1 /*- 2 * modified for EXT2FS support in Lites 1.1 3 * 4 * Aug 1995, Godmar Back (gback@cs.utah.edu) 5 * University of Utah, Department of Computer Science 6 */ 7 /*- 8 * Copyright (c) 1989, 1991, 1993, 1994 9 * The Regents of the University of California. All rights reserved. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)ffs_vfsops.c 8.8 (Berkeley) 4/18/94 36 * $FreeBSD$ 37 */ 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/namei.h> 42 #include <sys/priv.h> 43 #include <sys/proc.h> 44 #include <sys/kernel.h> 45 #include <sys/vnode.h> 46 #include <sys/mount.h> 47 #include <sys/bio.h> 48 #include <sys/buf.h> 49 #include <sys/conf.h> 50 #include <sys/endian.h> 51 #include <sys/fcntl.h> 52 #include <sys/malloc.h> 53 #include <sys/stat.h> 54 #include <sys/mutex.h> 55 56 #include <geom/geom.h> 57 #include <geom/geom_vfs.h> 58 59 #include <fs/ext2fs/ext2_mount.h> 60 #include <fs/ext2fs/inode.h> 61 62 #include <fs/ext2fs/fs.h> 63 #include <fs/ext2fs/ext2fs.h> 64 #include <fs/ext2fs/ext2_dinode.h> 65 #include <fs/ext2fs/ext2_extern.h> 66 67 static int ext2_flushfiles(struct mount *mp, int flags, struct thread *td); 68 static int ext2_mountfs(struct vnode *, struct mount *); 69 static int ext2_reload(struct mount *mp, struct thread *td); 70 static int ext2_sbupdate(struct ext2mount *, int); 71 static int ext2_cgupdate(struct ext2mount *, int); 72 static vfs_unmount_t ext2_unmount; 73 static vfs_root_t ext2_root; 74 static vfs_statfs_t ext2_statfs; 75 static vfs_sync_t ext2_sync; 76 static vfs_vget_t ext2_vget; 77 static vfs_fhtovp_t ext2_fhtovp; 78 static vfs_mount_t ext2_mount; 79 80 MALLOC_DEFINE(M_EXT2NODE, "ext2_node", "EXT2 vnode private part"); 81 static MALLOC_DEFINE(M_EXT2MNT, "ext2_mount", "EXT2 mount structure"); 82 83 static struct vfsops ext2fs_vfsops = { 84 .vfs_fhtovp = ext2_fhtovp, 85 .vfs_mount = ext2_mount, 86 .vfs_root = ext2_root, /* root inode via vget */ 87 .vfs_statfs = ext2_statfs, 88 .vfs_sync = ext2_sync, 89 .vfs_unmount = ext2_unmount, 90 .vfs_vget = ext2_vget, 91 }; 92 93 VFS_SET(ext2fs_vfsops, ext2fs, 0); 94 95 static int ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev, 96 int ronly); 97 static int compute_sb_data(struct vnode * devvp, 98 struct ext2fs * es, struct m_ext2fs * fs); 99 100 static const char *ext2_opts[] = { "acls", "async", "noatime", "noclusterr", 101 "noclusterw", "noexec", "export", "force", "from", "multilabel", 102 "suiddir", "nosymfollow", "sync", "union", NULL }; 103 104 /* 105 * VFS Operations. 106 * 107 * mount system call 108 */ 109 static int 110 ext2_mount(struct mount *mp) 111 { 112 struct vfsoptlist *opts; 113 struct vnode *devvp; 114 struct thread *td; 115 struct ext2mount *ump = NULL; 116 struct m_ext2fs *fs; 117 struct nameidata nd, *ndp = &nd; 118 accmode_t accmode; 119 char *path, *fspec; 120 int error, flags, len; 121 122 td = curthread; 123 opts = mp->mnt_optnew; 124 125 if (vfs_filteropt(opts, ext2_opts)) 126 return (EINVAL); 127 128 vfs_getopt(opts, "fspath", (void **)&path, NULL); 129 /* Double-check the length of path.. */ 130 if (strlen(path) >= MAXMNTLEN) 131 return (ENAMETOOLONG); 132 133 fspec = NULL; 134 error = vfs_getopt(opts, "from", (void **)&fspec, &len); 135 if (!error && fspec[len - 1] != '\0') 136 return (EINVAL); 137 138 /* 139 * If updating, check whether changing from read-only to 140 * read/write; if there is no device name, that's all we do. 141 */ 142 if (mp->mnt_flag & MNT_UPDATE) { 143 ump = VFSTOEXT2(mp); 144 fs = ump->um_e2fs; 145 error = 0; 146 if (fs->e2fs_ronly == 0 && 147 vfs_flagopt(opts, "ro", NULL, 0)) { 148 error = VFS_SYNC(mp, MNT_WAIT); 149 if (error) 150 return (error); 151 flags = WRITECLOSE; 152 if (mp->mnt_flag & MNT_FORCE) 153 flags |= FORCECLOSE; 154 error = ext2_flushfiles(mp, flags, td); 155 if ( error == 0 && fs->e2fs_wasvalid && ext2_cgupdate(ump, MNT_WAIT) == 0) { 156 fs->e2fs->e2fs_state |= E2FS_ISCLEAN; 157 ext2_sbupdate(ump, MNT_WAIT); 158 } 