lfs_rfw.c revision 1.26
1 /* $NetBSD: lfs_rfw.c,v 1.26 2015/08/12 18:25:52 dholland Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Konrad E. Schroder <perseant (at) hhhh.org>. 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 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 __KERNEL_RCSID(0, "$NetBSD: lfs_rfw.c,v 1.26 2015/08/12 18:25:52 dholland Exp $"); 34 35 #if defined(_KERNEL_OPT) 36 #include "opt_quota.h" 37 #endif 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/namei.h> 42 #include <sys/proc.h> 43 #include <sys/kernel.h> 44 #include <sys/vnode.h> 45 #include <sys/mount.h> 46 #include <sys/kthread.h> 47 #include <sys/buf.h> 48 #include <sys/device.h> 49 #include <sys/mbuf.h> 50 #include <sys/file.h> 51 #include <sys/disklabel.h> 52 #include <sys/ioctl.h> 53 #include <sys/errno.h> 54 #include <sys/malloc.h> 55 #include <sys/pool.h> 56 #include <sys/socket.h> 57 #include <sys/syslog.h> 58 #include <uvm/uvm_extern.h> 59 #include <sys/sysctl.h> 60 #include <sys/conf.h> 61 #include <sys/kauth.h> 62 63 #include <miscfs/specfs/specdev.h> 64 65 #include <ufs/lfs/ulfs_quotacommon.h> 66 #include <ufs/lfs/ulfs_inode.h> 67 #include <ufs/lfs/ulfsmount.h> 68 #include <ufs/lfs/ulfs_extern.h> 69 70 #include <uvm/uvm.h> 71 #include <uvm/uvm_stat.h> 72 #include <uvm/uvm_pager.h> 73 #include <uvm/uvm_pdaemon.h> 74 75 #include <ufs/lfs/lfs.h> 76 #include <ufs/lfs/lfs_accessors.h> 77 #include <ufs/lfs/lfs_kernel.h> 78 #include <ufs/lfs/lfs_extern.h> 79 80 #include <miscfs/genfs/genfs.h> 81 #include <miscfs/genfs/genfs_node.h> 82 83 /* 84 * Roll-forward code. 85 */ 86 static daddr_t check_segsum(struct lfs *, daddr_t, u_int64_t, 87 kauth_cred_t, int, int *, struct lwp *); 88 89 extern int lfs_do_rfw; 90 91 /* 92 * Allocate a particular inode with a particular version number, freeing 93 * any previous versions of this inode that may have gone before. 94 * Used by the roll-forward code. 95 * 96 * XXX this function does not have appropriate locking to be used on a live fs; 97 * XXX but something similar could probably be used for an "undelete" call. 98 * 99 * Called with the Ifile inode locked. 100 */ 101 int 102 lfs_rf_valloc(struct lfs *fs, ino_t ino, int vers, struct lwp *l, 103 struct vnode **vpp) 104 { 105 struct vattr va; 106 struct vnode *vp; 107 struct inode *ip; 108 int error; 109 110 ASSERT_SEGLOCK(fs); /* XXX it doesn't, really */ 111 112 /* 113 * First, just try a vget. If the version number is the one we want, 114 * we don't have to do anything else. If the version number is wrong, 115 * take appropriate action. 