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 = 0; 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 - 1) 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 printf( 295 "WARNING: mount of %s denied due to unsupported optional features\n", 296 devtoname(dev)); 297 return (1); 298 } 299 if (!ronly && 300 (es->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP)) { 301 printf("WARNING: R/W mount of %s denied due to " 302 "unsupported optional features\n", devtoname(dev)); 303 return (1); 304 } 305 } 306 return (0); 307 } 308 309 /* 310 * This computes the fields of the ext2_sb_info structure from the 311 * data in the ext2_super_block structure read in. 312 */ 313 static int 314 compute_sb_data(struct vnode *devvp, struct ext2fs *es, 315 struct m_ext2fs *fs) 316 { 317 int db_count, error; 318 int i; 319 int logic_sb_block = 1; /* XXX for now */ 320 struct buf *bp; 321 322 fs->e2fs_bsize = EXT2_MIN_BLOCK_SIZE << es->e2fs_log_bsize; 323 fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize; 324 fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1; 325 fs->e2fs_qbmask = fs->e2fs_bsize - 1; 326 fs->e2fs_blocksize_bits = es->e2fs_log_bsize + 10; 327 fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << es->e2fs_log_fsize; 328 if (fs->e2fs_fsize) 329 fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize; 330 fs->e2fs_bpg = es->e2fs_bpg; 331 fs->e2fs_fpg = es->e2fs_fpg; 332 fs->e2fs_ipg = es->e2fs_ipg; 333 if (es->e2fs_rev == E2FS_REV0) { 334 fs->e2fs_first_inode = EXT2_FIRSTINO; 335 fs->e2fs_isize = E2FS_REV0_INODE_SIZE ; 336 } else { 337 fs->e2fs_first_inode = es->e2fs_first_ino; 338 fs->e2fs_isize = es->e2fs_inode_size; 339 340 /* 341 * Simple sanity check for superblock inode size value. 342 */ 343 if (EXT2_INODE_SIZE(fs) < E2FS_REV0_INODE_SIZE || 344 EXT2_INODE_SIZE(fs) > fs->e2fs_bsize || 345 (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) { 346 printf("ext2fs: invalid inode size %d\n", 347 fs->e2fs_isize); 348 return (EIO); 349 } 350 } 351 /* Check for extra isize in big inodes. */ 352 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT4F_ROCOMPAT_EXTRA_ISIZE) && 353 EXT2_INODE_SIZE(fs) < sizeof(struct ext2fs_dinode)) { 354 printf("ext2fs: no space for extra inode timestamps\n"); 355 return (EINVAL); 356 } 357 358 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs); 359 fs->e2fs_itpg = fs->e2fs_ipg /fs->e2fs_ipb; 360 fs->e2fs_descpb = fs->e2fs_bsize / sizeof(struct ext2_gd); 361 /* s_resuid / s_resgid ? */ 362 fs->e2fs_gcount = (es->e2fs_bcount - es->e2fs_first_dblock + 363 EXT2_BLOCKS_PER_GROUP(fs) - 1) / EXT2_BLOCKS_PER_GROUP(fs); 364 db_count = (fs->e2fs_gcount + EXT2_DESC_PER_BLOCK(fs) - 1) / 365 EXT2_DESC_PER_BLOCK(fs); 366 fs->e2fs_gdbcount = db_count; 367 fs->e2fs_gd = malloc(db_count * fs->e2fs_bsize, 368 M_EXT2MNT, M_WAITOK); 369 fs->e2fs_contigdirs = malloc(fs->e2fs_gcount * 370 sizeof(*fs->e2fs_contigdirs), M_EXT2MNT, M_WAITOK); 371 372 /* 373 * Adjust logic_sb_block. 