159 fs->e2fs_ronly = 1; 160 vfs_flagopt(opts, "ro", &mp->mnt_flag, MNT_RDONLY); 161 DROP_GIANT(); 162 g_topology_lock(); 163 g_access(ump->um_cp, 0, -1, 0); 164 g_topology_unlock(); 165 PICKUP_GIANT(); 166 } 167 if (!error && (mp->mnt_flag & MNT_RELOAD)) 168 error = ext2_reload(mp, td); 169 if (error) 170 return (error); 171 devvp = ump->um_devvp; 172 if (fs->e2fs_ronly && !vfs_flagopt(opts, "ro", NULL, 0)) { 173 if (ext2_check_sb_compat(fs->e2fs, devvp->v_rdev, 0)) 174 return (EPERM); 175 176 /* 177 * If upgrade to read-write by non-root, then verify 178 * that user has necessary permissions on the device. 179 */ 180 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 181 error = VOP_ACCESS(devvp, VREAD | VWRITE, 182 td->td_ucred, td); 183 if (error) 184 error = priv_check(td, PRIV_VFS_MOUNT_PERM); 185 if (error) { 186 VOP_UNLOCK(devvp, 0); 187 return (error); 188 } 189 VOP_UNLOCK(devvp, 0); 190 DROP_GIANT(); 191 g_topology_lock(); 192 error = g_access(ump->um_cp, 0, 1, 0); 193 g_topology_unlock(); 194 PICKUP_GIANT(); 195 if (error) 196 return (error); 197 198 if ((fs->e2fs->e2fs_state & E2FS_ISCLEAN) == 0 || 199 (fs->e2fs->e2fs_state & E2FS_ERRORS)) { 200 if (mp->mnt_flag & MNT_FORCE) { 201 printf( 202 "WARNING: %s was not properly dismounted\n", fs->e2fs_fsmnt); 203 } else { 204 printf( 205 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", 206 fs->e2fs_fsmnt); 207 return (EPERM); 208 } 209 } 210 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; 211 (void)ext2_cgupdate(ump, MNT_WAIT); 212 fs->e2fs_ronly = 0; 213 MNT_ILOCK(mp); 214 mp->mnt_flag &= ~MNT_RDONLY; 215 MNT_IUNLOCK(mp); 216 } 217 if (vfs_flagopt(opts, "export", NULL, 0)) { 218 /* Process export requests in vfs_mount.c. */ 219 return (error); 220 } 221 } 222 223 /* 224 * Not an update, or updating the name: look up the name 225 * and verify that it refers to a sensible disk device. 226 */ 227 if (fspec == NULL) 228 return (EINVAL); 229 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td); 230 if ((error = namei(ndp)) != 0) 231 return (error); 232 NDFREE(ndp, NDF_ONLY_PNBUF); 233 devvp = ndp->ni_vp; 234 235 if (!vn_isdisk(devvp, &error)) { 236 vput(devvp); 237 return (error); 238 } 239 240 /* 241 * If mount by non-root, then verify that user has necessary 242 * permissions on the device. 243 * 244 * XXXRW: VOP_ACCESS() enough? 245 */ 246 accmode = VREAD; 247 if ((mp->mnt_flag & MNT_RDONLY) == 0) 248 accmode |= VWRITE; 249 error = VOP_ACCESS(devvp, accmode, td->td_ucred, td); 250 if (error) 251 error = priv_check(td, PRIV_VFS_MOUNT_PERM); 252 if (error) { 253 vput(devvp); 254 return (error); 255 } 256 257 if ((mp->mnt_flag & MNT_UPDATE) == 0) { 258 error = ext2_mountfs(devvp, mp); 259 } else { 260 if (devvp != ump->um_devvp) { 261 vput(devvp); 262 return (EINVAL); /* needs translation */ 263 } else 264 vput(devvp); 265 } 266 if (error) { 267 vrele(devvp); 268 return (error); 269 } 270 ump = VFSTOEXT2(mp); 271 fs = ump->um_e2fs; 272 273 /* 274 * Note that this strncpy() is ok because of a check at the start 275 * of ext2_mount(). 276 */ 277 strncpy(fs->e2fs_fsmnt, path, MAXMNTLEN); 278 fs->e2fs_fsmnt[MAXMNTLEN - 1] = '\0'; 279 vfs_mountedfrom(mp, fspec); 280 return (0); 281 } 282 283 static int 284 ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev, int ronly) 285 { 286 287 if (es->e2fs_magic != E2FS_MAGIC) { 288 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n", 289 devtoname(dev), es->e2fs_magic, E2FS_MAGIC); 290 return (1); 291 } 292 if (es->e2fs_rev > E2FS_REV0) { 293 if (es->e2fs_features_incompat & ~(EXT2F_INCOMPAT_SUPP | 294 EXT4F_RO_INCOMPAT_SUPP)) { 295 printf( 296 "WARNING: mount of %s denied due to unsupported optional features\n", 297 devtoname(dev)); 298 return (1); 299 } 300 if (!ronly && 301 (es->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP)) { 302 printf("WARNING: R/W mount of %s denied due to " 303 "unsupported optional features\n", devtoname(dev)); 304 