lfs.c revision 1.18
1/* $NetBSD: lfs.c,v 1.18 2005/10/08 03:21:17 chs Exp $ */ 2/*- 3 * Copyright (c) 2003 The NetBSD Foundation, Inc. 4 * All rights reserved. 5 * 6 * This code is derived from software contributed to The NetBSD Foundation 7 * by Konrad E. Schroder <perseant@hhhh.org>. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. All advertising materials mentioning features or use of this software 18 * must display the following acknowledgement: 19 * This product includes software developed by the NetBSD 20 * Foundation, Inc. and its contributors. 21 * 4. Neither the name of The NetBSD Foundation nor the names of its 22 * contributors may be used to endorse or promote products derived 23 * from this software without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 26 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 27 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 28 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 29 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 35 * POSSIBILITY OF SUCH DAMAGE. 36 */ 37/* 38 * Copyright (c) 1989, 1991, 1993 39 * The Regents of the University of California. All rights reserved. 40 * (c) UNIX System Laboratories, Inc. 41 * All or some portions of this file are derived from material licensed 42 * to the University of California by American Telephone and Telegraph 43 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 44 * the permission of UNIX System Laboratories, Inc. 45 * 46 * Redistribution and use in source and binary forms, with or without 47 * modification, are permitted provided that the following conditions 48 * are met: 49 * 1. Redistributions of source code must retain the above copyright 50 * notice, this list of conditions and the following disclaimer. 51 * 2. Redistributions in binary form must reproduce the above copyright 52 * notice, this list of conditions and the following disclaimer in the 53 * documentation and/or other materials provided with the distribution. 54 * 3. Neither the name of the University nor the names of its contributors 55 * may be used to endorse or promote products derived from this software 56 * without specific prior written permission. 57 * 58 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 59 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 60 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 61 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 62 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 63 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 64 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 65 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 66 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 67 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 68 * SUCH DAMAGE. 