| 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 | * 3. All advertising materials mentioning features or use of this software |
| 20 | * must display the following acknowledgement: |
| 21 | * This product includes software developed by the University of |
| 22 | * California, Berkeley and its contributors. |
| 23 | * 4. Neither the name of the University nor the names of its contributors |
| 24 | * may be used to endorse or promote products derived from this software |
| 25 | * without specific prior written permission. |
| 26 | * |
| 27 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
| 28 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 29 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 30 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
| 31 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 32 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 33 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 34 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 35 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 36 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 37 | * SUCH DAMAGE. |
| 38 | * |
| 39 | * @(#)ffs_vfsops.c 8.8 (Berkeley) 4/18/94 |
| 40 | * $FreeBSD: src/sys/gnu/ext2fs/ext2_vfsops.c,v 1.63.2.7 2002/07/01 00:18:51 iedowse Exp $ |
| 41 | * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_vfsops.c,v 1.52 2006/12/23 00:41:29 swildner Exp $ |
| 42 | */ |
| 43 | |
| 44 | #include "opt_quota.h" |
| 45 | |
| 46 | #include <sys/param.h> |
| 47 | #include <sys/systm.h> |
| 48 | #include <sys/nlookup.h> |
| 49 | #include <sys/proc.h> |
| 50 | #include <sys/kernel.h> |
| 51 | #include <sys/vnode.h> |
| 52 | #include <sys/mount.h> |
| 53 | #include <sys/buf.h> |
| 54 | #include <sys/conf.h> |
| 55 | #include <sys/fcntl.h> |
| 56 | #include <sys/disklabel.h> |
| 57 | #include <sys/malloc.h> |
| 58 | #include <sys/stat.h> |
| 59 | #include <vm/vm_zone.h> |
| 60 | |
| 61 | #include <sys/buf2.h> |
| 62 | #include <sys/thread2.h> |
| 63 | |
| 64 | #include "quota.h" |
| 65 | #include "dinode.h" |
| 66 | #include "inode.h" |
| 67 | #include "ext2mount.h" |
| 68 | #include "ext2_extern.h" |
| 69 | |
| 70 | #include "fs.h" |
| 71 | #include "ext2_fs.h" |
| 72 | #include "ext2_fs_sb.h" |
| 73 | |
| 74 | extern struct vop_ops ext2_vnode_vops; |
| 75 | extern struct vop_ops ext2_spec_vops; |
| 76 | extern struct vop_ops ext2_fifo_vops; |
| 77 | |
| 78 | static int ext2_fhtovp (struct mount *, struct fid *, struct vnode **); |
| 79 | static int ext2_flushfiles (struct mount *mp, int flags); |
| 80 | static int ext2_mount (struct mount *, char *, caddr_t, struct ucred *); |
| 81 | static int ext2_mountfs (struct vnode *, struct mount *, struct ucred *); |
| 82 | static int ext2_root(struct mount *, struct vnode **); |
| 83 | static int ext2_reload (struct mount *mountp, struct ucred *cred); |
| 84 | static int ext2_sbupdate (struct ext2mount *, int); |
| 85 | static int ext2_statfs (struct mount *, struct statfs *, struct ucred *); |
| 86 | static int ext2_sync (struct mount *, int); |
| 87 | static int ext2_unmount (struct mount *, int); |
| 88 | static int ext2_vget (struct mount *, ino_t, struct vnode **); |
| 89 | static int ext2_init(struct vfsconf *); |
| 90 | static int ext2_vptofh (struct vnode *, struct fid *); |
| 91 | |
| 92 | static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part"); |
| 93 | MALLOC_DEFINE(M_EXT2MNT, "EXT2 mount", "EXT2 mount structure"); |
| 94 | |
| 95 | static struct vfsops ext2fs_vfsops = { |
| 96 | .vfs_mount = ext2_mount, |
| 97 | .vfs_unmount = ext2_unmount, |
| 98 | .vfs_root = ext2_root, /* root inode via vget */ |
| 99 | .vfs_quotactl = ext2_quotactl, /* quota operations */ |
| 100 | .vfs_statfs = ext2_statfs, |
| 101 | .vfs_sync = ext2_sync, |
| 102 | .vfs_vget = ext2_vget, |
| 103 | .vfs_fhtovp = ext2_fhtovp, |
| 104 | .vfs_checkexp = ext2_check_export, |
| 105 | .vfs_vptofh = ext2_vptofh, |
| 106 | .vfs_init = ext2_init, |
| 107 | .vfs_uninit = ext2_uninit |
| 108 | }; |
| 109 | |
| 110 | VFS_SET(ext2fs_vfsops, ext2fs, 0); |
| 111 | #define bsd_malloc kmalloc |