374 * Godmar thinks: if the blocksize is greater than 1024, then 375 * the superblock is logically part of block zero. 376 */ 377 if(fs->e2fs_bsize > SBSIZE) 378 logic_sb_block = 0; 379 for (i = 0; i < db_count; i++) { 380 error = bread(devvp , 381 fsbtodb(fs, logic_sb_block + i + 1 ), 382 fs->e2fs_bsize, NOCRED, &bp); 383 if (error) { 384 free(fs->e2fs_gd, M_EXT2MNT); 385 brelse(bp); 386 return (error); 387 } 388 e2fs_cgload((struct ext2_gd *)bp->b_data, 389 &fs->e2fs_gd[ 390 i * fs->e2fs_bsize / sizeof(struct ext2_gd)], 391 fs->e2fs_bsize); 392 brelse(bp); 393 bp = NULL; 394 } 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 fs->e2fs_contigdirs[i] = 0; 399 } 400 if (es->e2fs_rev == E2FS_REV0 || 401 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_LARGEFILE)) 402 fs->e2fs_maxfilesize = 0x7fffffff; 403 else 404 fs->e2fs_maxfilesize = 0x7fffffffffffffff; 405 return (0); 406 } 407 408 /* 409 * Reload all incore data for a filesystem (used after running fsck on 410 * the root filesystem and finding things to fix). The filesystem must 411 * be mounted read-only. 412 * 413 * Things to do to update the mount: 414 * 1) invalidate all cached meta-data. 415 * 2) re-read superblock from disk. 416 * 3) invalidate all cluster summary information. 417 * 4) invalidate all inactive vnodes. 418 * 5) invalidate all cached file data. 419 * 6) re-read inode data for all active vnodes. 420 * XXX we are missing some steps, in particular # 3, this has to be reviewed. 421 */ 422 static int 423 ext2_reload(struct mount *mp, struct thread *td) 424 { 425 struct vnode *vp, *mvp, *devvp; 426 struct inode *ip; 427 struct buf *bp; 428 struct ext2fs *es; 429 struct m_ext2fs *fs; 430 struct csum *sump; 431 int error, i; 432 int32_t *lp; 433 434 if ((mp->mnt_flag & MNT_RDONLY) == 0) 435 return (EINVAL); 436 /* 437 * Step 1: invalidate all cached meta-data. 438 */ 439 devvp = VFSTOEXT2(mp)->um_devvp; 440 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 441 if (vinvalbuf(devvp, 0, 0, 0) != 0) 442 panic("ext2_reload: dirty1"); 443 VOP_UNLOCK(devvp, 0); 444 445 /* 446 * Step 2: re-read superblock from disk. 447 * constants have been adjusted for ext2 448 */ 449 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 450 return (error); 451 es = (struct ext2fs *)bp->b_data; 452 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) { 453 brelse(bp); 454 return (EIO); /* XXX needs translation */ 455 } 456 fs = VFSTOEXT2(mp)->um_e2fs; 457 bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs)); 458 459 if((error = compute_sb_data(devvp, es, fs)) != 0) { 460 brelse(bp); 461 return (error); 462 } 463 #ifdef UNKLAR 464 if (fs->fs_sbsize < SBSIZE) 465 bp->b_flags |= B_INVAL; 466 #endif 467 brelse(bp); 468 469 /* 470 * Step 3: invalidate all cluster summary information. 471 */ 472 if (fs->e2fs_contigsumsize > 0) { 473 lp = fs->e2fs_maxcluster; 474 sump = fs->e2fs_clustersum; 475 for (i = 0; i < fs->e2fs_gcount; i++, sump++) { 476 *lp++ = fs->e2fs_contigsumsize; 477 sump->cs_init = 0; 478 bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1); 479 } 480 } 481 482 loop: 483 MNT_ILOCK(mp); 484 MNT_VNODE_FOREACH(vp, mp, mvp) { 485 VI_LOCK(vp); 486 if (vp->v_iflag & VI_DOOMED) { 487 VI_UNLOCK(vp); 488 continue; 489 } 490 MNT_IUNLOCK(mp); 491 /* 492 * Step 4: invalidate all cached file data. 493 */ 494 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) { 495 MNT_VNODE_FOREACH_ABORT(mp, mvp); 496 goto loop; 497 } 498 if (vinvalbuf(vp, 0, 0, 0)) 499 panic("ext2_reload: dirty2"); 500 501 /* 502 * Step 5: re-read inode data for all active vnodes. 503 */ 504 ip = VTOI(vp); 505 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 506 (int)fs->e2fs_bsize, NOCRED, &bp); 507 if (error) { 508 VOP_UNLOCK(vp, 0); 509 vrele(vp); 510 MNT_VNODE_FOREACH_ABORT(mp, mvp); 511 return (error); 512 } 513 ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + 514 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip); 515 brelse(bp); 516 VOP_UNLOCK(vp, 0); 517 vrele(vp); 518 MNT_ILOCK(mp); 519 } 520 MNT_IUNLOCK(mp); 521 return (0); 522 } 523 524 /* 525 * Common code for mount and mountroot. 