69 * 70 * @(#)ufs_bmap.c 8.8 (Berkeley) 8/11/95 71 */ 72 73 74#include <sys/types.h> 75#include <sys/param.h> 76#include <sys/time.h> 77#include <sys/buf.h> 78#include <sys/mount.h> 79 80#include <ufs/ufs/inode.h> 81#include <ufs/ufs/ufsmount.h> 82#define vnode uvnode 83#include <ufs/lfs/lfs.h> 84#undef vnode 85 86#include <assert.h> 87#include <err.h> 88#include <errno.h> 89#include <stdarg.h> 90#include <stdio.h> 91#include <stdlib.h> 92#include <string.h> 93#include <unistd.h> 94 95#include "bufcache.h" 96#include "vnode.h" 97#include "lfs_user.h" 98#include "segwrite.h" 99 100#define panic call_panic 101 102extern u_int32_t cksum(void *, size_t); 103extern u_int32_t lfs_sb_cksum(struct dlfs *); 104extern void pwarn(const char *, ...); 105 106extern struct uvnodelst vnodelist; 107extern struct uvnodelst getvnodelist[VNODE_HASH_MAX]; 108extern int nvnodes; 109 110int fsdirty = 0; 111void (*panic_func)(int, const char *, va_list) = my_vpanic; 112 113/* 114 * LFS buffer and uvnode operations 115 */ 116 117int 118lfs_vop_strategy(struct ubuf * bp) 119{ 120 int count; 121 122 if (bp->b_flags & B_READ) { 123 count = pread(bp->b_vp->v_fd, bp->b_data, bp->b_bcount, 124 dbtob(bp->b_blkno)); 125 if (count == bp->b_bcount) 126 bp->b_flags |= B_DONE; 127 } else { 128 count = pwrite(bp->b_vp->v_fd, bp->b_data, bp->b_bcount, 129 dbtob(bp->b_blkno)); 130 if (count == 0) { 131 perror("pwrite"); 132 return -1; 133 } 134 bp->b_flags &= ~B_DELWRI; 135 reassignbuf(bp, bp->b_vp); 136 } 137 return 0; 138} 139 140int 141lfs_vop_bwrite(struct ubuf * bp) 142{ 143 struct lfs *fs; 144 145 fs = bp->b_vp->v_fs; 146 if (!(bp->b_flags & B_DELWRI)) { 147 fs->lfs_avail -= btofsb(fs, bp->b_bcount); 148 } 149 bp->b_flags |= B_DELWRI | B_LOCKED; 150 reassignbuf(bp, bp->b_vp); 151 brelse(bp); 152 return 0; 153} 154 155/* 156 * ufs_bmaparray does the bmap conversion, and if requested returns the 157 * array of logical blocks which must be traversed to get to a block. 158 * Each entry contains the offset into that block that gets you to the 159 * next block and the disk address of the block (if it is assigned). 160 */ 161int 162ufs_bmaparray(struct lfs * fs, struct uvnode * vp, daddr_t bn, daddr_t * bnp, struct indir * ap, int *nump) 163{ 164 struct inode *ip; 165 struct ubuf *bp; 166 struct indir a[NIADDR + 1], *xap; 167 daddr_t daddr; 168 daddr_t metalbn; 169 int error, num; 170 171 ip = VTOI(vp); 172 173 if (bn >= 0 && bn < NDADDR) { 174 if (nump != NULL) 175 *nump = 0; 176 *bnp = fsbtodb(fs, ip->i_ffs1_db[bn]); 177 if (*bnp == 0) 178 *bnp = -1; 179 return (0); 180 } 181 xap = ap == NULL ? a : ap; 182 if (!nump) 183 nump = # 184 if ((error = ufs_getlbns(fs, vp, bn, xap, nump)) != 0) 185 return (error); 186 187 num = *nump; 188 189 /* Get disk address out of indirect block array */ 190 daddr = ip->i_ffs1_ib[xap->in_off]; 191 192 for (bp = NULL, ++xap; --num; ++xap) { 193 /* Exit the loop if there is no disk address assigned yet and 194 * the indirect block isn't in the cache, or if we were 195 * looking for an indirect block and we've found it. */ 196 197 metalbn = xap->in_lbn; 198 if ((daddr == 0 && !incore(vp, metalbn)) || metalbn == bn) 199 break; 200 /* 201 * If we get here, we've either got the block in the cache 202 * or we have a disk address for it, go fetch it. 203 */ 204 if (bp) 205 brelse(bp); 206 207 xap->in_exists = 1; 208 bp = getblk(vp, metalbn, fs->lfs_bsize); 209 210 if (!