| 112 | #define bsd_free kfree |
| 113 | |
| 114 | static int ext2fs_inode_hash_lock; |
| 115 | |
| 116 | static int ext2_check_sb_compat (struct ext2_super_block *es, |
| 117 | cdev_t dev, int ronly); |
| 118 | static int compute_sb_data (struct vnode * devvp, |
| 119 | struct ext2_super_block * es, |
| 120 | struct ext2_sb_info * fs); |
| 121 | |
| 122 | static int |
| 123 | ext2_root(struct mount *mp, struct vnode **vpp) |
| 124 | { |
| 125 | struct vnode *nvp; |
| 126 | int error; |
| 127 | |
| 128 | error = VFS_VGET(mp, (ino_t)ROOTINO, &nvp); |
| 129 | if (error) |
| 130 | return (error); |
| 131 | *vpp = nvp; |
| 132 | return (0); |
| 133 | } |
| 134 | |
| 135 | /* |
| 136 | * Do operations associated with quotas |
| 137 | */ |
| 138 | int |
| 139 | ext2_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t arg, |
| 140 | struct ucred *cred) |
| 141 | { |
| 142 | #ifndef QUOTA |
| 143 | return (EOPNOTSUPP); |
| 144 | #else |
| 145 | int cmd, type, error; |
| 146 | |
| 147 | type = cmds & SUBCMDMASK; |
| 148 | cmd = cmds >> SUBCMDSHIFT; |
| 149 | |
| 150 | if (uid == -1) { |
| 151 | switch(type) { |
| 152 | case USRQUOTA: |
| 153 | uid = cred->cr_ruid; |
| 154 | break; |
| 155 | case GRPQUOTA: |
| 156 | uid = cred->cr_rgid; |
| 157 | break; |
| 158 | default: |
| 159 | return (EINVAL); |
| 160 | } |
| 161 | } |
| 162 | |
| 163 | switch (cmd) { |
| 164 | case Q_SYNC: |
| 165 | break; |
| 166 | case Q_GETQUOTA: |
| 167 | if (uid == cred->cr_ruid) |
| 168 | break; |
| 169 | /* fall through */ |
| 170 | default: |
| 171 | if ((error = suser_cred(cred, PRISON_ROOT)) != 0) |
| 172 | return (error); |
| 173 | } |
| 174 | |
| 175 | type = cmds & SUBCMDMASK; |
| 176 | if ((uint)type >= MAXQUOTAS) |
| 177 | return (EINVAL); |
| 178 | if (vfs_busy(mp, LK_NOWAIT)) |
| 179 | return (0); |
| 180 | |
| 181 | switch (cmd) { |
| 182 | |
| 183 | case Q_QUOTAON: |
| 184 | error = ext2_quotaon(cred, mp, type, arg); |
| 185 | break; |
| 186 | |
| 187 | case Q_QUOTAOFF: |
| 188 | error = ext2_quotaoff(mp, type); |
| 189 | break; |
| 190 | |
| 191 | case Q_SETQUOTA: |
| 192 | error = ext2_setquota(mp, uid, type, arg); |
| 193 | break; |
| 194 | |
| 195 | case Q_SETUSE: |
| 196 | error = ext2_setuse(mp, uid, type, arg); |
| 197 | break; |
| 198 | |
| 199 | case Q_GETQUOTA: |
| 200 | error = ext2_getquota(mp, uid, type, arg); |
| 201 | break; |
| 202 | |
| 203 | case Q_SYNC: |
| 204 | error = ext2_qsync(mp); |
| 205 | break; |
| 206 | |
| 207 | default: |
| 208 | error = EINVAL; |
| 209 | break; |
| 210 | } |
| 211 | vfs_unbusy(mp); |
| 212 | return (error); |
| 213 | #endif |
| 214 | } |
| 215 | |
| 216 | /* |
| 217 | * Initial UFS filesystems, done only once. |
| 218 | */ |
| 219 | int |
| 220 | ext2_init(struct vfsconf *vfsp) |
| 221 | { |
| 222 | static int done; |
| 223 | |
| 224 | if (done) |
| 225 | return (0); |
| 226 | done = 1; |
| 227 | ext2_ihashinit(); |
| 228 | #ifdef QUOTA |
| 229 | ext2_dqinit(); |
| 230 | #endif |
| 231 | return (0); |
| 232 | } |
| 233 | |
| 234 | /* |
| 235 | * VFS Operations. |
| 236 | * |
| 237 | * mount system call |
| 238 | * |
| 239 | * Parameters: |
| 240 | * data: this is actually a (struct ext2_args *) |
| 241 | */ |
| 242 | static int |
| 243 | ext2_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred) |
| 244 | { |
| 245 | struct vnode *devvp; |
| 246 | struct ext2_args args; |
| 247 | struct ext2mount *ump = 0; |
| 248 | struct ext2_sb_info *fs; |
| 249 | size_t size; |
| 250 | int error, flags; |
| 251 | mode_t accessmode; |
| 252 | struct nlookupdata nd; |
| 253 | |
| 254 | if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0) |
| 255 | return (error); |
| 256 | |
| 257 | /* |
| 258 | * If updating, check whether changing from read-only to |
| 259 | * read/write; if there is no device name, that's all we do. |
| 260 | */ |
| 261 | if (mp->mnt_flag & MNT_UPDATE) { |
| 262 | ump = VFSTOEXT2(mp); |
| 263 | fs = ump->um_e2fs; |
| 264 | devvp = ump->um_devvp; |
| 265 | error = 0; |
| 266 | if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) { |
| 267 | flags = WRITECLOSE; |
| 268 | if (mp->mnt_flag & MNT_FORCE) |
