 bdev_strategy(bp); 912 913 FSS_LOCK(sc, s); 914 scp->fc_xfercount++; 915 FSS_UNLOCK(sc, s); 916 917 dblk += btodb(len); 918 addr += len; 919 todo -= len; 920 } 921 922 /* 923 * Wait for all read requests to complete. 924 */ 925 FSS_LOCK(sc, s); 926 while (scp->fc_xfercount > 0) 927 ltsleep(&scp->fc_data, PRIBIO, "cowwait", 0, &sc->sc_slock); 928 929 scp->fc_type = FSS_CACHE_VALID; 930 setbit(sc->sc_copied, scp->fc_cluster); 931 FSS_UNLOCK(sc, s); 932 933 wakeup(&sc->sc_bs_lwp); 934 } 935 936 /* 937 * Read/write clusters from/to backing store. 938 * For persistent snapshots must be called with cl == 0. off is the 939 * offset into the snapshot. 940 */ 941 static int 942 fss_bs_io(struct fss_softc *sc, fss_io_type rw, 943 u_int32_t cl, off_t off, int len, void *data) 944 { 945 int error; 946 947 off += FSS_CLTOB(sc, cl); 948 949 vn_lock(sc->sc_bs_vp, LK_EXCLUSIVE|LK_RETRY); 950 951 error = vn_rdwr((rw == FSS_READ ? UIO_READ : UIO_WRITE), sc->sc_bs_vp, 952 data, len, off, UIO_SYSSPACE, IO_UNIT|IO_NODELOCKED, 953 sc->sc_bs_lwp->l_cred, NULL, NULL); 954 if (error == 0) { 955 mutex_enter(&sc->sc_bs_vp->v_interlock); 956 error = VOP_PUTPAGES(sc->sc_bs_vp, trunc_page(off), 957 round_page(off+len), PGO_CLEANIT|PGO_SYNCIO|PGO_FREE); 958 } 959 960 VOP_UNLOCK(sc->sc_bs_vp, 0); 961 962 return error; 963 } 964 965 /* 966 * Get a pointer to the indirect slot for this cluster. 967 */ 968 static u_int32_t * 969 fss_bs_indir(struct fss_softc *sc, u_int32_t cl) 970 { 971 u_int32_t icl; 972 int ioff; 973 974 icl = cl/(FSS_CLSIZE(sc)/sizeof(u_int32_t)); 975 ioff = cl%(FSS_CLSIZE(sc)/sizeof(u_int32_t)); 976 977 if (sc->sc_indir_cur == icl) 978 return &sc->sc_indir_data[ioff]; 979 980 if (sc->sc_indir_dirty) { 981 FSS_STAT_INC(sc, indir_write); 982 if (fss_bs_io(sc, FSS_WRITE, sc->sc_indir_cur, 0, 983 FSS_CLSIZE(sc), (void *)sc->sc_indir_data) != 0) 984 return NULL; 985 setbit(sc->sc_indir_valid, sc->sc_indir_cur); 986 } 987 988 sc->sc_indir_dirty = 0; 989 sc->sc_indir_cur = icl; 990 991 if (isset(sc->sc_indir_valid, sc->sc_indir_cur)) { 992 FSS_STAT_INC(sc, indir_read); 993 if (fss_bs_io(sc, FSS_READ, sc->sc_indir_cur, 0, 994 FSS_CLSIZE(sc), (void *)sc->sc_indir_data) != 0) 995 return NULL; 996 } else 997 memset(sc->sc_indir_data, 0, FSS_CLSIZE(sc)); 998 999 return &sc->sc_indir_data[ioff]; 1000 } 1001 1002 /* 1003 * The kernel thread (one for every active snapshot). 1004 * 1005 * After wakeup it cleans the cache and runs the I/O requests. 1006 */ 1007 static void 1008 fss_bs_thread(void *arg) 1009 { 1010 int error, len, nfreed, nio, s; 1011 long off; 1012 char *addr; 1013 u_int32_t c, cl, ch, *indirp; 1014 struct buf *bp, *nbp; 1015 struct fss_softc *sc; 1016 struct fss_cache *scp, *scl; 1017 1018 sc = arg; 1019 1020 scl = sc->sc_cache+sc->sc_cache_size; 1021 1022 nbp = getiobuf(NULL, true); 1023 1024 nfreed = nio = 1; /* Dont sleep the first time */ 1025 1026 FSS_LOCK(sc, s); 1027 1028 for (;;) { 1029 if (nfreed == 0 && nio == 0) 1030 ltsleep(&sc->sc_bs_lwp, PVM-1, "fssbs", 0, 1031 &sc->sc_slock); 1032 1033 if ((sc->sc_flags & FSS_BS_THREAD) == 0) { 1034 sc->sc_bs_lwp = NULL; 1035 wakeup(&sc->sc_bs_lwp); 1036 1037 FSS_UNLOCK(sc, s); 1038 1039 putiobuf(nbp); 1040 #ifdef FSS_STATISTICS 1041 if ((sc->sc_flags & FSS_PERSISTENT) == 0) { 1042 printf("fss%d: cow called %" PRId64 " times," 1043 " copied %" PRId64 " clusters," 1044 " cache full %" PRId64 " times\n", 1045 sc->sc_unit, 1046 FSS_STAT_VAL(sc, cow_calls), 1047 FSS_STAT_VAL(sc, cow_copied), 1048 FSS_STAT_VAL(sc, cow_cache_full)); 1049 printf("fss%d: %" PRId64 " indir reads," 1050 " %" PRId64 " indir writes\n", 1051 sc->sc_unit, 1052 FSS_STAT_VAL(sc, indir_read), 1053 FSS_STAT_VAL(sc, indir_write)); 1054 } 1055 #endif /* FSS_STATISTICS */ 