(bp->b_flags & (B_DONE | B_DELWRI))) { 211 bp->b_blkno = fsbtodb(fs, daddr); 212 bp->b_flags |= B_READ; 213 VOP_STRATEGY(bp); 214 } 215 daddr = ((ufs_daddr_t *) bp->b_data)[xap->in_off]; 216 } 217 if (bp) 218 brelse(bp); 219 220 daddr = fsbtodb(fs, (ufs_daddr_t) daddr); 221 *bnp = daddr == 0 ? -1 : daddr; 222 return (0); 223} 224 225/* 226 * Create an array of logical block number/offset pairs which represent the 227 * path of indirect blocks required to access a data block. The first "pair" 228 * contains the logical block number of the appropriate single, double or 229 * triple indirect block and the offset into the inode indirect block array. 230 * Note, the logical block number of the inode single/double/triple indirect 231 * block appears twice in the array, once with the offset into the i_ffs1_ib and 232 * once with the offset into the page itself. 233 */ 234int 235ufs_getlbns(struct lfs * fs, struct uvnode * vp, daddr_t bn, struct indir * ap, int *nump) 236{ 237 daddr_t metalbn, realbn; 238 int64_t blockcnt; 239 int lbc; 240 int i, numlevels, off; 241 int lognindir, indir; 242 243 if (nump) 244 *nump = 0; 245 numlevels = 0; 246 realbn = bn; 247 if (bn < 0) 248 bn = -bn; 249 250 lognindir = -1; 251 for (indir = fs->lfs_nindir; indir; indir >>= 1) 252 ++lognindir; 253 254 /* Determine the number of levels of indirection. After this loop is 255 * done, blockcnt indicates the number of data blocks possible at the 256 * given level of indirection, and NIADDR - i is the number of levels 257 * of indirection needed to locate the requested block. */ 258 259 bn -= NDADDR; 260 for (lbc = 0, i = NIADDR;; i--, bn -= blockcnt) { 261 if (i == 0) 262 return (EFBIG); 263 264 lbc += lognindir; 265 blockcnt = (int64_t) 1 << lbc; 266 267 if (bn < blockcnt) 268 break; 269 } 270 271 /* Calculate the address of the first meta-block. */ 272 metalbn = -((realbn >= 0 ? realbn : -realbn) - bn + NIADDR - i); 273 274 /* At each iteration, off is the offset into the bap array which is an 275 * array of disk addresses at the current level of indirection. The 276 * logical block number and the offset in that block are stored into 277 * the argument array. */ 278 ap->in_lbn = metalbn; 279 ap->in_off = off = NIADDR - i; 280 ap->in_exists = 0; 281 ap++; 282 for (++numlevels; i <= NIADDR; i++) { 283 /* If searching for a meta-data block, quit when found. */ 284 if (metalbn == realbn) 285 break; 286 287 lbc -= lognindir; 288 blockcnt = (int64_t) 1 << lbc; 289 off = (bn >> lbc) & (fs->lfs_nindir - 1); 290 291 ++numlevels; 292 ap->in_lbn = metalbn; 293 ap->in_off = off; 294 ap->in_exists = 0; 295 ++ap; 296 297 metalbn -= -1 + (off << lbc); 298 } 299 if (nump) 300 *nump = numlevels; 301 return (0); 302} 303 304int 305lfs_vop_bmap(struct uvnode * vp, daddr_t lbn, daddr_t * daddrp) 306{ 307 return ufs_bmaparray(vp->v_fs, vp, lbn, daddrp, NULL, NULL); 308} 309 310/* Search a block for a specific dinode. */ 311struct ufs1_dinode * 312lfs_ifind(struct lfs * fs, ino_t ino, struct ubuf * bp) 313{ 314 struct ufs1_dinode *dip = (struct ufs1_dinode *) bp->b_data; 315 struct ufs1_dinode *ldip, *fin; 316 317 fin = dip + INOPB(fs); 318 319 /* 320 * Read the inode block backwards, since later versions of the 321 * inode will supercede earlier ones. Though it is unlikely, it is 322 * possible that the same inode will appear in the same inode block. 