| 269 | flags |= FORCECLOSE; |
| 270 | if (vfs_busy(mp, LK_NOWAIT)) |
| 271 | return (EBUSY); |
| 272 | error = ext2_flushfiles(mp, flags); |
| 273 | vfs_unbusy(mp); |
| 274 | if (!error && fs->s_wasvalid) { |
| 275 | fs->s_es->s_state |= EXT2_VALID_FS; |
| 276 | ext2_sbupdate(ump, MNT_WAIT); |
| 277 | } |
| 278 | fs->s_rd_only = 1; |
| 279 | vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); |
| 280 | VOP_OPEN(devvp, FREAD, FSCRED, NULL); |
| 281 | VOP_CLOSE(devvp, FREAD|FWRITE); |
| 282 | vn_unlock(devvp); |
| 283 | } |
| 284 | if (!error && (mp->mnt_flag & MNT_RELOAD)) |
| 285 | error = ext2_reload(mp, cred); |
| 286 | if (error) |
| 287 | return (error); |
| 288 | if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev, |
| 289 | (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0) |
| 290 | return (EPERM); |
| 291 | if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { |
| 292 | /* |
| 293 | * If upgrade to read-write by non-root, then verify |
| 294 | * that user has necessary permissions on the device. |
| 295 | */ |
| 296 | if (cred->cr_uid != 0) { |
| 297 | vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); |
| 298 | error = VOP_ACCESS(devvp, VREAD | VWRITE, cred); |
| 299 | if (error) { |
| 300 | vn_unlock(devvp); |
| 301 | return (error); |
| 302 | } |
| 303 | vn_unlock(devvp); |
| 304 | } |
| 305 | |
| 306 | if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 || |
| 307 | (fs->s_es->s_state & EXT2_ERROR_FS)) { |
| 308 | if (mp->mnt_flag & MNT_FORCE) { |
| 309 | kprintf( |
| 310 | "WARNING: %s was not properly dismounted\n", |
| 311 | fs->fs_fsmnt); |
| 312 | } else { |
| 313 | kprintf( |
| 314 | "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", |
| 315 | fs->fs_fsmnt); |
| 316 | return (EPERM); |
| 317 | } |
| 318 | } |
| 319 | fs->s_es->s_state &= ~EXT2_VALID_FS; |
| 320 | ext2_sbupdate(ump, MNT_WAIT); |
| 321 | fs->s_rd_only = 0; |
| 322 | vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); |
| 323 | VOP_OPEN(devvp, FREAD|FWRITE, FSCRED, NULL); |
| 324 | VOP_CLOSE(devvp, FREAD); |
| 325 | vn_unlock(devvp); |
| 326 | } |
| 327 | if (args.fspec == 0) { |
| 328 | /* |
| 329 | * Process export requests. |
| 330 | */ |
| 331 | return (vfs_export(mp, &ump->um_export, &args.export)); |
| 332 | } |
| 333 | } |
| 334 | /* |
| 335 | * Not an update, or updating the name: look up the name |
| 336 | * and verify that it refers to a sensible block device. |
| 337 | */ |
| 338 | devvp = NULL; |
| 339 | error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW); |
| 340 | if (error == 0) |
| 341 | error = nlookup(&nd); |
| 342 | if (error == 0) |
| 343 | error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp); |
| 344 | nlookup_done(&nd); |
| 345 | if (error) |
| 346 | return (error); |
| 347 | |
| 348 | if (!vn_isdisk(devvp, &error)) { |
| 349 | vrele(devvp); |
| 350 | return (error); |
| 351 | } |
| 352 | |
| 353 | /* |
| 354 | * If mount by non-root, then verify that user has necessary |
| 355 | * permissions on the device. |
| 356 | */ |
| 357 | if (cred->cr_uid != 0) { |
| 358 | accessmode = VREAD; |
| 359 | if ((mp->mnt_flag & MNT_RDONLY) == 0) |
| 360 | accessmode |= VWRITE; |
| 361 | vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); |
| 362 | if ((error = VOP_ACCESS(devvp, accessmode, cred)) != 0) { |
| 363 | vput(devvp); |
| 364 | return (error); |
| 365 | } |
| 366 | vn_unlock(devvp); |
| 367 | } |
| 368 | |
| 369 | if ((mp->mnt_flag & MNT_UPDATE) == 0) { |
| 370 | error = ext2_mountfs(devvp, mp, cred); |
| 371 | } else { |
| 372 | if (devvp != ump->um_devvp) |
| 373 | error = EINVAL; /* needs translation */ |
| 374 | else |
| 375 | vrele(devvp); |
| 376 | } |
| 377 | if (error) { |
| 378 | vrele(devvp); |
| 379 | return (error); |
| 380 | } |
| 381 | ump = VFSTOEXT2(mp); |
| 382 | fs = ump->um_e2fs; |
| 383 | copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size); |
| 384 | bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size); |
| 385 | copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size); |
| 386 | bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); |