526 */ 527 static int 528 ext2_mountfs(struct vnode *devvp, struct mount *mp) 529 { 530 struct ext2mount *ump; 531 struct buf *bp; 532 struct m_ext2fs *fs; 533 struct ext2fs *es; 534 struct cdev *dev = devvp->v_rdev; 535 struct g_consumer *cp; 536 struct bufobj *bo; 537 struct csum *sump; 538 int error; 539 int ronly; 540 int i, size; 541 int32_t *lp; 542 543 ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0); 544 /* XXX: use VOP_ACESS to check FS perms */ 545 DROP_GIANT(); 546 g_topology_lock(); 547 error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1); 548 g_topology_unlock(); 549 PICKUP_GIANT(); 550 VOP_UNLOCK(devvp, 0); 551 if (error) 552 return (error); 553 554 /* XXX: should we check for some sectorsize or 512 instead? */ 555 if (((SBSIZE % cp->provider->sectorsize) != 0) || 556 (SBSIZE < cp->provider->sectorsize)) { 557 DROP_GIANT(); 558 g_topology_lock(); 559 g_vfs_close(cp); 560 g_topology_unlock(); 561 PICKUP_GIANT(); 562 return (EINVAL); 563 } 564 565 bo = &devvp->v_bufobj; 566 bo->bo_private = cp; 567 bo->bo_ops = g_vfs_bufops; 568 if (devvp->v_rdev->si_iosize_max != 0) 569 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max; 570 if (mp->mnt_iosize_max > MAXPHYS) 571 mp->mnt_iosize_max = MAXPHYS; 572 573 bp = NULL; 574 ump = NULL; 575 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 576 goto out; 577 es = (struct ext2fs *)bp->b_data; 578 if (ext2_check_sb_compat(es, dev, ronly) != 0) { 579 error = EINVAL; /* XXX needs translation */ 580 goto out; 581 } 582 if ((es->e2fs_state & E2FS_ISCLEAN) == 0 || 583 (es->e2fs_state & E2FS_ERRORS)) { 584 if (ronly || (mp->mnt_flag & MNT_FORCE)) { 585 printf( 586 "WARNING: Filesystem was not properly dismounted\n"); 587 } else { 588 printf( 589 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n"); 590 error = EPERM; 591 goto out; 592 } 593 } 594 ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO); 595 596 /* 597 * I don't know whether this is the right strategy. Note that 598 * we dynamically allocate both an ext2_sb_info and an ext2_super_block 599 * while Linux keeps the super block in a locked buffer. 600 */ 601 ump->um_e2fs = malloc(sizeof(struct m_ext2fs), 602 M_EXT2MNT, M_WAITOK); 603 ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs), 604 M_EXT2MNT, M_WAITOK); 605 mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF); 606 bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs)); 607 if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs))) 608 goto out; 609 610 /* 611 * Calculate the maximum contiguous blocks and size of cluster summary 612 * array. In FFS this is done by newfs; however, the superblock 613 * in ext2fs doesn't have these variables, so we can calculate 614 * them here. 615 */ 616 ump->um_e2fs->e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize); 617 if (ump->um_e2fs->e2fs_maxcontig > 0) 618 ump->um_e2fs->e2fs_contigsumsize = 619 MIN(ump->um_e2fs->e2fs_maxcontig, EXT2_MAXCONTIG); 620 else 621 ump->um_e2fs->e2fs_contigsumsize = 0; 622 if (ump->um_e2fs->e2fs_contigsumsize > 0) { 623 size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t); 624 ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK); 625 size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum); 626 ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK); 627 lp = ump->um_e2fs->e2fs_maxcluster; 