323 */ 324 for (ldip = fin - 1; ldip >= dip; --ldip) 325 if (ldip->di_inumber == ino) 326 return (ldip); 327 return NULL; 328} 329 330/* 331 * lfs_raw_vget makes us a new vnode from the inode at the given disk address. 332 * XXX it currently loses atime information. 333 */ 334struct uvnode * 335lfs_raw_vget(struct lfs * fs, ino_t ino, int fd, ufs_daddr_t daddr) 336{ 337 struct uvnode *vp; 338 struct inode *ip; 339 struct ufs1_dinode *dip; 340 struct ubuf *bp; 341 int i, hash; 342 343 vp = (struct uvnode *) malloc(sizeof(*vp)); 344 memset(vp, 0, sizeof(*vp)); 345 vp->v_fd = fd; 346 vp->v_fs = fs; 347 vp->v_usecount = 0; 348 vp->v_strategy_op = lfs_vop_strategy; 349 vp->v_bwrite_op = lfs_vop_bwrite; 350 vp->v_bmap_op = lfs_vop_bmap; 351 LIST_INIT(&vp->v_cleanblkhd); 352 LIST_INIT(&vp->v_dirtyblkhd); 353 354 ip = (struct inode *) malloc(sizeof(*ip)); 355 memset(ip, 0, sizeof(*ip)); 356 357 ip->i_din.ffs1_din = (struct ufs1_dinode *) 358 malloc(sizeof(struct ufs1_dinode)); 359 memset(ip->i_din.ffs1_din, 0, sizeof (struct ufs1_dinode)); 360 361 /* Initialize the inode -- from lfs_vcreate. */ 362 ip->inode_ext.lfs = malloc(sizeof(struct lfs_inode_ext)); 363 memset(ip->inode_ext.lfs, 0, sizeof(struct lfs_inode_ext)); 364 vp->v_data = ip; 365 /* ip->i_vnode = vp; */ 366 ip->i_number = ino; 367 ip->i_lockf = 0; 368 ip->i_diroff = 0; 369 ip->i_lfs_effnblks = 0; 370 ip->i_flag = 0; 371 372 /* Load inode block and find inode */ 373 if (daddr > 0) { 374 bread(fs->lfs_devvp, fsbtodb(fs, daddr), fs->lfs_ibsize, NULL, &bp); 375 bp->b_flags |= B_AGE; 376 dip = lfs_ifind(fs, ino, bp); 377 if (dip == NULL) { 378 brelse(bp); 379 free(ip); 380 free(vp); 381 return NULL; 382 } 383 memcpy(ip->i_din.ffs1_din, dip, sizeof(*dip)); 384 brelse(bp); 385 } 386 ip->i_number = ino; 387 /* ip->i_devvp = fs->lfs_devvp; */ 388 ip->i_lfs = fs; 389 390 ip->i_ffs_effnlink = ip->i_ffs1_nlink; 391 ip->i_lfs_effnblks = ip->i_ffs1_blocks; 392 ip->i_lfs_osize = ip->i_ffs1_size; 393#if 0 394 if (fs->lfs_version > 1) { 395 ip->i_ffs1_atime = ts.tv_sec; 396 ip->i_ffs1_atimensec = ts.tv_nsec; 397 } 398#endif 399 400 memset(ip->i_lfs_fragsize, 0, NDADDR * sizeof(*ip->i_lfs_fragsize)); 401 for (i = 0; i < NDADDR; i++) 402 if (ip->i_ffs1_db[i] != 0) 403 ip->i_lfs_fragsize[i] = blksize(fs, ip, i); 404 405 ++nvnodes; 406 hash = ((int)(intptr_t)fs + ino) & (VNODE_HASH_MAX - 1); 407 LIST_INSERT_HEAD(&getvnodelist[hash], vp, v_getvnodes); 408 LIST_INSERT_HEAD(&vnodelist, vp, v_mntvnodes); 409 410 return vp; 411} 412 413static struct uvnode * 414lfs_vget(void *vfs, ino_t ino) 415{ 416 struct lfs *fs = (struct lfs *)vfs; 417 ufs_daddr_t daddr; 418 struct ubuf *bp; 419 IFILE *ifp; 420 421 LFS_IENTRY(ifp, fs, ino, bp); 422 daddr = ifp->if_daddr; 423 brelse(bp); 424 if (daddr <= 0 || dtosn(fs, daddr) >= fs->lfs_nseg) 425 return NULL; 426 return lfs_raw_vget(fs, ino, fs->lfs_ivnode->v_fd, daddr); 427} 428 429/* Check superblock magic number and checksum */ 430static int 431check_sb(struct lfs *fs) 432{ 433 u_int32_t checksum; 434 435 if (fs->lfs_magic != LFS_MAGIC) { 436 printf("Superblock magic number (0x%lx) does not match " 437 "expected 0x%lx\n", (unsigned long) fs->lfs_magic, 