| 387 | ext2_statfs(mp, &mp->mnt_stat, cred); |
| 388 | return (0); |
| 389 | } |
| 390 | |
| 391 | /* |
| 392 | * checks that the data in the descriptor blocks make sense |
| 393 | * this is taken from ext2/super.c |
| 394 | */ |
| 395 | static int |
| 396 | ext2_check_descriptors(struct ext2_sb_info *sb) |
| 397 | { |
| 398 | int i; |
| 399 | int desc_block = 0; |
| 400 | unsigned long block = sb->s_es->s_first_data_block; |
| 401 | struct ext2_group_desc * gdp = NULL; |
| 402 | |
| 403 | /* ext2_debug ("Checking group descriptors"); */ |
| 404 | |
| 405 | for (i = 0; i < sb->s_groups_count; i++) |
| 406 | { |
| 407 | /* examine next descriptor block */ |
| 408 | if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0) |
| 409 | gdp = (struct ext2_group_desc *) |
| 410 | sb->s_group_desc[desc_block++]->b_data; |
| 411 | if (gdp->bg_block_bitmap < block || |
| 412 | gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) |
| 413 | { |
| 414 | kprintf ("ext2_check_descriptors: " |
| 415 | "Block bitmap for group %d" |
| 416 | " not in group (block %lu)!\n", |
| 417 | i, (unsigned long) gdp->bg_block_bitmap); |
| 418 | return 0; |
| 419 | } |
| 420 | if (gdp->bg_inode_bitmap < block || |
| 421 | gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) |
| 422 | { |
| 423 | kprintf ("ext2_check_descriptors: " |
| 424 | "Inode bitmap for group %d" |
| 425 | " not in group (block %lu)!\n", |
| 426 | i, (unsigned long) gdp->bg_inode_bitmap); |
| 427 | return 0; |
| 428 | } |
| 429 | if (gdp->bg_inode_table < block || |
| 430 | gdp->bg_inode_table + sb->s_itb_per_group >= |
| 431 | block + EXT2_BLOCKS_PER_GROUP(sb)) |
| 432 | { |
| 433 | kprintf ("ext2_check_descriptors: " |
| 434 | "Inode table for group %d" |
| 435 | " not in group (block %lu)!\n", |
| 436 | i, (unsigned long) gdp->bg_inode_table); |
| 437 | return 0; |
| 438 | } |
| 439 | block += EXT2_BLOCKS_PER_GROUP(sb); |
| 440 | gdp++; |
| 441 | } |
| 442 | return 1; |
| 443 | } |
| 444 | |
| 445 | static int |
| 446 | ext2_check_sb_compat(struct ext2_super_block *es, cdev_t dev, int ronly) |
| 447 | { |
| 448 | if (es->s_magic != EXT2_SUPER_MAGIC) { |
| 449 | kprintf("ext2fs: %s: wrong magic number %#x (expected %#x)\n", |
| 450 | devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC); |
| 451 | return (1); |
| 452 | } |
| 453 | if (es->s_rev_level > EXT2_GOOD_OLD_REV) { |
| 454 | if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) { |
| 455 | kprintf( |
| 456 | "WARNING: mount of %s denied due to unsupported optional features\n", |
| 457 | devtoname(dev)); |
| 458 | return (1); |
| 459 | } |
| 460 | if (!ronly && |
| 461 | (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) { |
| 462 | kprintf( |
| 463 | "WARNING: R/W mount of %s denied due to unsupported optional features\n", |
| 464 | devtoname(dev)); |
| 465 | return (1); |
| 466 | } |
| 467 | } |
| 468 | return (0); |
| 469 | } |
| 470 | |
| 471 | /* |
| 472 | * this computes the fields of the ext2_sb_info structure from the |
| 473 | * data in the ext2_super_block structure read in |
| 474 | */ |
| 475 | static int |
| 476 | compute_sb_data(struct vnode *devvp, struct ext2_super_block *es, |
| 477 | struct ext2_sb_info *fs) |
| 478 | { |
| 479 | int db_count, error; |
| 480 | int i, j; |
| 481 | int logic_sb_block = 1; /* XXX for now */ |
| 482 | |
| 483 | #if 1 |
| 484 | #define V(v) |
| 485 | #else |
| 486 | #define V(v) kprintf(#v"= %d\n", fs->v); |
| 487 | #endif |
| 488 | |
| 489 | fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size; |
| 490 | V(s_blocksize) |
| 491 | fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size; |
| 492 | V(s_bshift) |
| 493 | fs->s_fsbtodb = es->s_log_block_size + 1; |
| 494 | V(s_fsbtodb) |
| 495 | fs->s_qbmask = fs->s_blocksize - 1; |
| 496 | V(s_bmask) |
| 497 | fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es); |
| 498 | V(s_blocksize_bits) |
| 499 | fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size; |
| 500 | V(s_frag_size) |
| 501 | if (fs->s_frag_size) |
| 502 | fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size; |
| 503 | V(s_frags_per_block) |