return (1); 305 } 306 } 307 return (0); 308 } 309 310 /* 311 * This computes the fields of the m_ext2fs structure from the 312 * data in the ext2fs structure read in. 313 */ 314 static int 315 compute_sb_data(struct vnode *devvp, struct ext2fs *es, 316 struct m_ext2fs *fs) 317 { 318 int db_count, error; 319 int i; 320 int logic_sb_block = 1; /* XXX for now */ 321 struct buf *bp; 322 uint32_t e2fs_descpb; 323 324 fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize; 325 fs->e2fs_bsize = 1U << fs->e2fs_bshift; 326 fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1; 327 fs->e2fs_qbmask = fs->e2fs_bsize - 1; 328 fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << es->e2fs_log_fsize; 329 if (fs->e2fs_fsize) 330 fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize; 331 fs->e2fs_bpg = es->e2fs_bpg; 332 fs->e2fs_fpg = es->e2fs_fpg; 333 fs->e2fs_ipg = es->e2fs_ipg; 334 if (es->e2fs_rev == E2FS_REV0) { 335 fs->e2fs_isize = E2FS_REV0_INODE_SIZE ; 336 } else { 337 fs->e2fs_isize = es->e2fs_inode_size; 338 339 /* 340 * Simple sanity check for superblock inode size value. 341 */ 342 if (EXT2_INODE_SIZE(fs) < E2FS_REV0_INODE_SIZE || 343 EXT2_INODE_SIZE(fs) > fs->e2fs_bsize || 344 (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) { 345 printf("ext2fs: invalid inode size %d\n", 346 fs->e2fs_isize); 347 return (EIO); 348 } 349 } 350 /* Check for extra isize in big inodes. */ 351 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_EXTRA_ISIZE) && 352 EXT2_INODE_SIZE(fs) < sizeof(struct ext2fs_dinode)) { 353 printf("ext2fs: no space for extra inode timestamps\n"); 354 return (EINVAL); 355 } 356 357 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs); 358 fs->e2fs_itpg = fs->e2fs_ipg / fs->e2fs_ipb; 359 /* s_resuid / s_resgid ? */ 360 fs->e2fs_gcount = howmany(es->e2fs_bcount - es->e2fs_first_dblock, 361 EXT2_BLOCKS_PER_GROUP(fs)); 362 e2fs_descpb = fs->e2fs_bsize / sizeof(struct ext2_gd); 363 db_count = howmany(fs->e2fs_gcount, e2fs_descpb); 364 fs->e2fs_gdbcount = db_count; 365 fs->e2fs_gd = malloc(db_count * fs->e2fs_bsize, 366 M_EXT2MNT, M_WAITOK); 367 fs->e2fs_contigdirs = malloc(fs->e2fs_gcount * 368 sizeof(*fs->e2fs_contigdirs), M_EXT2MNT, M_WAITOK | M_ZERO); 369 370 /* 371 * Adjust logic_sb_block. 372 * Godmar thinks: if the blocksize is greater than 1024, then 373 * the superblock is logically part of block zero. 374 */ 375 if(fs->e2fs_bsize > SBSIZE) 376 logic_sb_block = 0; 377 for (i = 0; i < db_count; i++) { 378 error = bread(devvp , 379 fsbtodb(fs, logic_sb_block + i + 1 ), 380 fs->e2fs_bsize, NOCRED, &bp); 381 if (error) { 382 free(fs->e2fs_contigdirs, M_EXT2MNT); 383 free(fs->e2fs_gd, M_EXT2MNT); 384 brelse(bp); 385 return (error); 386 } 387 e2fs_cgload((struct ext2_gd *)bp->b_data, 388 &fs->e2fs_gd[ 389 i * fs->e2fs_bsize / sizeof(struct ext2_gd)], 390 fs->e2fs_bsize); 391 brelse(bp); 392 bp = NULL; 393 } 394 /* Initialization for the ext2 Orlov allocator variant. */ 395 fs->e2fs_total_dir = 0; 396 for (i = 0; i < fs->e2fs_gcount; i++) 397 fs->e2fs_total_dir += fs->e2fs_gd[i].ext2bgd_ndirs; 398 399 if (es->e2fs_rev == E2FS_REV0 || 400 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_LARGEFILE)) 401 fs->e2fs_maxfilesize = 0x7fffffff; 402 else { 403 fs->e2fs_maxfilesize = 0xffffffffffff; 404 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_HUGE_FILE)) 405 fs->e2fs_maxfilesize = 0x7fffffffffffffff; 406 } 407 if (es->e4fs_flags & E2FS_UNSIGNED_HASH) { 408 fs->e2fs_uhash = 3; 409 } else if ((es->e4fs_flags & E2FS_SIGNED_HASH) == 0) { 410 #ifdef __CHAR_UNSIGNED__ 411 es->e4fs_flags |= E2FS_UNSIGNED_HASH; 412 fs->e2fs_uhash = 3; 413 #else 414 es->e4fs_flags |= E2FS_SIGNED_HASH; 415 #endif 416 } 417 418 return (0); 419 } 420 421 /* 422 * Reload all incore data for a filesystem (used after running fsck on 423 * the root filesystem and finding things to fix). The filesystem must 424 * be mounted read-only. 425 * 426 * Things to do to update the mount: 427 * 1) invalidate all cached meta-data. 428 * 2) re-read superblock from disk. 429 * 3) invalidate all cluster summary information. 