438 (unsigned long) LFS_MAGIC); 439 return 1; 440 } 441 /* checksum */ 442 checksum = lfs_sb_cksum(&(fs->lfs_dlfs)); 443 if (fs->lfs_cksum != checksum) { 444 printf("Superblock checksum (%lx) does not match computed checksum (%lx)\n", 445 (unsigned long) fs->lfs_cksum, (unsigned long) checksum); 446 return 1; 447 } 448 return 0; 449} 450 451/* Initialize LFS library; load superblocks and choose which to use. */ 452struct lfs * 453lfs_init(int devfd, daddr_t sblkno, daddr_t idaddr, int dummy_read, int debug) 454{ 455 struct uvnode *devvp; 456 struct ubuf *bp; 457 int tryalt; 458 struct lfs *fs, *altfs; 459 int error; 460 461 vfs_init(); 462 463 devvp = (struct uvnode *) malloc(sizeof(*devvp)); 464 memset(devvp, 0, sizeof(*devvp)); 465 devvp->v_fs = NULL; 466 devvp->v_fd = devfd; 467 devvp->v_strategy_op = raw_vop_strategy; 468 devvp->v_bwrite_op = raw_vop_bwrite; 469 devvp->v_bmap_op = raw_vop_bmap; 470 LIST_INIT(&devvp->v_cleanblkhd); 471 LIST_INIT(&devvp->v_dirtyblkhd); 472 473 tryalt = 0; 474 if (dummy_read) { 475 if (sblkno == 0) 476 sblkno = btodb(LFS_LABELPAD); 477 fs = (struct lfs *) malloc(sizeof(*fs)); 478 memset(fs, 0, sizeof(*fs)); 479 fs->lfs_devvp = devvp; 480 } else { 481 if (sblkno == 0) { 482 sblkno = btodb(LFS_LABELPAD); 483 tryalt = 1; 484 } else if (debug) { 485 printf("No -b flag given, not attempting to verify checkpoint\n"); 486 } 487 error = bread(devvp, sblkno, LFS_SBPAD, NOCRED, &bp); 488 fs = (struct lfs *) malloc(sizeof(*fs)); 489 memset(fs, 0, sizeof(*fs)); 490 fs->lfs_dlfs = *((struct dlfs *) bp->b_data); 491 fs->lfs_devvp = devvp; 492 bp->b_flags |= B_INVAL; 493 brelse(bp); 494 495 if (tryalt) { 496 error = bread(devvp, fsbtodb(fs, fs->lfs_sboffs[1]), 497 LFS_SBPAD, NOCRED, &bp); 498 altfs = (struct lfs *) malloc(sizeof(*altfs)); 499 memset(altfs, 0, sizeof(*altfs)); 500 altfs->lfs_dlfs = *((struct dlfs *) bp->b_data); 501 altfs->lfs_devvp = devvp; 502 bp->b_flags |= B_INVAL; 503 brelse(bp); 504 505 if (check_sb(fs) || fs->lfs_idaddr <= 0) { 506 if (debug) 507 printf("Primary superblock is no good, using first alternate\n"); 508 free(fs); 509 fs = altfs; 510 } else { 511 /* If both superblocks check out, try verification */ 512 if (check_sb(altfs)) { 513 if (debug) 514 printf("First alternate superblock is no good, using primary\n"); 515 free(altfs); 516 } else { 517 if (lfs_verify(fs, altfs, devvp, debug) == fs) { 518 free(altfs); 519 } else { 520 free(fs); 521 fs = altfs; 522 } 523 } 524 } 525 } 526 if (check_sb(fs)) { 527 free(fs); 528 return NULL; 529 } 530 } 531 532 /* Compatibility */ 533 if (fs->lfs_version < 2) { 534 fs->lfs_sumsize = LFS_V1_SUMMARY_SIZE; 535 fs->lfs_ibsize = fs->lfs_bsize; 536 fs->lfs_start = fs->lfs_sboffs[0]; 537 fs->lfs_tstamp = fs->lfs_otstamp; 538 fs->lfs_fsbtodb = 0; 539 } 540 541 if (!dummy_read) { 542 fs->lfs_suflags = (u_int32_t **) malloc(2 * sizeof(u_int32_t *)); 543 fs->lfs_suflags[0] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t)); 544 fs->lfs_suflags[1] = (u_int32_t *) malloc(fs->lfs_nseg * sizeof(u_int32_t)); 545 } 546 547 if (idaddr == 0) 548 idaddr = fs->lfs_idaddr; 549 else 550 fs->lfs_idaddr = idaddr; 551 /* NB: If dummy_read!