628 sump = ump->um_e2fs->e2fs_clustersum; 629 for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) { 630 *lp++ = ump->um_e2fs->e2fs_contigsumsize; 631 sump->cs_init = 0; 632 sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) * 633 sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO); 634 } 635 } 636 637 brelse(bp); 638 bp = NULL; 639 fs = ump->um_e2fs; 640 fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */ 641 642 /* 643 * If the fs is not mounted read-only, make sure the super block is 644 * always written back on a sync(). 645 */ 646 fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0; 647 if (ronly == 0) { 648 fs->e2fs_fmod = 1; /* mark it modified */ 649 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; /* set fs invalid */ 650 } 651 mp->mnt_data = ump; 652 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev); 653 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; 654 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; 655 MNT_ILOCK(mp); 656 mp->mnt_flag |= MNT_LOCAL; 657 MNT_IUNLOCK(mp); 658 ump->um_mountp = mp; 659 ump->um_dev = dev; 660 ump->um_devvp = devvp; 661 ump->um_bo = &devvp->v_bufobj; 662 ump->um_cp = cp; 663 664 /* 665 * Setting those two parameters allowed us to use 666 * ufs_bmap w/o changse! 667 */ 668 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); 669 ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1; 670 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); 671 if (ronly == 0) 672 ext2_sbupdate(ump, MNT_WAIT); 673 /* 674 * Initialize filesystem stat information in mount struct. 675 */ 676 MNT_ILOCK(mp); 677 mp->mnt_kern_flag |= MNTK_MPSAFE | MNTK_LOOKUP_SHARED | 678 MNTK_EXTENDED_SHARED; 679 MNT_IUNLOCK(mp); 680 return (0); 681 out: 682 if (bp) 683 brelse(bp); 684 if (cp != NULL) { 685 DROP_GIANT(); 686 g_topology_lock(); 687 g_vfs_close(cp); 688 g_topology_unlock(); 689 PICKUP_GIANT(); 690 } 691 if (ump) { 692 mtx_destroy(EXT2_MTX(ump)); 693 free(ump->um_e2fs->e2fs_gd, M_EXT2MNT); 694 free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT); 695 free(ump->um_e2fs->e2fs, M_EXT2MNT); 696 free(ump->um_e2fs, M_EXT2MNT); 697 free(ump, M_EXT2MNT); 698 mp->mnt_data = NULL; 699 } 700 return (error); 701 } 702 703 /* 704 * Unmount system call. 705 */ 706 static int 707 ext2_unmount(struct mount *mp, int mntflags) 708 { 709 struct ext2mount *ump; 710 struct m_ext2fs *fs; 711 struct csum *sump; 712 int error, flags, i, ronly; 713 714 flags = 0; 715 if (mntflags & MNT_FORCE) { 716 if (mp->mnt_flag & MNT_ROOTFS) 717 return (EINVAL); 718 flags |= FORCECLOSE; 719 } 720 if ((error = ext2_flushfiles(mp, flags, curthread)) != 0) 721 return (error); 722 ump = VFSTOEXT2(mp); 723 fs = ump->um_e2fs; 724 ronly = fs->e2fs_ronly; 725 if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) { 726 if (fs->e2fs_wasvalid) 727 fs->e2fs->e2fs_state |= E2FS_ISCLEAN; 728 ext2_sbupdate(ump, MNT_WAIT); 729 } 730 731 DROP_GIANT(); 732 g_topology_lock(); 733 g_vfs_close(ump->um_cp); 734 g_topology_unlock(); 735 PICKUP_GIANT(); 736 vrele(ump->um_devvp); 737 sump = fs->e2fs_clustersum; 738 for (i = 0; i < fs->e2fs_gcount; i++, sump++) 739 free(sump->cs_sum, M_EXT2MNT); 740 free(fs->e2fs_clustersum, M_EXT2MNT); 741 free(fs->e2fs_maxcluster, M_EXT2MNT); 742 free(fs->e2fs_gd, M_EXT2MNT); 743 free(fs->e2fs_contigdirs, M_EXT2MNT); 744 free(fs->e2fs, M_EXT2MNT); 745 free(fs, M_EXT2MNT); 746 free(ump, M_EXT2MNT); 747 mp->mnt_data = NULL; 748 MNT_ILOCK(mp); 749 mp->mnt_flag &= ~MNT_LOCAL; 750 MNT_IUNLOCK(mp); 751 return (error); 752 } 753 754 /* 755 * Flush out all the files in a filesystem. 