| 504 | fs->s_blocks_per_group = es->s_blocks_per_group; |
| 505 | V(s_blocks_per_group) |
| 506 | fs->s_frags_per_group = es->s_frags_per_group; |
| 507 | V(s_frags_per_group) |
| 508 | fs->s_inodes_per_group = es->s_inodes_per_group; |
| 509 | V(s_inodes_per_group) |
| 510 | fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE; |
| 511 | V(s_inodes_per_block) |
| 512 | fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block; |
| 513 | V(s_itb_per_group) |
| 514 | fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc); |
| 515 | V(s_desc_per_block) |
| 516 | /* s_resuid / s_resgid ? */ |
| 517 | fs->s_groups_count = (es->s_blocks_count - |
| 518 | es->s_first_data_block + |
| 519 | EXT2_BLOCKS_PER_GROUP(fs) - 1) / |
| 520 | EXT2_BLOCKS_PER_GROUP(fs); |
| 521 | V(s_groups_count) |
| 522 | db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) / |
| 523 | EXT2_DESC_PER_BLOCK(fs); |
| 524 | fs->s_db_per_group = db_count; |
| 525 | V(s_db_per_group) |
| 526 | |
| 527 | fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *), |
| 528 | M_EXT2MNT, M_WAITOK); |
| 529 | |
| 530 | /* adjust logic_sb_block */ |
| 531 | if(fs->s_blocksize > SBSIZE) |
| 532 | /* Godmar thinks: if the blocksize is greater than 1024, then |
| 533 | the superblock is logically part of block zero. |
| 534 | */ |
| 535 | logic_sb_block = 0; |
| 536 | |
| 537 | for (i = 0; i < db_count; i++) { |
| 538 | error = bread(devvp, fsbtodoff(fs, logic_sb_block + i + 1), |
| 539 | fs->s_blocksize, &fs->s_group_desc[i]); |
| 540 | if(error) { |
| 541 | for (j = 0; j < i; j++) |
| 542 | brelse(fs->s_group_desc[j]); |
| 543 | bsd_free(fs->s_group_desc, M_EXT2MNT); |
| 544 | kprintf("EXT2-fs: unable to read group descriptors (%d)\n", error); |
| 545 | return EIO; |
| 546 | } |
| 547 | /* Set the B_LOCKED flag on the buffer, then brelse() it */ |
| 548 | LCK_BUF(fs->s_group_desc[i]) |
| 549 | } |
| 550 | if(!ext2_check_descriptors(fs)) { |
| 551 | for (j = 0; j < db_count; j++) |
| 552 | ULCK_BUF(fs->s_group_desc[j]) |
| 553 | bsd_free(fs->s_group_desc, M_EXT2MNT); |
| 554 | kprintf("EXT2-fs: (ext2_check_descriptors failure) " |
| 555 | "unable to read group descriptors\n"); |
| 556 | return EIO; |
| 557 | } |
| 558 | |
| 559 | for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) { |
| 560 | fs->s_inode_bitmap_number[i] = 0; |
| 561 | fs->s_inode_bitmap[i] = NULL; |
| 562 | fs->s_block_bitmap_number[i] = 0; |
| 563 | fs->s_block_bitmap[i] = NULL; |
| 564 | } |
| 565 | fs->s_loaded_inode_bitmaps = 0; |
| 566 | fs->s_loaded_block_bitmaps = 0; |
| 567 | return 0; |
| 568 | } |
| 569 | |
| 570 | /* |
| 571 | * Reload all incore data for a filesystem (used after running fsck on |
| 572 | * the root filesystem and finding things to fix). The filesystem must |
| 573 | * be mounted read-only. |
| 574 | * |
| 575 | * Things to do to update the mount: |
| 576 | * 1) invalidate all cached meta-data. |
| 577 | * 2) re-read superblock from disk. |
| 578 | * 3) re-read summary information from disk. |
| 579 | * 4) invalidate all inactive vnodes. |
| 580 | * 5) invalidate all cached file data. |
| 581 | * 6) re-read inode data for all active vnodes. |
| 582 | */ |
| 583 | static int ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *rescan); |
| 584 | static int ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *rescan); |
| 585 | |
| 586 | struct scaninfo { |
| 587 | int rescan; |
| 588 | int allerror; |
| 589 | int waitfor; |
| 590 | struct vnode *devvp; |
| 591 | struct ext2_sb_info *fs; |
| 592 | }; |
| 593 | |
| 594 | static int |
| 595 | ext2_reload(struct mount *mountp, struct ucred *cred) |
| 596 | { |
| 597 | struct vnode *devvp; |
| 598 | struct buf *bp; |
| 599 | struct ext2_super_block * es; |
| 600 | struct ext2_sb_info *fs; |
| 601 | int error; |
| 602 | struct scaninfo scaninfo; |
| 603 | |
| 604 | if ((mountp->mnt_flag & MNT_RDONLY) == 0) |
| 605 | return (EINVAL); |
| 606 | /* |
| 607 | * Step 1: invalidate all cached meta-data. |
| 608 | */ |
| 609 | devvp = VFSTOEXT2(mountp)->um_devvp; |
| 610 | if (vinvalbuf(devvp, 0, 0, 0)) |
| 611 | panic("ext2_reload: dirty1"); |
| 612 | /* |