=0, idaddr==0 here so we get a fake inode. */ 552 fs->lfs_ivnode = lfs_raw_vget(fs, 553 (dummy_read ? LFS_IFILE_INUM : fs->lfs_ifile), devvp->v_fd, 554 idaddr); 555 556 register_vget((void *)fs, lfs_vget); 557 558 return fs; 559} 560 561/* 562 * Check partial segment validity between fs->lfs_offset and the given goal. 563 * 564 * If goal == 0, just keep on going until the segments stop making sense, 565 * and return the address of the last valid partial segment. 566 * 567 * If goal != 0, return the address of the first partial segment that failed, 568 * or "goal" if we reached it without failure (the partial segment *at* goal 569 * need not be valid). 570 */ 571ufs_daddr_t 572try_verify(struct lfs *osb, struct uvnode *devvp, ufs_daddr_t goal, int debug) 573{ 574 ufs_daddr_t daddr, odaddr; 575 SEGSUM *sp; 576 int bc, flag; 577 struct ubuf *bp; 578 ufs_daddr_t nodirop_daddr; 579 u_int64_t serial; 580 581 odaddr = -1; 582 daddr = osb->lfs_offset; 583 nodirop_daddr = daddr; 584 serial = osb->lfs_serial; 585 while (daddr != goal) { 586 flag = 0; 587oncemore: 588 /* Read in summary block */ 589 bread(devvp, fsbtodb(osb, daddr), osb->lfs_sumsize, NULL, &bp); 590 sp = (SEGSUM *)bp->b_data; 591 592 /* 593 * Could be a superblock instead of a segment summary. 594 * XXX should use gseguse, but right now we need to do more 595 * setup before we can...fix this 596 */ 597 if (sp->ss_magic != SS_MAGIC || 598 sp->ss_ident != osb->lfs_ident || 599 sp->ss_serial < serial || 600 sp->ss_sumsum != cksum(&sp->ss_datasum, osb->lfs_sumsize - 601 sizeof(sp->ss_sumsum))) { 602 brelse(bp); 603 if (flag == 0) { 604 flag = 1; 605 daddr += btofsb(osb, LFS_SBPAD); 606 goto oncemore; 607 } 608 break; 609 } 610 ++serial; 611 bc = check_summary(osb, sp, daddr, debug, devvp, NULL); 612 if (bc == 0) { 613 brelse(bp); 614 break; 615 } 616 assert (bc > 0); 617 odaddr = daddr; 618 daddr += btofsb(osb, osb->lfs_sumsize + bc); 619 if (dtosn(osb, odaddr) != dtosn(osb, daddr) || 620 dtosn(osb, daddr) != dtosn(osb, daddr + 621 btofsb(osb, osb->lfs_sumsize + osb->lfs_bsize))) { 622 daddr = sp->ss_next; 623 } 624 if (!(sp->ss_flags & SS_CONT)) 625 nodirop_daddr = daddr; 626 brelse(bp); 627 } 628 629 if (goal == 0) 630 return nodirop_daddr; 631 else 632 return daddr; 633} 634 635/* Use try_verify to check whether the newer superblock is valid. */ 636struct lfs * 637lfs_verify(struct lfs *sb0, struct lfs *sb1, struct uvnode *devvp, int debug) 638{ 639 ufs_daddr_t daddr; 640 struct lfs *osb, *nsb; 641 642 /* 643 * Verify the checkpoint of the newer superblock, 644 * if the timestamp/serial number of the two superblocks is 645 * different. 646 */ 647 648 osb = NULL; 649 if (debug) 650 printf("sb0 %lld, sb1 %lld\n", (long long) sb0->lfs_serial, 651 (long long) sb1->lfs_serial); 652 653 if ((sb0->lfs_version == 1 && 654 sb0->lfs_otstamp != sb1->lfs_otstamp) || 655 (sb0->lfs_version > 1 && 656 sb0->lfs_serial != sb1->lfs_serial)) { 657 if (sb0->lfs_version == 1) { 658 if (sb0->lfs_otstamp > sb1->lfs_otstamp) { 659 osb = sb1; 660 nsb = sb0; 661 } else { 662 osb = sb0; 663 nsb = sb1; 664 } 665 } else { 666 if (sb0->lfs_serial > sb1->lfs_serial) { 667 osb = sb1; 668 nsb = sb0; 669 } else { 670 osb = sb0; 671 nsb = sb1; 672 } 673 } 674 if (debug) { 675 printf("Attempting to verify newer checkpoint..."); 676 fflush(stdout); 677 } 678 daddr = try_verify(osb, devvp, nsb->lfs_offset, debug); 679 680 if (debug) 681 printf("done.