116 */ 117 error = VFS_VGET(fs->lfs_ivnode->v_mount, ino, &vp); 118 if (error == 0) { 119 DLOG((DLOG_RF, "lfs_rf_valloc[1]: ino %d vp %p\n", ino, vp)); 120 121 *vpp = vp; 122 ip = VTOI(vp); 123 if (ip->i_gen == vers) 124 return 0; 125 else if (ip->i_gen < vers) { 126 lfs_truncate(vp, (off_t)0, 0, NOCRED); 127 ip->i_gen = ip->i_ffs1_gen = vers; 128 LFS_SET_UINO(ip, IN_CHANGE | IN_UPDATE); 129 return 0; 130 } else { 131 DLOG((DLOG_RF, "ino %d: sought version %d, got %d\n", 132 ino, vers, ip->i_ffs1_gen)); 133 vput(vp); 134 *vpp = NULLVP; 135 return EEXIST; 136 } 137 } 138 139 /* Not found, create as regular file. */ 140 vattr_null(&va); 141 va.va_type = VREG; 142 va.va_mode = 0; 143 va.va_fileid = ino; 144 va.va_gen = vers; 145 error = vcache_new(fs->lfs_ivnode->v_mount, NULL, &va, NOCRED, &vp); 146 if (error) 147 return error; 148 error = vn_lock(vp, LK_EXCLUSIVE); 149 if (error) { 150 vrele(vp); 151 *vpp = NULLVP; 152 return error; 153 } 154 ip = VTOI(vp); 155 ip->i_nlink = ip->i_ffs1_nlink = 1; 156 *vpp = vp; 157 return 0; 158 } 159 160 /* 161 * Load the appropriate indirect block, and change the appropriate pointer. 162 * Mark the block dirty. Do segment and avail accounting. 163 */ 164 static int 165 update_meta(struct lfs *fs, ino_t ino, int vers, daddr_t lbn, 166 daddr_t ndaddr, size_t size, struct lwp *l) 167 { 168 int error; 169 struct vnode *vp; 170 struct inode *ip; 171 #ifdef DEBUG 172 daddr_t odaddr; 173 struct indir a[ULFS_NIADDR]; 174 int num; 175 int i; 176 #endif /* DEBUG */ 177 struct buf *bp; 178 SEGUSE *sup; 179 180 KASSERT(lbn >= 0); /* no indirect blocks */ 181 182 if ((error = lfs_rf_valloc(fs, ino, vers, l, &vp)) != 0) { 183 DLOG((DLOG_RF, "update_meta: ino %d: lfs_rf_valloc" 184 " returned %d\n", ino, error)); 185 return error; 186 } 187 188 if ((error = lfs_balloc(vp, (lbn << lfs_sb_getbshift(fs)), size, 189 NOCRED, 0, &bp)) != 0) { 190 vput(vp); 191 return (error); 192 } 193 /* No need to write, the block is already on disk */ 194 if (bp->b_oflags & BO_DELWRI) { 195 LFS_UNLOCK_BUF(bp); 196 lfs_sb_addavail(fs, lfs_btofsb(fs, bp->b_bcount)); 197 /* XXX should this wake up fs->lfs_availsleep? */ 198 } 199 brelse(bp, BC_INVAL); 200 201 /* 202 * Extend the file, if it is not large enough already. 203 * XXX this is not exactly right, we don't know how much of the 204 * XXX last block is actually used. We hope that an inode will 205 * XXX appear later to give the correct size. 206 */ 207 ip = VTOI(vp); 208 if (ip->i_size <= (lbn << lfs_sb_getbshift(fs))) { 209 u_int64_t newsize; 210 211 if (lbn < ULFS_NDADDR) 212 newsize = ip->i_ffs1_size = (lbn << lfs_sb_getbshift(fs)) + 213 (size - lfs_sb_getfsize(fs)) + 1; 214 else 215 newsize = ip->i_ffs1_size = (lbn << lfs_sb_getbshift(fs)) + 1; 216 217 if (ip->i_size < newsize) { 218 ip->i_size = newsize; 219 /* 220 * tell vm our new size for the case the inode won't 221 * appear later. 