| 613 | * Step 2: re-read superblock from disk. |
| 614 | * constants have been adjusted for ext2 |
| 615 | */ |
| 616 | if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0) |
| 617 | return (error); |
| 618 | es = (struct ext2_super_block *)bp->b_data; |
| 619 | if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) { |
| 620 | brelse(bp); |
| 621 | return (EIO); /* XXX needs translation */ |
| 622 | } |
| 623 | fs = VFSTOEXT2(mountp)->um_e2fs; |
| 624 | bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block)); |
| 625 | |
| 626 | if((error = compute_sb_data(devvp, es, fs)) != 0) { |
| 627 | brelse(bp); |
| 628 | return error; |
| 629 | } |
| 630 | #ifdef UNKLAR |
| 631 | if (fs->fs_sbsize < SBSIZE) |
| 632 | bp->b_flags |= B_INVAL; |
| 633 | #endif |
| 634 | brelse(bp); |
| 635 | |
| 636 | scaninfo.rescan = 1; |
| 637 | scaninfo.devvp = devvp; |
| 638 | scaninfo.fs = fs; |
| 639 | while (error == 0 && scaninfo.rescan) { |
| 640 | scaninfo.rescan = 0; |
| 641 | error = vmntvnodescan(mountp, VMSC_GETVX, ext2_reload_scan1, |
| 642 | ext2_reload_scan2, &scaninfo); |
| 643 | } |
| 644 | return(error); |
| 645 | } |
| 646 | |
| 647 | static int |
| 648 | ext2_reload_scan1(struct mount *mp, struct vnode *vp, void *data) |
| 649 | { |
| 650 | /*struct scaninfo *info = data;*/ |
| 651 | |
| 652 | return(0); |
| 653 | } |
| 654 | |
| 655 | static int |
| 656 | ext2_reload_scan2(struct mount *mp, struct vnode *vp, void *data) |
| 657 | { |
| 658 | struct scaninfo *info = data; |
| 659 | struct inode *ip; |
| 660 | struct buf *bp; |
| 661 | int error; |
| 662 | |
| 663 | /* |
| 664 | * Try to recycle |
| 665 | */ |
| 666 | if (vrecycle(vp)) |
| 667 | return(0); |
| 668 | |
| 669 | /* |
| 670 | * Step 5: invalidate all cached file data. |
| 671 | */ |
| 672 | if (vinvalbuf(vp, 0, 0, 0)) |
| 673 | panic("ext2_reload: dirty2"); |
| 674 | /* |
| 675 | * Step 6: re-read inode data for all active vnodes. |
| 676 | */ |
| 677 | ip = VTOI(vp); |
| 678 | error = bread(info->devvp, |
| 679 | fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)), |
| 680 | (int)info->fs->s_blocksize, &bp); |
| 681 | if (error) |
| 682 | return (error); |
| 683 | ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + |
| 684 | EXT2_INODE_SIZE * ino_to_fsbo(info->fs, ip->i_number)), |
| 685 | &ip->i_din); |
| 686 | brelse(bp); |
| 687 | return(0); |
| 688 | } |
| 689 | |
| 690 | /* |
| 691 | * Common code for mount and mountroot |
| 692 | */ |
| 693 | static int |
| 694 | ext2_mountfs(struct vnode *devvp, struct mount *mp, struct ucred *cred) |
| 695 | { |
| 696 | struct ext2mount *ump; |
| 697 | struct buf *bp; |
| 698 | struct ext2_sb_info *fs; |
| 699 | struct ext2_super_block * es; |
| 700 | cdev_t dev; |
| 701 | struct partinfo dpart; |
| 702 | int havepart = 0; |
| 703 | int error, i, size; |
| 704 | int ronly; |
| 705 | |
| 706 | /* |
| 707 | * Disallow multiple mounts of the same device. |
| 708 | * Disallow mounting of a device that is currently in use |
| 709 | * (except for root, which might share swap device for miniroot). |
| 710 | * Flush out any old buffers remaining from a previous use. |
| 711 | */ |
| 712 | if ((error = vfs_mountedon(devvp)) != 0) |
| 713 | return (error); |
| 714 | if (count_udev(devvp->v_udev) > 0) |
| 715 | return (EBUSY); |
| 716 | if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0) |
| 717 | return (error); |
| 718 | #ifdef READONLY |
| 719 | /* turn on this to force it to be read-only */ |
| 720 | mp->mnt_flag |= MNT_RDONLY; |
| 721 | #endif |
| 722 | |
| 723 | ronly = (mp->mnt_flag & MNT_RDONLY) != 0; |
| 724 | vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); |
| 725 | error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL); |
| 726 | vn_unlock(devvp); |
| 727 | if (error) |
| 728 | return (error); |
| 729 | dev = devvp->v_rdev; |
| 730 | if (dev->si_iosize_max != 0) |
| 731 | mp->mnt_iosize_max = dev->si_iosize_max; |
| 732 | if (mp->mnt_iosize_max > MAXPHYS) |
| 733 | mp->mnt_iosize_max = MAXPHYS; |
| 734 | if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred) != 0) |
| 735 | size = DEV_BSIZE; |