\n"); 682 if (daddr == nsb->lfs_offset) { 683 pwarn("** Newer checkpoint verified, recovered %lld seconds of data\n", 684 (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp); 685 sbdirty(); 686 } else { 687 pwarn("** Newer checkpoint invalid, lost %lld seconds of data\n", (long long) nsb->lfs_tstamp - (long long) osb->lfs_tstamp); 688 } 689 return (daddr == nsb->lfs_offset ? nsb : osb); 690 } 691 /* Nothing to check */ 692 return osb; 693} 694 695/* Verify a partial-segment summary; return the number of bytes on disk. */ 696int 697check_summary(struct lfs *fs, SEGSUM *sp, ufs_daddr_t pseg_addr, int debug, 698 struct uvnode *devvp, void (func(ufs_daddr_t, FINFO *))) 699{ 700 FINFO *fp; 701 int bc; /* Bytes in partial segment */ 702 int nblocks; 703 ufs_daddr_t seg_addr, daddr; 704 ufs_daddr_t *dp, *idp; 705 struct ubuf *bp; 706 int i, j, k, datac, len; 707 long sn; 708 u_int32_t *datap; 709 u_int32_t ccksum; 710 711 sn = dtosn(fs, pseg_addr); 712 seg_addr = sntod(fs, sn); 713 714 /* We've already checked the sumsum, just do the data bounds and sum */ 715 716 /* Count the blocks. */ 717 nblocks = howmany(sp->ss_ninos, INOPB(fs)); 718 bc = nblocks << (fs->lfs_version > 1 ? fs->lfs_ffshift : fs->lfs_bshift); 719 assert(bc >= 0); 720 721 fp = (FINFO *) (sp + 1); 722 for (i = 0; i < sp->ss_nfinfo; i++) { 723 nblocks += fp->fi_nblocks; 724 bc += fp->fi_lastlength + ((fp->fi_nblocks - 1) 725 << fs->lfs_bshift); 726 assert(bc >= 0); 727 fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks); 728 } 729 datap = (u_int32_t *) malloc(nblocks * sizeof(*datap)); 730 datac = 0; 731 732 dp = (ufs_daddr_t *) sp; 733 dp += fs->lfs_sumsize / sizeof(ufs_daddr_t); 734 dp--; 735 736 idp = dp; 737 daddr = pseg_addr + btofsb(fs, fs->lfs_sumsize); 738 fp = (FINFO *) (sp + 1); 739 for (i = 0, j = 0; 740 i < sp->ss_nfinfo || j < howmany(sp->ss_ninos, INOPB(fs)); i++) { 741 if (i >= sp->ss_nfinfo && *idp != daddr) { 742 pwarn("Not enough inode blocks in pseg at 0x%" PRIx32 743 ": found %d, wanted %d\n", 744 pseg_addr, j, howmany(sp->ss_ninos, INOPB(fs))); 745 if (debug) 746 pwarn("*idp=%x, daddr=%" PRIx32 "\n", *idp, 747 daddr); 748 break; 749 } 750 while (j < howmany(sp->ss_ninos, INOPB(fs)) && *idp == daddr) { 751 bread(devvp, fsbtodb(fs, daddr), fs->lfs_ibsize, NOCRED, &bp); 752 datap[datac++] = ((u_int32_t *) (bp->b_data))[0]; 753 brelse(bp); 754 755 ++j; 756 daddr += btofsb(fs, fs->lfs_ibsize); 757 --idp; 758 } 759 if (i < sp->ss_nfinfo) { 760 if (func) 761 func(daddr, fp); 762 for (k = 0; k < fp->fi_nblocks; k++) { 763 len = (k == fp->fi_nblocks - 1 ? 