430 * 4) invalidate all inactive vnodes. 431 * 5) invalidate all cached file data. 432 * 6) re-read inode data for all active vnodes. 433 * XXX we are missing some steps, in particular # 3, this has to be reviewed. 434 */ 435 static int 436 ext2_reload(struct mount *mp, struct thread *td) 437 { 438 struct vnode *vp, *mvp, *devvp; 439 struct inode *ip; 440 struct buf *bp; 441 struct ext2fs *es; 442 struct m_ext2fs *fs; 443 struct csum *sump; 444 int error, i; 445 int32_t *lp; 446 447 if ((mp->mnt_flag & MNT_RDONLY) == 0) 448 return (EINVAL); 449 /* 450 * Step 1: invalidate all cached meta-data. 451 */ 452 devvp = VFSTOEXT2(mp)->um_devvp; 453 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 454 if (vinvalbuf(devvp, 0, 0, 0) != 0) 455 panic("ext2_reload: dirty1"); 456 VOP_UNLOCK(devvp, 0); 457 458 /* 459 * Step 2: re-read superblock from disk. 460 * constants have been adjusted for ext2 461 */ 462 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 463 return (error); 464 es = (struct ext2fs *)bp->b_data; 465 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) { 466 brelse(bp); 467 return (EIO); /* XXX needs translation */ 468 } 469 fs = VFSTOEXT2(mp)->um_e2fs; 470 bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs)); 471 472 if((error = compute_sb_data(devvp, es, fs)) != 0) { 473 brelse(bp); 474 return (error); 475 } 476 #ifdef UNKLAR 477 if (fs->fs_sbsize < SBSIZE) 478 bp->b_flags |= B_INVAL; 479 #endif 480 brelse(bp); 481 482 /* 483 * Step 3: invalidate all cluster summary information. 484 */ 485 if (fs->e2fs_contigsumsize > 0) { 486 lp = fs->e2fs_maxcluster; 487 sump = fs->e2fs_clustersum; 488 for (i = 0; i < fs->e2fs_gcount; i++, sump++) { 489 *lp++ = fs->e2fs_contigsumsize; 490 sump->cs_init = 0; 491 bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1); 492 } 493 } 494 495 loop: 496 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 497 /* 498 * Step 4: invalidate all cached file data. 499 */ 500 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) { 501 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 502 goto loop; 503 } 504 if (vinvalbuf(vp, 0, 0, 0)) 505 panic("ext2_reload: dirty2"); 506 507 /* 508 * Step 5: re-read inode data for all active vnodes. 509 */ 510 ip = VTOI(vp); 511 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 512 (int)fs->e2fs_bsize, NOCRED, &bp); 513 if (error) { 514 VOP_UNLOCK(vp, 0); 515 vrele(vp); 516 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 517 return (error); 518 } 519 ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + 520 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip); 521 brelse(bp); 522 VOP_UNLOCK(vp, 0); 523 vrele(vp); 524 } 525 return (0); 526 } 527 528 /* 529 * Common code for mount and mountroot. 530 */ 531 static int 532 ext2_mountfs(struct vnode *devvp, struct mount *mp) 533 { 534 struct ext2mount *ump; 535 struct buf *bp; 536 struct m_ext2fs *fs; 537 struct ext2fs *es; 538 struct cdev *dev = devvp->v_rdev; 539 struct g_consumer *cp; 540 struct bufobj *bo; 541 struct csum *sump; 542 int error; 543 int ronly; 544 int i, size; 545 int32_t *lp; 546 int32_t e2fs_maxcontig; 547 548 ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0); 549 /* XXX: use VOP_ACESS to check FS perms */ 550 DROP_GIANT(); 551 g_topology_lock(); 552 error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1); 553 g_topology_unlock(); 554 PICKUP_GIANT(); 555 VOP_UNLOCK(devvp, 0); 556 if (error) 557 return (error); 558 559 /* XXX: should we check for some sectorsize or 512 instead? */ 560 if (((SBSIZE % cp->provider->sectorsize) != 0) || 561 (SBSIZE < cp->provider->sectorsize)) { 562 DROP_GIANT(); 563 g_topology_lock(); 564 g_vfs_close(cp); 565 g_topology_unlock(); 566 PICKUP_GIANT(); 567 return (EINVAL); 568 } 569 570 bo = &devvp->v_bufobj; 571 bo->bo_private = cp; 572 bo->bo_ops = g_vfs_bufops; 573 if (devvp->v_rdev->si_iosize_max != 0) 574 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max; 575 if (mp->mnt_iosize_max > MAXPHYS) 576 mp->mnt_iosize_max = MAXPHYS; 577 578 bp = NULL; 579 ump = NULL; 580 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 581 goto out; 582 es = (struct ext2fs *)bp->b_data; 583 if (ext2_check_sb_compat(es, dev, ronly) != 0) { 584 error = EINVAL; /* XXX needs translation */ 585 goto out; 586 } 587 if ((es->e2fs_state & E2FS_ISCLEAN) == 0 || 588 (es->e2fs_state & E2FS_ERRORS)) { 589 if (ronly || (mp->mnt_flag & MNT_FORCE)) { 590 printf( 591 "WARNING: Filesystem was not properly dismounted\n"); 592 } else { 593 printf( 594 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n"); 595 error = EPERM; 596 goto out; 597 } 598 } 599 ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO); 600 601 /* 602 * I don't know whether this is the right strategy. Note that 603 * we dynamically allocate both an m_ext2fs and an ext2fs 604 * while Linux keeps the super block in a locked buffer. 605 */ 606 ump->um_e2fs = malloc(sizeof(struct m_ext2fs), 607 M_EXT2MNT, M_WAITOK | M_ZERO); 608 ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs), 609 M_EXT2MNT, M_WAITOK); 610 mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF); 611 bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs)); 612 if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs))) 613 goto out; 614 615 /* 616 * Calculate the maximum contiguous blocks and size of cluster summary 617 * array. In FFS this is done by newfs; however, the superblock 618 * in ext2fs doesn't have these variables, so we can calculate 619 * them here. 620 */ 621 e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize); 622 ump->um_e2fs->e2fs_contigsumsize = MIN(e2fs_maxcontig, EXT2_MAXCONTIG); 623 if (ump->um_e2fs->e2fs_contigsumsize > 0) { 624 size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t); 625 ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK); 626 size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum); 627 ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK); 628 lp = ump->um_e2fs->e2fs_maxcluster; 629 sump = ump->um_e2fs->e2fs_clustersum; 630 for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) { 631 *lp++ = ump->um_e2fs->e2fs_contigsumsize; 632 sump->cs_init = 0; 633 sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) * 634 sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO); 635 } 636 } 637 638 brelse(bp); 639 bp = NULL; 640 fs = ump->um_e2fs; 641 fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */ 642 643 /* 644 * If the fs is not mounted read-only, make sure the super block is 645 * always written back on a sync(). 646 */ 647 fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0; 648 if (ronly == 0) { 649 fs->e2fs_fmod = 1; /* mark it modified */ 650 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; /* set fs invalid */ 651 } 652 mp->mnt_data = ump; 653 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev); 654 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; 655 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; 656 MNT_ILOCK(mp); 657 mp->mnt_flag |= MNT_LOCAL; 658 MNT_IUNLOCK(mp); 659 ump->um_mountp = mp; 660 ump->um_dev = dev; 661 ump->um_devvp = devvp; 662 ump->um_bo = &devvp->v_bufobj; 663 ump->um_cp = cp; 664 665 /* 666 * Setting those two parameters allowed us to use 667 * ufs_bmap w/o changse! 668 */ 669 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); 670 ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1; 671 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); 672 if (ronly == 0) 673 ext2_sbupdate(ump, MNT_WAIT); 674 /* 675 * Initialize filesystem stat information in mount struct. 