| 736 | else { |
| 737 | havepart = 1; |
| 738 | size = dpart.disklab->d_secsize; |
| 739 | } |
| 740 | |
| 741 | bp = NULL; |
| 742 | ump = NULL; |
| 743 | if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0) |
| 744 | goto out; |
| 745 | es = (struct ext2_super_block *)bp->b_data; |
| 746 | if (ext2_check_sb_compat(es, dev, ronly) != 0) { |
| 747 | error = EINVAL; /* XXX needs translation */ |
| 748 | goto out; |
| 749 | } |
| 750 | if ((es->s_state & EXT2_VALID_FS) == 0 || |
| 751 | (es->s_state & EXT2_ERROR_FS)) { |
| 752 | if (ronly || (mp->mnt_flag & MNT_FORCE)) { |
| 753 | kprintf( |
| 754 | "WARNING: Filesystem was not properly dismounted\n"); |
| 755 | } else { |
| 756 | kprintf( |
| 757 | "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n"); |
| 758 | error = EPERM; |
| 759 | goto out; |
| 760 | } |
| 761 | } |
| 762 | ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK); |
| 763 | bzero((caddr_t)ump, sizeof *ump); |
| 764 | ump->um_malloctype = M_EXT2NODE; |
| 765 | ump->um_blkatoff = ext2_blkatoff; |
| 766 | ump->um_truncate = ext2_truncate; |
| 767 | ump->um_update = ext2_update; |
| 768 | ump->um_valloc = ext2_valloc; |
| 769 | ump->um_vfree = ext2_vfree; |
| 770 | /* I don't know whether this is the right strategy. Note that |
| 771 | we dynamically allocate both a ext2_sb_info and a ext2_super_block |
| 772 | while Linux keeps the super block in a locked buffer |
| 773 | */ |
| 774 | ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info), |
| 775 | M_EXT2MNT, M_WAITOK); |
| 776 | ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block), |
| 777 | M_EXT2MNT, M_WAITOK); |
| 778 | bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block)); |
| 779 | if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs))) |
| 780 | goto out; |
| 781 | /* |
| 782 | * We don't free the group descriptors allocated by compute_sb_data() |
| 783 | * until ext2_unmount(). This is OK since the mount will succeed. |
| 784 | */ |
| 785 | brelse(bp); |
| 786 | bp = NULL; |
| 787 | fs = ump->um_e2fs; |
| 788 | fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */ |
| 789 | /* if the fs is not mounted read-only, make sure the super block is |
| 790 | always written back on a sync() |
| 791 | */ |
| 792 | fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0; |
| 793 | if (ronly == 0) { |
| 794 | fs->s_dirt = 1; /* mark it modified */ |
| 795 | fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */ |
| 796 | } |
| 797 | mp->mnt_data = (qaddr_t)ump; |
| 798 | mp->mnt_stat.f_fsid.val[0] = dev2udev(dev); |
| 799 | mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; |
| 800 | mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; |
| 801 | mp->mnt_flag |= MNT_LOCAL; |
| 802 | ump->um_mountp = mp; |
| 803 | ump->um_dev = dev; |
| 804 | ump->um_devvp = devvp; |
| 805 | /* setting those two parameters allows us to use |
| 806 | ext2_bmap w/o changse ! |
| 807 | */ |
| 808 | ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); |
| 809 | ump->um_bptrtodb = fs->s_es->s_log_block_size + 1; |
| 810 | ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); |
| 811 | for (i = 0; i < MAXQUOTAS; i++) |
| 812 | ump->um_quotas[i] = NULLVP; |
| 813 | dev->si_mountpoint = mp; |
| 814 | |
| 815 | vfs_add_vnodeops(mp, &ext2_vnode_vops, &mp->mnt_vn_norm_ops); |
| 816 | vfs_add_vnodeops(mp, &ext2_spec_vops, &mp->mnt_vn_spec_ops); |
| 817 | vfs_add_vnodeops(mp, &ext2_fifo_vops, &mp->mnt_vn_fifo_ops); |
| 818 | |
| 819 | if (ronly == 0) |
| 820 | ext2_sbupdate(ump, MNT_WAIT); |
| 821 | return (0); |
| 822 | out: |
| 823 | if (bp) |
| 824 | brelse(bp); |
| 825 | VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE); |
| 826 | if (ump) { |
| 827 | bsd_free(ump->um_e2fs->s_es, M_EXT2MNT); |
| 828 | bsd_free(ump->um_e2fs, M_EXT2MNT); |
| 829 | bsd_free(ump, M_EXT2MNT); |
| 830 | mp->mnt_data = (qaddr_t)0; |
| 831 | } |
| 832 | return (error); |
| 833 | } |
| 834 | |
| 835 | /* |
| 836 | * unmount system call |
| 837 | */ |
| 838 | static int |
| 839 | ext2_unmount(struct mount *mp, int mntflags) |
| 840 | { |
| 841 | struct ext2mount *ump; |