756 */ 757 static int 758 ext2_flushfiles(struct mount *mp, int flags, struct thread *td) 759 { 760 int error; 761 762 error = vflush(mp, 0, flags, td); 763 return (error); 764 } 765 /* 766 * Get file system statistics. 767 */ 768 int 769 ext2_statfs(struct mount *mp, struct statfs *sbp) 770 { 771 struct ext2mount *ump; 772 struct m_ext2fs *fs; 773 uint32_t overhead, overhead_per_group, ngdb; 774 int i, ngroups; 775 776 ump = VFSTOEXT2(mp); 777 fs = ump->um_e2fs; 778 if (fs->e2fs->e2fs_magic != E2FS_MAGIC) 779 panic("ext2fs_statvfs"); 780 781 /* 782 * Compute the overhead (FS structures) 783 */ 784 overhead_per_group = 785 1 /* block bitmap */ + 786 1 /* inode bitmap */ + 787 fs->e2fs_itpg; 788 overhead = fs->e2fs->e2fs_first_dblock + 789 fs->e2fs_gcount * overhead_per_group; 790 if (fs->e2fs->e2fs_rev > E2FS_REV0 && 791 fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) { 792 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) { 793 if (cg_has_sb(i)) 794 ngroups++; 795 } 796 } else { 797 ngroups = fs->e2fs_gcount; 798 } 799 ngdb = fs->e2fs_gdbcount; 800 if (fs->e2fs->e2fs_rev > E2FS_REV0 && 801 fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE) 802 ngdb += fs->e2fs->e2fs_reserved_ngdb; 803 overhead += ngroups * (1 /* superblock */ + ngdb); 804 805 sbp->f_bsize = EXT2_FRAG_SIZE(fs); 806 sbp->f_iosize = EXT2_BLOCK_SIZE(fs); 807 sbp->f_blocks = fs->e2fs->e2fs_bcount - overhead; 808 sbp->f_bfree = fs->e2fs->e2fs_fbcount; 809 sbp->f_bavail = sbp->f_bfree - fs->e2fs->e2fs_rbcount; 810 sbp->f_files = fs->e2fs->e2fs_icount; 811 sbp->f_ffree = fs->e2fs->e2fs_ficount; 812 return (0); 813 } 814 815 /* 816 * Go through the disk queues to initiate sandbagged IO; 817 * go through the inodes to write those that have been modified; 818 * initiate the writing of the super block if it has been modified. 819 * 820 * Note: we are always called with the filesystem marked `MPBUSY'. 821 */ 822 static int 823 ext2_sync(struct mount *mp, int waitfor) 824 { 825 struct vnode *mvp, *vp; 826 struct thread *td; 827 struct inode *ip; 828 struct ext2mount *ump = VFSTOEXT2(mp); 829 struct m_ext2fs *fs; 830 int error, allerror = 0; 831 832 td = curthread; 833 fs = ump->um_e2fs; 834 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */ 835 printf("fs = %s\n", fs->e2fs_fsmnt); 836 panic("ext2_sync: rofs mod"); 837 } 838 839 /* 840 * Write back each (modified) inode. 841 */ 842 MNT_ILOCK(mp); 843 loop: 844 MNT_VNODE_FOREACH(vp, mp, mvp) { 845 VI_LOCK(vp); 846 if (vp->v_type == VNON || (vp->v_iflag & VI_DOOMED)) { 847 VI_UNLOCK(vp); 848 continue; 849 } 850 MNT_IUNLOCK(mp); 851 ip = VTOI(vp); 852 if ((ip->i_flag & 853 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 854 (vp->v_bufobj.bo_dirty.bv_cnt == 0 || 855 waitfor == MNT_LAZY)) { 856 VI_UNLOCK(vp); 857 MNT_ILOCK(mp); 858 continue; 859 } 860 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td); 861 if (error) { 862 MNT_ILOCK(mp); 863 if (error == ENOENT) { 864 MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp); 865 goto loop; 866 } 867 continue; 868 } 869 if ((error = VOP_FSYNC(vp, waitfor, td)) != 0) 870 allerror = error; 871 VOP_UNLOCK(vp, 0); 872 vrele(vp); 873 MNT_ILOCK(mp); 874 } 875 MNT_IUNLOCK(mp); 876 877 /* 878 * Force stale file system control information to be flushed. 