676 */ 677 MNT_ILOCK(mp); 678 mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_EXTENDED_SHARED | 679 MNTK_USES_BCACHE; 680 MNT_IUNLOCK(mp); 681 return (0); 682 out: 683 if (bp) 684 brelse(bp); 685 if (cp != NULL) { 686 DROP_GIANT(); 687 g_topology_lock(); 688 g_vfs_close(cp); 689 g_topology_unlock(); 690 PICKUP_GIANT(); 691 } 692 if (ump) { 693 mtx_destroy(EXT2_MTX(ump)); 694 free(ump->um_e2fs->e2fs_gd, M_EXT2MNT); 695 free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT); 696 free(ump->um_e2fs->e2fs, M_EXT2MNT); 697 free(ump->um_e2fs, M_EXT2MNT); 698 free(ump, M_EXT2MNT); 699 mp->mnt_data = NULL; 700 } 701 return (error); 702 } 703 704 /* 705 * Unmount system call. 706 */ 707 static int 708 ext2_unmount(struct mount *mp, int mntflags) 709 { 710 struct ext2mount *ump; 711 struct m_ext2fs *fs; 712 struct csum *sump; 713 int error, flags, i, ronly; 714 715 flags = 0; 716 if (mntflags & MNT_FORCE) { 717 if (mp->mnt_flag & MNT_ROOTFS) 718 return (EINVAL); 719 flags |= FORCECLOSE; 720 } 721 if ((error = ext2_flushfiles(mp, flags, curthread)) != 0) 722 return (error); 723 ump = VFSTOEXT2(mp); 724 fs = ump->um_e2fs; 725 ronly = fs->e2fs_ronly; 726 if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) { 727 if (fs->e2fs_wasvalid) 728 fs->e2fs->e2fs_state |= E2FS_ISCLEAN; 729 ext2_sbupdate(ump, MNT_WAIT); 730 } 731 732 DROP_GIANT(); 733 g_topology_lock(); 734 g_vfs_close(ump->um_cp); 735 g_topology_unlock(); 736 PICKUP_GIANT(); 737 vrele(ump->um_devvp); 738 sump = fs->e2fs_clustersum; 739 for (i = 0; i < fs->e2fs_gcount; i++, sump++) 740 free(sump->cs_sum, M_EXT2MNT); 741 free(fs->e2fs_clustersum, M_EXT2MNT); 742 free(fs->e2fs_maxcluster, M_EXT2MNT); 743 free(fs->e2fs_gd, M_EXT2MNT); 744 free(fs->e2fs_contigdirs, M_EXT2MNT); 745 free(fs->e2fs, M_EXT2MNT); 746 free(fs, M_EXT2MNT); 747 free(ump, M_EXT2MNT); 748 mp->mnt_data = NULL; 749 MNT_ILOCK(mp); 750 mp->mnt_flag &= ~MNT_LOCAL; 751 MNT_IUNLOCK(mp); 752 return (error); 753 } 754 755 /* 756 * Flush out all the files in a filesystem. 757 */ 758 static int 759 ext2_flushfiles(struct mount *mp, int flags, struct thread *td) 760 { 761 int error; 762 763 error = vflush(mp, 0, flags, td); 764 return (error); 765 } 766 /* 767 * Get filesystem statistics. 768 */ 769 int 770 ext2_statfs(struct mount *mp, struct statfs *sbp) 771 { 772 struct ext2mount *ump; 773 struct m_ext2fs *fs; 774 uint32_t overhead, overhead_per_group, ngdb; 775 int i, ngroups; 776 777 ump = VFSTOEXT2(mp); 778 fs = ump->um_e2fs; 779 if (fs->e2fs->e2fs_magic != E2FS_MAGIC) 780 panic("ext2_statfs"); 781 782 /* 783 * Compute the overhead (FS structures) 784 */ 785 overhead_per_group = 786 1 /* block bitmap */ + 787 1 /* inode bitmap */ + 788 fs->e2fs_itpg; 789 overhead = fs->e2fs->e2fs_first_dblock + 790 fs->e2fs_gcount * overhead_per_group; 791 if (fs->e2fs->e2fs_rev > E2FS_REV0 && 792 fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) { 793 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) { 794 if (cg_has_sb(i)) 795 ngroups++; 796 } 797 } else { 798 ngroups = fs->e2fs_gcount; 799 } 800 ngdb = fs->e2fs_gdbcount; 801 if (fs->e2fs->e2fs_rev > E2FS_REV0 && 802 fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE) 803 ngdb += fs->e2fs->e2fs_reserved_ngdb; 804 overhead += ngroups * (1 /* superblock */ + ngdb); 805 806 sbp->f_bsize = EXT2_FRAG_SIZE(fs); 807 sbp->f_iosize = EXT2_BLOCK_SIZE(fs); 808 sbp->f_blocks = fs->e2fs->e2fs_bcount - overhead; 809 sbp->f_bfree = fs->e2fs->e2fs_fbcount; 810 sbp->f_bavail = sbp->f_bfree - fs->e2fs->e2fs_rbcount; 811 sbp->f_files = fs->e2fs->e2fs_icount; 812 sbp->f_ffree = fs->e2fs->e2fs_ficount; 813 return (0); 814 } 815 816 /* 817 * Go through the disk queues to initiate sandbagged IO; 818 * go through the inodes to write those that have been modified; 819 * initiate the writing of the super block if it has been modified. 820 * 821 * Note: we are always called with the filesystem marked `MPBUSY'. 