222 */ 223 uvm_vnp_setsize(vp, newsize); 224 } 225 } 226 227 lfs_update_single(fs, NULL, vp, lbn, ndaddr, size); 228 229 LFS_SEGENTRY(sup, fs, lfs_dtosn(fs, ndaddr), bp); 230 sup->su_nbytes += size; 231 LFS_WRITESEGENTRY(sup, fs, lfs_dtosn(fs, ndaddr), bp); 232 233 /* differences here should be due to UNWRITTEN indirect blocks. */ 234 KASSERT((lfs_lblkno(fs, ip->i_size) > ULFS_NDADDR && 235 ip->i_lfs_effnblks == ip->i_ffs1_blocks) || 236 ip->i_lfs_effnblks >= ip->i_ffs1_blocks); 237 238 #ifdef DEBUG 239 /* Now look again to make sure it worked */ 240 ulfs_bmaparray(vp, lbn, &odaddr, &a[0], &num, NULL, NULL); 241 for (i = num; i > 0; i--) { 242 if (!a[i].in_exists) 243 panic("update_meta: absent %d lv indirect block", i); 244 } 245 if (LFS_DBTOFSB(fs, odaddr) != ndaddr) 246 DLOG((DLOG_RF, "update_meta: failed setting ino %d lbn %" 247 PRId64 " to %" PRId64 "\n", ino, lbn, ndaddr)); 248 #endif /* DEBUG */ 249 vput(vp); 250 return 0; 251 } 252 253 static int 254 update_inoblk(struct lfs *fs, daddr_t offset, kauth_cred_t cred, 255 struct lwp *l) 256 { 257 struct vnode *devvp, *vp; 258 struct inode *ip; 259 struct ulfs1_dinode *dip; 260 struct buf *dbp, *ibp; 261 int error; 262 daddr_t daddr; 263 IFILE *ifp; 264 SEGUSE *sup; 265 266 devvp = VTOI(fs->lfs_ivnode)->i_devvp; 267 268 /* 269 * Get the inode, update times and perms. 270 * DO NOT update disk blocks, we do that separately. 271 */ 272 error = bread(devvp, LFS_FSBTODB(fs, offset), lfs_sb_getibsize(fs), 273 0, &dbp); 274 if (error) { 275 DLOG((DLOG_RF, "update_inoblk: bread returned %d\n", error)); 276 return error; 277 } 278 dip = ((struct ulfs1_dinode *)(dbp->b_data)) + LFS_INOPB(fs); 279 while (--dip >= (struct ulfs1_dinode *)dbp->b_data) { 280 if (dip->di_inumber > LFS_IFILE_INUM) { 281 error = lfs_rf_valloc(fs, dip->di_inumber, dip->di_gen, 282 l, &vp); 283 if (error) { 284 DLOG((DLOG_RF, "update_inoblk: lfs_rf_valloc" 285 " returned %d\n", error)); 286 continue; 287 } 288 ip = VTOI(vp); 289 if (dip->di_size != ip->i_size) 290 lfs_truncate(vp, dip->di_size, 0, NOCRED); 291 /* Get mode, link count, size, and times */ 292 memcpy(ip->i_din.ffs1_din, dip, 293 offsetof(struct ulfs1_dinode, di_db[0])); 294 295 /* Then the rest, except di_blocks */ 296 ip->i_flags = ip->i_ffs1_flags = dip->di_flags; 297 ip->i_gen = ip->i_ffs1_gen = dip->di_gen; 298 ip->i_uid = ip->i_ffs1_uid = dip->di_uid; 299 ip->i_gid = ip->i_ffs1_gid = dip->di_gid; 300 301 ip->i_mode = ip->i_ffs1_mode; 302 ip->i_nlink = ip->i_ffs1_nlink; 303 ip->i_size = ip->i_ffs1_size; 304 305 LFS_SET_UINO(ip, IN_CHANGE | IN_UPDATE); 306 307 /* Re-initialize to get type right */ 308 ulfs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, 309 &vp); 310 vput(vp); 311 312 /* Record change in location */ 313 LFS_IENTRY(ifp, fs, dip->di_inumber, ibp); 314 daddr = lfs_if_getdaddr(fs, ifp); 315 lfs_if_setdaddr(fs, ifp, LFS_DBTOFSB(fs, dbp->b_blkno)); 316 error = LFS_BWRITE_LOG(ibp); /* Ifile */ 317 /* And do segment accounting */ 318 if (lfs_dtosn(fs, daddr) != lfs_dtosn(fs, LFS_DBTOFSB(fs, dbp->b_blkno))) { 319 if (daddr > 0) { 320 LFS_SEGENTRY(sup, fs, lfs_dtosn(fs, daddr), 321 ibp); 322 sup->su_nbytes -= sizeof (struct ulfs1_dinode); 323 LFS_WRITESEGENTRY(sup, fs, 324 lfs_dtosn(fs, daddr), 325 ibp); 326 } 327 LFS_SEGENTRY(sup, fs, lfs_dtosn(fs, LFS_DBTOFSB(fs, dbp->b_blkno)), 328 ibp); 329 sup->su_nbytes += sizeof (struct ulfs1_dinode); 330 LFS_WRITESEGENTRY(sup, fs, 331 lfs_dtosn(fs, LFS_DBTOFSB(fs, dbp->b_blkno)), 332 ibp); 333 } 334 } 335 } 336 brelse(dbp, BC_AGE); 337 338 return 0; 339 } 340 341 #define CHECK_CKSUM 0x0001 /* Check the checksum to make sure it's valid */ 342 #define CHECK_UPDATE 0x0002 /* Update Ifile for new data blocks / inodes */ 343 344 static daddr_t 345 check_segsum(struct lfs *fs, daddr_t offset, u_int64_t nextserial, 346 kauth_cred_t cred, int flags, int *pseg_flags, struct lwp *l) 347 { 348 struct vnode *devvp; 349 struct buf *bp, *dbp; 350 int error, nblocks = 0, ninos, i, j; /* XXX: gcc */ 351 SEGSUM *ssp; 352 u_long *dp = NULL, *datap = NULL; /* XXX u_int32_t */ 353 daddr_t oldoffset; 354 int32_t *iaddr; /* XXX ondisk32 */ 355 FINFO *fip; 356 SEGUSE *sup; 357 size_t size; 358 359 devvp = VTOI(fs->lfs_ivnode)->i_devvp; 360 /* 361 * If the segment has a superblock and we're at the top 362 * of the segment, skip the superblock. 363 */ 364 if (lfs_sntod(fs, lfs_dtosn(fs, offset)) == offset) { 365 LFS_SEGENTRY(sup, fs, lfs_dtosn(fs, offset), bp); 366 if (sup->su_flags & SEGUSE_SUPERBLOCK) 367 offset += lfs_btofsb(fs, LFS_SBPAD); 368 brelse(bp, 0); 369 } 370 371 /* Read in the segment summary */ 372 error = bread(devvp, LFS_FSBTODB(fs, offset), lfs_sb_getsumsize(fs), 373 0, &bp); 374 if (error) 375 return -1; 376 377 /* Check summary checksum */ 378 ssp = (SEGSUM *)bp->b_data; 379 if (flags & CHECK_CKSUM) { 380 if (ssp->ss_sumsum != cksum(&ssp->ss_datasum, 381 lfs_sb_getsumsize(fs) - 382 sizeof(ssp->ss_sumsum))) { 383 DLOG((DLOG_RF, "Sumsum error at 0x%" PRIx64 "\n", offset)); 384 offset = -1; 385 goto err1; 386 } 387 if (ssp->ss_nfinfo == 0 && ssp->ss_ninos == 0) { 388 DLOG((DLOG_RF, "Empty pseg at 0x%" PRIx64 "\n", offset)); 389 offset = -1; 390 goto err1; 391 } 392 if (ssp->ss_create < lfs_sb_gettstamp(fs)) { 393 DLOG((DLOG_RF, "Old data at 0x%" PRIx64 "\n", offset)); 394 offset = -1; 395 goto err1; 396 } 397 } 398 if (lfs_sb_getversion(fs) > 1) { 399 if (ssp->ss_serial != nextserial) { 400 DLOG((DLOG_RF, "Unexpected serial number at 0x%" PRIx64 401 "\n", offset)); 402 offset = -1; 403 goto err1; 404 } 405 if (ssp->ss_ident != lfs_sb_getident(fs)) { 406 DLOG((DLOG_RF, "Incorrect fsid (0x%x vs 0x%x) at 0x%" 407 PRIx64 "\n", ssp->ss_ident, lfs_sb_getident(fs), offset)); 408 offset = -1; 409 goto err1; 410 } 411 } 412 if (pseg_flags) 413 *pseg_flags = ssp->ss_flags; 414 oldoffset = offset; 415 offset += lfs_btofsb(fs, lfs_sb_getsumsize(fs)); 416 417 ninos = howmany(ssp->ss_ninos, LFS_INOPB(fs)); 418 /* XXX ondisk32 */ 419 iaddr = (int32_t *)((char*)bp->b_data + lfs_sb_getsumsize(fs) - sizeof(int32_t)); 420 if (flags & CHECK_CKSUM) { 421 /* Count blocks */ 422 nblocks = 0; 423 fip = (FINFO *)((char*)bp->b_data + SEGSUM_SIZE(fs)); 424 for (i = 0; i < ssp->ss_nfinfo; ++i) { 425 nblocks += fip->fi_nblocks; 426 if (fip->fi_nblocks <= 0) 427 break; 428 /* XXX ondisk32 */ 429 fip = (FINFO *)(((char *)fip) + FINFOSIZE + 430 (fip->fi_nblocks * sizeof(int32_t))); 431 } 432 nblocks += ninos; 433 /* Create the sum array */ 434 datap = dp = malloc(nblocks * sizeof(u_long), 435 M_SEGMENT, M_WAITOK); 436 } 437 438 /* Handle individual blocks */ 439 fip = (FINFO *)((char*)bp->b_data + SEGSUM_SIZE(fs)); 440 for (i = 0; i < ssp->ss_nfinfo || ninos; ++i) { 441 /* Inode block? */ 442 if (ninos && *iaddr == offset) { 443 if (flags & CHECK_CKSUM) { 444 /* Read in the head and add to the buffer */ 445 error = bread(devvp, LFS_FSBTODB(fs, offset), lfs_sb_getbsize(fs), 446 0, &dbp); 447 if (error) { 448 offset = -1; 449 goto err2; 450 } 451 (*dp++) = ((u_long *)(dbp->b_data))[0]; 452 brelse(dbp, BC_AGE); 453 } 454 if (flags & CHECK_UPDATE) { 455 if ((error = update_inoblk(fs, offset, cred, l)) 456 != 0) { 457 offset = -1; 458 goto err2; 459 } 460 } 461 offset += lfs_btofsb(fs, lfs_sb_getibsize(fs)); 462 --iaddr; 463 --ninos; 464 --i; /* compensate */ 465 continue; 466 } 467 size = lfs_sb_getbsize(fs); 468 for (j = 0; j < fip->fi_nblocks; ++j) { 469 if (j == fip->fi_nblocks - 1) 470 size = fip->fi_lastlength; 471 if (flags & CHECK_CKSUM) { 472 error = bread(devvp, LFS_FSBTODB(fs, offset), size, 473 0, &dbp); 474 if (error) { 475 offset = -1; 476 goto err2; 477 } 478 (*dp++) = ((u_long *)(dbp->b_data))[0]; 479 brelse(dbp, BC_AGE); 480 } 481 /* Account for and update any direct blocks */ 482 if ((flags & CHECK_UPDATE) && 483 fip->fi_ino > LFS_IFILE_INUM && 484 fip->fi_blocks[j] >= 0) { 485 update_meta(fs, fip->fi_ino, fip->fi_version, 486 fip->fi_blocks[j], offset, size, l); 487 } 488 offset += lfs_btofsb(fs, size); 489 } 