879 */ 880 if (waitfor != MNT_LAZY) { 881 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 882 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0) 883 allerror = error; 884 VOP_UNLOCK(ump->um_devvp, 0); 885 } 886 887 /* 888 * Write back modified superblock. 889 */ 890 if (fs->e2fs_fmod != 0) { 891 fs->e2fs_fmod = 0; 892 fs->e2fs->e2fs_wtime = time_second; 893 if ((error = ext2_cgupdate(ump, waitfor)) != 0) 894 allerror = error; 895 } 896 return (allerror); 897 } 898 899 /* 900 * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it 901 * in from disk. If it is in core, wait for the lock bit to clear, then 902 * return the inode locked. Detection and handling of mount points must be 903 * done by the calling routine. 904 */ 905 static int 906 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp) 907 { 908 struct m_ext2fs *fs; 909 struct inode *ip; 910 struct ext2mount *ump; 911 struct buf *bp; 912 struct vnode *vp; 913 struct cdev *dev; 914 struct thread *td; 915 int i, error; 916 int used_blocks; 917 918 td = curthread; 919 error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL); 920 if (error || *vpp != NULL) 921 return (error); 922 923 ump = VFSTOEXT2(mp); 924 dev = ump->um_dev; 925 926 /* 927 * If this malloc() is performed after the getnewvnode() 928 * it might block, leaving a vnode with a NULL v_data to be 929 * found by ext2_sync() if a sync happens to fire right then, 930 * which will cause a panic because ext2_sync() blindly 931 * dereferences vp->v_data (as well it should). 932 */ 933 ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO); 934 935 /* Allocate a new vnode/inode. */ 936 if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) { 937 *vpp = NULL; 938 free(ip, M_EXT2NODE); 939 return (error); 940 } 941 vp->v_data = ip; 942 ip->i_vnode = vp; 943 ip->i_e2fs = fs = ump->um_e2fs; 944 ip->i_ump = ump; 945 ip->i_number = ino; 946 947 lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL); 948 error = insmntque(vp, mp); 949 if (error != 0) { 950 free(ip, M_EXT2NODE); 951 *vpp = NULL; 952 return (error); 953 } 954 error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL); 955 if (error || *vpp != NULL) 956 return (error); 957 958 /* Read in the disk contents for the inode, copy into the inode. */ 959 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 960 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) { 961 /* 962 * The inode does not contain anything useful, so it would 963 * be misleading to leave it on its hash chain. With mode 964 * still zero, it will be unlinked and returned to the free 965 * list by vput(). 966 */ 967 brelse(bp); 968 vput(vp); 969 *vpp = NULL; 970 return (error); 971 } 972 /* convert ext2 inode to dinode */ 973 ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) * 974 ino_to_fsbo(fs, ino)), ip); 975 ip->i_block_group = ino_to_cg(fs, ino); 976 ip->i_next_alloc_block = 0; 977 ip->i_next_alloc_goal = 0; 978 979 /* 980 * Now we want to make sure that block pointers for unused 981 * blocks are zeroed out - ext2_balloc depends on this 982 * although for regular files and directories only 983 */ 984 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) { 985 used_blocks = (ip->i_size+fs->e2fs_bsize-1) / fs->e2fs_bsize; 986 for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) 987 ip->i_db[i] = 0; 988 } 989 /* 990 ext2_print_inode(ip); 991 */ 992 bqrelse(bp); 993 994 /* 995 * Initialize the vnode from the inode, check for aliases. 996 * Note that the underlying vnode may have changed. 