822 */ 823 static int 824 ext2_sync(struct mount *mp, int waitfor) 825 { 826 struct vnode *mvp, *vp; 827 struct thread *td; 828 struct inode *ip; 829 struct ext2mount *ump = VFSTOEXT2(mp); 830 struct m_ext2fs *fs; 831 int error, allerror = 0; 832 833 td = curthread; 834 fs = ump->um_e2fs; 835 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */ 836 printf("fs = %s\n", fs->e2fs_fsmnt); 837 panic("ext2_sync: rofs mod"); 838 } 839 840 /* 841 * Write back each (modified) inode. 842 */ 843 loop: 844 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) { 845 if (vp->v_type == VNON) { 846 VI_UNLOCK(vp); 847 continue; 848 } 849 ip = VTOI(vp); 850 if ((ip->i_flag & 851 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 852 (vp->v_bufobj.bo_dirty.bv_cnt == 0 || 853 waitfor == MNT_LAZY)) { 854 VI_UNLOCK(vp); 855 continue; 856 } 857 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td); 858 if (error) { 859 if (error == ENOENT) { 860 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp); 861 goto loop; 862 } 863 continue; 864 } 865 if ((error = VOP_FSYNC(vp, waitfor, td)) != 0) 866 allerror = error; 867 VOP_UNLOCK(vp, 0); 868 vrele(vp); 869 } 870 871 /* 872 * Force stale filesystem control information to be flushed. 873 */ 874 if (waitfor != MNT_LAZY) { 875 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 876 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0) 877 allerror = error; 878 VOP_UNLOCK(ump->um_devvp, 0); 879 } 880 881 /* 882 * Write back modified superblock. 883 */ 884 if (fs->e2fs_fmod != 0) { 885 fs->e2fs_fmod = 0; 886 fs->e2fs->e2fs_wtime = time_second; 887 if ((error = ext2_cgupdate(ump, waitfor)) != 0) 888 allerror = error; 889 } 890 return (allerror); 891 } 892 893 /* 894 * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it 895 * in from disk. If it is in core, wait for the lock bit to clear, then 896 * return the inode locked. Detection and handling of mount points must be 897 * done by the calling routine. 898 */ 899 static int 900 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp) 901 { 902 struct m_ext2fs *fs; 903 struct inode *ip; 904 struct ext2mount *ump; 905 struct buf *bp; 906 struct vnode *vp; 907 struct thread *td; 908 int i, error; 909 int used_blocks; 910 911 td = curthread; 912 error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL); 913 if (error || *vpp != NULL) 914 return (error); 915 916 ump = VFSTOEXT2(mp); 917 ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO); 918 919 /* Allocate a new vnode/inode. */ 920 if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) { 921 *vpp = NULL; 922 free(ip, M_EXT2NODE); 923 return (error); 924 } 925 vp->v_data = ip; 926 ip->i_vnode = vp; 927 ip->i_e2fs = fs = ump->um_e2fs; 928 ip->i_ump = ump; 929 ip->i_number = ino; 930 931 lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL); 932 error = insmntque(vp, mp); 933 if (error != 0) { 934 free(ip, M_EXT2NODE); 935 *vpp = NULL; 936 return (error); 937 } 938 error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL); 939 if (error || *vpp != NULL) 940 return (error); 941 942 /* Read in the disk contents for the inode, copy into the inode. */ 943 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 944 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) { 945 /* 946 * The inode does not contain anything useful, so it would 947 * be misleading to leave it on its hash chain. With mode 948 * still zero, it will be unlinked and returned to the free 949 * list by vput(). 950 */ 951 brelse(bp); 952 vput(vp); 953 *vpp = NULL; 954 return (error); 955 } 956 /* convert ext2 inode to dinode */ 957 ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) * 958 ino_to_fsbo(fs, ino)), ip); 959 ip->i_block_group = ino_to_cg(fs, ino); 960 ip->i_next_alloc_block = 0; 961 ip->i_next_alloc_goal = 0; 962 963 /* 964 * Now we want to make sure that block pointers for unused 965 * blocks are zeroed out - ext2_balloc depends on this 966 * although for regular files and directories only 967 * 968 * If IN_E4EXTENTS is enabled, unused blocks are not zeroed 969 * out because we could corrupt the extent tree. 970 */ 971 if (!