490 /* XXX ondisk32 */ 491 fip = (FINFO *)(((char *)fip) + FINFOSIZE 492 + fip->fi_nblocks * sizeof(int32_t)); 493 } 494 /* Checksum the array, compare */ 495 if ((flags & CHECK_CKSUM) && 496 ssp->ss_datasum != cksum(datap, nblocks * sizeof(u_long))) 497 { 498 DLOG((DLOG_RF, "Datasum error at 0x%" PRIx64 499 " (wanted %x got %x)\n", 500 offset, ssp->ss_datasum, cksum(datap, nblocks * 501 sizeof(u_long)))); 502 offset = -1; 503 goto err2; 504 } 505 506 /* If we're at the end of the segment, move to the next */ 507 if (lfs_dtosn(fs, offset + lfs_btofsb(fs, lfs_sb_getsumsize(fs) + lfs_sb_getbsize(fs))) != 508 lfs_dtosn(fs, offset)) { 509 if (lfs_dtosn(fs, offset) == lfs_dtosn(fs, ssp->ss_next)) { 510 offset = -1; 511 goto err2; 512 } 513 offset = ssp->ss_next; 514 DLOG((DLOG_RF, "LFS roll forward: moving to offset 0x%" PRIx64 515 " -> segment %d\n", offset, lfs_dtosn(fs,offset))); 516 } 517 518 if (flags & CHECK_UPDATE) { 519 lfs_sb_subavail(fs, offset - oldoffset); 520 /* Don't clog the buffer queue */ 521 mutex_enter(&lfs_lock); 522 if (locked_queue_count > LFS_MAX_BUFS || 523 locked_queue_bytes > LFS_MAX_BYTES) { 524 lfs_flush(fs, SEGM_CKP, 0); 525 } 526 mutex_exit(&lfs_lock); 527 } 528 529 err2: 530 if (flags & CHECK_CKSUM) 531 free(datap, M_SEGMENT); 532 err1: 533 brelse(bp, BC_AGE); 534 535 /* XXX should we update the serial number even for bad psegs? */ 536 if ((flags & CHECK_UPDATE) && offset > 0 && lfs_sb_getversion(fs) > 1) 537 lfs_sb_setserial(fs, nextserial); 538 return offset; 539 } 540 541 void 542 lfs_roll_forward(struct lfs *fs, struct mount *mp, struct lwp *l) 543 { 544 int flags, dirty; 545 daddr_t offset, oldoffset, lastgoodpseg; 546 int sn, curseg, do_rollforward; 547 struct proc *p; 548 kauth_cred_t cred; 549 SEGUSE *sup; 550 struct buf *bp; 551 552 p = l ? l->l_proc : NULL; 553 cred = p ? p->p_cred : NOCRED; 554 555 /* 556 * Roll forward. 557 * 558 * We don't roll forward for v1 filesystems, because 559 * of the danger that the clock was turned back between the last 560 * checkpoint and crash. This would roll forward garbage. 561 * 562 * v2 filesystems don't have this problem because they use a 563 * monotonically increasing serial number instead of a timestamp. 564 */ 565 do_rollforward = (!(lfs_sb_getpflags(fs) & LFS_PF_CLEAN) && 566 lfs_do_rfw && lfs_sb_getversion(fs) > 1 && p != NULL); 567 if (do_rollforward) { 568 u_int64_t nextserial; 569 /* 570 * Phase I: Find the address of the last good partial 571 * segment that was written after the checkpoint. Mark 572 * the segments in question dirty, so they won't be 573 * reallocated. 