997 */ 998 if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) { 999 vput(vp); 1000 *vpp = NULL; 1001 return (error); 1002 } 1003 1004 /* 1005 * Finish inode initialization. 1006 */ 1007 1008 /* 1009 * Set up a generation number for this inode if it does not 1010 * already have one. This should only happen on old filesystems. 1011 */ 1012 if (ip->i_gen == 0) { 1013 ip->i_gen = random() / 2 + 1; 1014 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1015 ip->i_flag |= IN_MODIFIED; 1016 } 1017 *vpp = vp; 1018 return (0); 1019 } 1020 1021 /* 1022 * File handle to vnode 1023 * 1024 * Have to be really careful about stale file handles: 1025 * - check that the inode number is valid 1026 * - call ext2_vget() to get the locked inode 1027 * - check for an unallocated inode (i_mode == 0) 1028 * - check that the given client host has export rights and return 1029 * those rights via. exflagsp and credanonp 1030 */ 1031 static int 1032 ext2_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp) 1033 { 1034 struct inode *ip; 1035 struct ufid *ufhp; 1036 struct vnode *nvp; 1037 struct m_ext2fs *fs; 1038 int error; 1039 1040 ufhp = (struct ufid *)fhp; 1041 fs = VFSTOEXT2(mp)->um_e2fs; 1042 if (ufhp->ufid_ino < EXT2_ROOTINO || 1043 ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg) 1044 return (ESTALE); 1045 1046 error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp); 1047 if (error) { 1048 *vpp = NULLVP; 1049 return (error); 1050 } 1051 ip = VTOI(nvp); 1052 if (ip->i_mode == 0 || 1053 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) { 1054 vput(nvp); 1055 *vpp = NULLVP; 1056 return (ESTALE); 1057 } 1058 *vpp = nvp; 1059 vnode_create_vobject(*vpp, 0, curthread); 1060 return (0); 1061 } 1062 1063 /* 1064 * Write a superblock and associated information back to disk. 1065 */ 1066 static int 1067 ext2_sbupdate(struct ext2mount *mp, int waitfor) 1068 { 1069 struct m_ext2fs *fs = mp->um_e2fs; 1070 struct ext2fs *es = fs->e2fs; 1071 struct buf *bp; 1072 int error = 0; 1073 1074 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0); 1075 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs)); 1076 if (waitfor == MNT_WAIT) 1077 error = bwrite(bp); 1078 else 1079 bawrite(bp); 1080 1081 /* 1082 * The buffers for group descriptors, inode bitmaps and block bitmaps 1083 * are not busy at this point and are (hopefully) written by the 1084 * usual sync mechanism. No need to write them here. 1085 */ 1086 return (error); 1087 } 1088 int 1089 ext2_cgupdate(struct ext2mount *mp, int waitfor) 1090 { 1091 struct m_ext2fs *fs = mp->um_e2fs; 1092 struct buf *bp; 1093 int i, error = 0, allerror = 0; 1094 1095 allerror = ext2_sbupdate(mp, waitfor); 1096 for (i = 0; i < fs->e2fs_gdbcount; i++) { 1097 bp = getblk(mp->um_devvp, fsbtodb(fs, 1098 fs->e2fs->e2fs_first_dblock + 1099 1 /* superblock */ + i), fs->e2fs_bsize, 0, 0, 0); 1100 e2fs_cgsave(&fs->e2fs_gd[ 1101 i * fs->e2fs_bsize / sizeof(struct ext2_gd)], 1102 (struct ext2_gd *)bp->b_data, fs->e2fs_bsize); 1103 if (waitfor == MNT_WAIT) 1104 error = bwrite(bp); 1105 else 1106 bawrite(bp); 1107 } 1108 1109 if (!allerror && error) 1110 allerror = error; 1111 return (allerror); 1112 } 1113 /* 1114 * Return the root of a filesystem. 1115 */ 1116 static int 1117 ext2_root(struct mount *mp, int flags, struct vnode **vpp) 1118 { 1119 struct vnode *nvp; 1120 int error; 1121 1122 error = VFS_VGET(mp, EXT2_ROOTINO, LK_EXCLUSIVE, &nvp); 1123 if (error) 1124 return (error); 1125 *vpp = nvp; 1126 return (0); 1127 } 1128