1056 kthread_exit(0); 1057 } 1058 1059 /* 1060 * Process I/O requests (persistent) 1061 */ 1062 1063 if (sc->sc_flags & FSS_PERSISTENT) { 1064 nfreed = nio = 0; 1065 1066 if ((bp = BUFQ_GET(sc->sc_bufq)) == NULL) 1067 continue; 1068 1069 nio++; 1070 1071 if (FSS_ISVALID(sc)) { 1072 FSS_UNLOCK(sc, s); 1073 1074 error = fss_bs_io(sc, FSS_READ, 0, 1075 dbtob(bp->b_blkno), bp->b_bcount, 1076 bp->b_data); 1077 1078 FSS_LOCK(sc, s); 1079 } else 1080 error = ENXIO; 1081 1082 if (error) { 1083 bp->b_error = error; 1084 bp->b_resid = bp->b_bcount; 1085 } else 1086 bp->b_resid = 0; 1087 1088 biodone(bp); 1089 1090 continue; 1091 } 1092 1093 /* 1094 * Clean the cache 1095 */ 1096 nfreed = 0; 1097 for (scp = sc->sc_cache; scp < scl; scp++) { 1098 if (scp->fc_type != FSS_CACHE_VALID) 1099 continue; 1100 1101 FSS_UNLOCK(sc, s); 1102 1103 indirp = fss_bs_indir(sc, scp->fc_cluster); 1104 if (indirp != NULL) { 1105 error = fss_bs_io(sc, FSS_WRITE, sc->sc_clnext, 1106 0, FSS_CLSIZE(sc), scp->fc_data); 1107 } else 1108 error = EIO; 1109 1110 FSS_LOCK(sc, s); 1111 1112 if (error == 0) { 1113 *indirp = sc->sc_clnext++; 1114 sc->sc_indir_dirty = 1; 1115 } else 1116 fss_error(sc, "write bs error %d", error); 1117 1118 scp->fc_type = FSS_CACHE_FREE; 1119 nfreed++; 1120 wakeup(&scp->fc_type); 1121 } 1122 1123 if (nfreed) 1124 wakeup(&sc->sc_cache); 1125 1126 /* 1127 * Process I/O requests 1128 */ 1129 nio = 0; 1130 1131 if ((bp = BUFQ_GET(sc->sc_bufq)) == NULL) 1132 continue; 1133 1134 nio++; 1135 1136 if (!FSS_ISVALID(sc)) { 1137 bp->b_error = ENXIO; 1138 bp->b_resid = bp->b_bcount; 1139 biodone(bp); 1140 continue; 1141 } 1142 1143 /* 1144 * First read from the snapshotted block device. 1145 * XXX Split to only read those parts that have not 1146 * been saved to backing store? 1147 */ 1148 1149 FSS_UNLOCK(sc, s); 1150 1151 buf_init(nbp); 1152 nbp->b_flags = B_READ; 1153 nbp->b_bcount = bp->b_bcount; 1154 nbp->b_bufsize = bp->b_bcount; 1155 nbp->b_error = 0; 1156 nbp->b_data = bp->b_data; 1157 nbp->b_blkno = bp->b_blkno; 1158 nbp->b_proc = bp->b_proc; 1159 nbp->b_dev = sc->sc_bdev; 1160 1161 bdev_strategy(nbp); 1162 1163 if (biowait(nbp) != 0) { 1164 bp->b_resid = bp->b_bcount; 1165 bp->b_error = nbp->b_error; 1166 biodone(bp); 1167 FSS_LOCK(sc, s); 1168 continue; 1169 } 1170 1171 cl = FSS_BTOCL(sc, dbtob(bp->b_blkno)); 1172 off = FSS_CLOFF(sc, dbtob(bp->b_blkno)); 1173 ch = FSS_BTOCL(sc, dbtob(bp->b_blkno)+bp->b_bcount-1); 1174 bp->b_resid = bp->b_bcount; 1175 addr = bp->b_data; 1176 1177 FSS_LOCK(sc, s); 1178 1179 /* 1180 * Replace those parts that have been saved to backing store. 1181 */ 1182 1183 for (c = cl; c <= ch; 1184 c++, off = 0, bp->b_resid -= len, addr += len) { 1185 len = FSS_CLSIZE(sc)-off; 1186 if (len > bp->b_resid) 1187 len = bp->b_resid; 1188 1189 if (isclr(sc->sc_copied, c)) 1190 continue; 1191 1192 FSS_UNLOCK(sc, s); 1193 1194 indirp = fss_bs_indir(sc, c); 1195 1196 FSS_LOCK(sc, s); 1197 1198 if (indirp == NULL || *indirp == 0) { 1199 /* 1200 * Not on backing store. Either in cache 1201 * or hole in the snapshotted block device. 1202 */ 1203 for (scp = sc->sc_cache; scp < scl; scp++) 1204 if (scp->fc_type == FSS_CACHE_VALID && 1205 scp->fc_cluster == c) 1206 break; 1207 if (scp < scl) 1208 memcpy(addr, (char *)scp->fc_data+off, len); 1209 else 1210 memset(addr, 0, len); 1211 continue; 1212 } 1213 /* 1214 * Read from backing store. 1215 */ 1216 1217 FSS_UNLOCK(sc, s); 1218 1219 if ((error = fss_bs_io(sc, FSS_READ, *indirp, 1220 off, len, addr)) != 0) { 1221 bp->b_resid = bp->b_bcount; 1222 bp->b_error = error; 1223 FSS_LOCK(sc, s); 1224 break; 1225 } 1226 1227 FSS_LOCK(sc, s); 1228 1229 } 1230 1231 biodone(bp); 1232 } 1233 } 1234
Indexes created Wed Nov 11 00:26:22 UTC 2026