574 */ 575 lastgoodpseg = oldoffset = offset = lfs_sb_getoffset(fs); 576 flags = 0x0; 577 DLOG((DLOG_RF, "LFS roll forward phase 1: start at offset 0x%" 578 PRIx64 "\n", offset)); 579 LFS_SEGENTRY(sup, fs, lfs_dtosn(fs, offset), bp); 580 if (!(sup->su_flags & SEGUSE_DIRTY)) 581 lfs_sb_subnclean(fs, 1); 582 sup->su_flags |= SEGUSE_DIRTY; 583 LFS_WRITESEGENTRY(sup, fs, lfs_dtosn(fs, offset), bp); 584 nextserial = lfs_sb_getserial(fs) + 1; 585 while ((offset = check_segsum(fs, offset, nextserial, 586 cred, CHECK_CKSUM, &flags, l)) > 0) { 587 nextserial++; 588 if (lfs_sntod(fs, oldoffset) != lfs_sntod(fs, offset)) { 589 LFS_SEGENTRY(sup, fs, lfs_dtosn(fs, oldoffset), 590 bp); 591 if (!(sup->su_flags & SEGUSE_DIRTY)) 592 lfs_sb_subnclean(fs, 1); 593 sup->su_flags |= SEGUSE_DIRTY; 594 LFS_WRITESEGENTRY(sup, fs, lfs_dtosn(fs, oldoffset), 595 bp); 596 } 597 598 DLOG((DLOG_RF, "LFS roll forward phase 1: offset=0x%" 599 PRIx64 "\n", offset)); 600 if (flags & SS_DIROP) { 601 DLOG((DLOG_RF, "lfs_mountfs: dirops at 0x%" 602 PRIx64 "\n", oldoffset)); 603 if (!(flags & SS_CONT)) { 604 DLOG((DLOG_RF, "lfs_mountfs: dirops end " 605 "at 0x%" PRIx64 "\n", oldoffset)); 606 } 607 } 608 if (!(flags & SS_CONT)) 609 lastgoodpseg = offset; 610 oldoffset = offset; 611 } 612 if (flags & SS_CONT) { 613 DLOG((DLOG_RF, "LFS roll forward: warning: incomplete " 614 "dirops discarded\n")); 615 } 616 DLOG((DLOG_RF, "LFS roll forward phase 1: completed: " 617 "lastgoodpseg=0x%" PRIx64 "\n", lastgoodpseg)); 618 oldoffset = lfs_sb_getoffset(fs); 619 if (lfs_sb_getoffset(fs) != lastgoodpseg) { 620 /* Don't overwrite what we're trying to preserve */ 621 offset = lfs_sb_getoffset(fs); 622 lfs_sb_setoffset(fs, lastgoodpseg); 623 lfs_sb_setcurseg(fs, lfs_sntod(fs, lfs_dtosn(fs, lfs_sb_getoffset(fs)))); 624 for (sn = curseg = lfs_dtosn(fs, lfs_sb_getcurseg(fs));;) { 625 sn = (sn + 1) % lfs_sb_getnseg(fs); 626 if (sn == curseg) 627 panic("lfs_mountfs: no clean segments"); 628 LFS_SEGENTRY(sup, fs, sn, bp); 629 dirty = (sup->su_flags & SEGUSE_DIRTY); 630 brelse(bp, 0); 631 if (!dirty) 632 break; 633 } 634 lfs_sb_setnextseg(fs, lfs_sntod(fs, sn)); 635 636 /* 637 * Phase II: Roll forward from the first superblock. 638 */ 639 while (offset != lastgoodpseg) { 640 DLOG((DLOG_RF, "LFS roll forward phase 2: 0x%" 641 PRIx64 "\n", offset)); 642 offset = check_segsum(fs, offset, 643 lfs_sb_getserial(fs) + 1, cred, CHECK_UPDATE, 644 NULL, l); 645 } 646 647 /* 648 * Finish: flush our changes to disk. 649 */ 650 lfs_segwrite(mp, SEGM_CKP | SEGM_SYNC); 651 DLOG((DLOG_RF, "lfs_mountfs: roll forward ", 652 "recovered %jd blocks\n", 653 (intmax_t)(lastgoodpseg - oldoffset))); 654 } 655 DLOG((DLOG_RF, "LFS roll forward complete\n")); 656 } 657 } 658
Indexes created Mon Sep 29 21:09:56 GMT 2025