(ip->i_flag & IN_E4EXTENTS) && 972 (S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode))) { 973 used_blocks = howmany(ip->i_size, fs->e2fs_bsize); 974 for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) 975 ip->i_db[i] = 0; 976 } 977 #ifdef EXT2FS_DEBUG 978 ext2_print_inode(ip); 979 #endif 980 bqrelse(bp); 981 982 /* 983 * Initialize the vnode from the inode, check for aliases. 984 * Note that the underlying vnode may have changed. 985 */ 986 if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) { 987 vput(vp); 988 *vpp = NULL; 989 return (error); 990 } 991 992 /* 993 * Finish inode initialization. 994 */ 995 996 /* 997 * Set up a generation number for this inode if it does not 998 * already have one. This should only happen on old filesystems. 999 */ 1000 if (ip->i_gen == 0) { 1001 ip->i_gen = random() + 1; 1002 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1003 ip->i_flag |= IN_MODIFIED; 1004 } 1005 *vpp = vp; 1006 return (0); 1007 } 1008 1009 /* 1010 * File handle to vnode 1011 * 1012 * Have to be really careful about stale file handles: 1013 * - check that the inode number is valid 1014 * - call ext2_vget() to get the locked inode 1015 * - check for an unallocated inode (i_mode == 0) 1016 * - check that the given client host has export rights and return 1017 * those rights via. exflagsp and credanonp 1018 */ 1019 static int 1020 ext2_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp) 1021 { 1022 struct inode *ip; 1023 struct ufid *ufhp; 1024 struct vnode *nvp; 1025 struct m_ext2fs *fs; 1026 int error; 1027 1028 ufhp = (struct ufid *)fhp; 1029 fs = VFSTOEXT2(mp)->um_e2fs; 1030 if (ufhp->ufid_ino < EXT2_ROOTINO || 1031 ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg) 1032 return (ESTALE); 1033 1034 error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp); 1035 if (error) { 1036 *vpp = NULLVP; 1037 return (error); 1038 } 1039 ip = VTOI(nvp); 1040 if (ip->i_mode == 0 || 1041 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) { 1042 vput(nvp); 1043 *vpp = NULLVP; 1044 return (ESTALE); 1045 } 1046 *vpp = nvp; 1047 vnode_create_vobject(*vpp, 0, curthread); 1048 return (0); 1049 } 1050 1051 /* 1052 * Write a superblock and associated information back to disk. 1053 */ 1054 static int 1055 ext2_sbupdate(struct ext2mount *mp, int waitfor) 1056 { 1057 struct m_ext2fs *fs = mp->um_e2fs; 1058 struct ext2fs *es = fs->e2fs; 1059 struct buf *bp; 1060 int error = 0; 1061 1062 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0); 1063 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs)); 1064 if (waitfor == MNT_WAIT) 1065 error = bwrite(bp); 1066 else 1067 bawrite(bp); 1068 1069 /* 1070 * The buffers for group descriptors, inode bitmaps and block bitmaps 1071 * are not busy at this point and are (hopefully) written by the 1072 * usual sync mechanism. No need to write them here. 1073 */ 1074 return (error); 1075 } 1076 int 1077 ext2_cgupdate(struct ext2mount *mp, int waitfor) 1078 { 1079 struct m_ext2fs *fs = mp->um_e2fs; 1080 struct buf *bp; 1081 int i, error = 0, allerror = 0; 1082 1083 allerror = ext2_sbupdate(mp, waitfor); 1084 for (i = 0; i < fs->e2fs_gdbcount; i++) { 1085 bp = getblk(mp->um_devvp, fsbtodb(fs, 1086 fs->e2fs->e2fs_first_dblock + 1087 1 /* superblock */ + i), fs->e2fs_bsize, 0, 0, 0); 1088 e2fs_cgsave(&fs->e2fs_gd[ 1089 i * fs->e2fs_bsize / sizeof(struct ext2_gd)], 1090 (struct ext2_gd *)bp->b_data, fs->e2fs_bsize); 1091 if (waitfor == MNT_WAIT) 1092 error = bwrite(bp); 1093 else 1094 bawrite(bp); 1095 } 1096 1097 if (!allerror && error) 1098 allerror = error; 1099 return (allerror); 1100 } 1101 /* 1102 * Return the root of a filesystem. 1103 */ 1104 static int 1105 ext2_root(struct mount *mp, int flags, struct vnode **vpp) 1106 { 1107 struct vnode *nvp; 1108 int error; 1109 1110 error = VFS_VGET(mp, EXT2_ROOTINO, LK_EXCLUSIVE, &nvp); 1111 if (error) 1112 return (error); 1113 *vpp = nvp; 1114 return (0); 1115 } 1116