764 fp->fi_lastlength 765 : fs->lfs_bsize); 766 bread(devvp, fsbtodb(fs, daddr), len, NOCRED, &bp); 767 datap[datac++] = ((u_int32_t *) (bp->b_data))[0]; 768 brelse(bp); 769 daddr += btofsb(fs, len); 770 } 771 fp = (FINFO *) (fp->fi_blocks + fp->fi_nblocks); 772 } 773 } 774 775 if (datac != nblocks) { 776 pwarn("Partial segment at 0x%llx expected %d blocks counted %d\n", 777 (long long) pseg_addr, nblocks, datac); 778 } 779 ccksum = cksum(datap, nblocks * sizeof(u_int32_t)); 780 /* Check the data checksum */ 781 if (ccksum != sp->ss_datasum) { 782 pwarn("Partial segment at 0x%" PRIx32 " data checksum" 783 " mismatch: given 0x%x, computed 0x%x\n", 784 pseg_addr, sp->ss_datasum, ccksum); 785 free(datap); 786 return 0; 787 } 788 free(datap); 789 assert(bc >= 0); 790 return bc; 791} 792 793/* print message and exit */ 794void 795my_vpanic(int fatal, const char *fmt, va_list ap) 796{ 797 (void) vprintf(fmt, ap); 798 exit(8); 799} 800 801void 802call_panic(const char *fmt, ...) 803{ 804 va_list ap; 805 806 va_start(ap, fmt); 807 panic_func(1, fmt, ap); 808 va_end(ap); 809} 810 811/* Allocate a new inode. */ 812struct uvnode * 813lfs_valloc(struct lfs *fs, ino_t ino) 814{ 815 struct ubuf *bp, *cbp; 816 struct ifile *ifp; 817 ino_t new_ino; 818 int error; 819 int new_gen; 820 CLEANERINFO *cip; 821 822 /* Get the head of the freelist. */ 823 LFS_GET_HEADFREE(fs, cip, cbp, &new_ino); 824 825 /* 826 * Remove the inode from the free list and write the new start 827 * of the free list into the superblock. 828 */ 829 LFS_IENTRY(ifp, fs, new_ino, bp); 830 if (ifp->if_daddr != LFS_UNUSED_DADDR) 831 panic("lfs_valloc: inuse inode %d on the free list", new_ino); 832 LFS_PUT_HEADFREE(fs, cip, cbp, ifp->if_nextfree); 833 834 new_gen = ifp->if_version; /* version was updated by vfree */ 835 brelse(bp); 836 837 /* Extend IFILE so that the next lfs_valloc will succeed. */ 838 if (fs->lfs_freehd == LFS_UNUSED_INUM) { 839 if ((error = extend_ifile(fs)) != 0) { 840 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino); 841 return NULL; 842 } 843 } 844 845 /* Set superblock modified bit and increment file count. */ 846 sbdirty(); 847 ++fs->lfs_nfiles; 848 849 return lfs_raw_vget(fs, ino, fs->lfs_devvp->v_fd, 0x0); 850} 851 852/* 853 * Add a new block to the Ifile, to accommodate future file creations. 854 */ 855int 856extend_ifile(struct lfs *fs) 857{ 858 struct uvnode *vp; 859 struct inode *ip; 860 IFILE *ifp; 861 IFILE_V1 *ifp_v1; 862 struct ubuf *bp, *cbp; 863 daddr_t i, blkno, max; 864 ino_t oldlast; 865 CLEANERINFO *cip; 866 867 vp = fs->lfs_ivnode; 868 ip = VTOI(vp); 869 blkno = lblkno(fs, ip->i_ffs1_size); 870 871 bp = getblk(vp, blkno, fs->lfs_bsize); /* XXX VOP_BALLOC() */ 872 ip->i_ffs1_size += fs->lfs_bsize; 873 874 i = (blkno - fs->lfs_segtabsz - fs->lfs_cleansz) * 875 fs->lfs_ifpb; 876 LFS_GET_HEADFREE(fs, cip, cbp, &oldlast); 877 LFS_PUT_HEADFREE(fs, cip, cbp, i); 878 max = i + fs->lfs_ifpb; 879 fs->lfs_bfree -= btofsb(fs, fs->lfs_bsize); 880 881 if (fs->lfs_version == 1) { 882 for (ifp_v1 = (IFILE_V1 *)bp->b_data; i < max; ++ifp_v1) { 883 ifp_v1->if_version = 1; 884 ifp_v1->if_daddr = LFS_UNUSED_DADDR; 885 ifp_v1->if_nextfree = ++i; 886 } 887 ifp_v1--; 888 ifp_v1->if_nextfree = oldlast; 889 } else { 890 for (ifp = (IFILE *)bp->b_data; i < max; ++ifp) { 891 ifp->if_version = 1; 892 ifp->if_daddr = LFS_UNUSED_DADDR; 893 ifp->if_nextfree = ++i; 894 } 895 ifp--; 896 ifp->if_nextfree = oldlast; 897 } 898 LFS_PUT_TAILFREE(fs, cip, cbp, max - 1); 899 900 LFS_BWRITE_LOG(bp); 901 902 return 0; 903} 904 905