| 842 | struct ext2_sb_info *fs; |
| 843 | int error, flags, ronly, i; |
| 844 | |
| 845 | flags = 0; |
| 846 | if (mntflags & MNT_FORCE) { |
| 847 | if (mp->mnt_flag & MNT_ROOTFS) |
| 848 | return (EINVAL); |
| 849 | flags |= FORCECLOSE; |
| 850 | } |
| 851 | if ((error = ext2_flushfiles(mp, flags)) != 0) |
| 852 | return (error); |
| 853 | ump = VFSTOEXT2(mp); |
| 854 | fs = ump->um_e2fs; |
| 855 | ronly = fs->s_rd_only; |
| 856 | if (ronly == 0) { |
| 857 | if (fs->s_wasvalid) |
| 858 | fs->s_es->s_state |= EXT2_VALID_FS; |
| 859 | ext2_sbupdate(ump, MNT_WAIT); |
| 860 | } |
| 861 | |
| 862 | /* release buffers containing group descriptors */ |
| 863 | for(i = 0; i < fs->s_db_per_group; i++) |
| 864 | ULCK_BUF(fs->s_group_desc[i]) |
| 865 | bsd_free(fs->s_group_desc, M_EXT2MNT); |
| 866 | |
| 867 | /* release cached inode/block bitmaps */ |
| 868 | for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) |
| 869 | if (fs->s_inode_bitmap[i]) |
| 870 | ULCK_BUF(fs->s_inode_bitmap[i]) |
| 871 | |
| 872 | for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) |
| 873 | if (fs->s_block_bitmap[i]) |
| 874 | ULCK_BUF(fs->s_block_bitmap[i]) |
| 875 | |
| 876 | ump->um_devvp->v_rdev->si_mountpoint = NULL; |
| 877 | error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE); |
| 878 | vrele(ump->um_devvp); |
| 879 | bsd_free(fs->s_es, M_EXT2MNT); |
| 880 | bsd_free(fs, M_EXT2MNT); |
| 881 | bsd_free(ump, M_EXT2MNT); |
| 882 | mp->mnt_data = (qaddr_t)0; |
| 883 | mp->mnt_flag &= ~MNT_LOCAL; |
| 884 | return (error); |
| 885 | } |
| 886 | |
| 887 | /* |
| 888 | * Flush out all the files in a filesystem. |
| 889 | */ |
| 890 | static int |
| 891 | ext2_flushfiles(struct mount *mp, int flags) |
| 892 | { |
| 893 | struct ext2mount *ump; |
| 894 | int error; |
| 895 | #if QUOTA |
| 896 | int i; |
| 897 | #endif |
| 898 | |
| 899 | ump = VFSTOEXT2(mp); |
| 900 | #if QUOTA |
| 901 | if (mp->mnt_flag & MNT_QUOTA) { |
| 902 | if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0) |
| 903 | return (error); |
| 904 | for (i = 0; i < MAXQUOTAS; i++) { |
| 905 | if (ump->um_quotas[i] == NULLVP) |
| 906 | continue; |
| 907 | ext2_quotaoff(mp, i); |
| 908 | } |
| 909 | /* |
| 910 | * Here we fall through to vflush again to ensure |
| 911 | * that we have gotten rid of all the system vnodes. |
| 912 | */ |
| 913 | } |
| 914 | #endif |
| 915 | error = vflush(mp, 0, flags); |
| 916 | return (error); |
| 917 | } |
| 918 | |
| 919 | /* |
| 920 | * Get file system statistics. |
| 921 | * taken from ext2/super.c ext2_statfs |
| 922 | */ |
| 923 | static int |
| 924 | ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred) |
| 925 | { |
| 926 | unsigned long overhead; |
| 927 | struct ext2mount *ump; |
| 928 | struct ext2_sb_info *fs; |
| 929 | struct ext2_super_block *es; |
| 930 | int i, nsb; |
| 931 | |
| 932 | ump = VFSTOEXT2(mp); |
| 933 | fs = ump->um_e2fs; |
| 934 | es = fs->s_es; |
| 935 | |
| 936 | if (es->s_magic != EXT2_SUPER_MAGIC) |
| 937 | panic("ext2_statfs - magic number spoiled"); |
| 938 | |
| 939 | /* |
| 940 | * Compute the overhead (FS structures) |
| 941 | */ |
| 942 | if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) { |
| 943 | nsb = 0; |
| 944 | for (i = 0 ; i < fs->s_groups_count; i++) |
| 945 | if (ext2_group_sparse(i)) |
| 946 | nsb++; |
| 947 | } else |
| 948 | nsb = fs->s_groups_count; |
| 949 | overhead = es->s_first_data_block + |
| 950 | /* Superblocks and block group descriptors: */ |
| 951 | nsb * (1 + fs->s_db_per_group) + |
| 952 | /* Inode bitmap, block bitmap, and inode table: */ |
| 953 | fs->s_groups_count * (1 + 1 + fs->s_itb_per_group); |
| 954 | |
| 955 | sbp->f_bsize = EXT2_FRAG_SIZE(fs); |
| 956 | sbp->f_iosize = EXT2_BLOCK_SIZE(fs); |
| 957 | sbp->f_blocks = es->s_blocks_count - overhead; |
| 958 | sbp->f_bfree = es->s_free_blocks_count; |
| 959 | sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count; |
| 960 | sbp->f_files = es->s_inodes_count; |
| 961 | sbp->f_ffree = es->s_free_inodes_count; |
| 962 | if (sbp != &mp->mnt_stat) { |
| 963 | sbp->f_type = mp->mnt_vfc->vfc_typenum; |
| 964 | bcopy((caddr_t)mp->mnt_stat.f_mntfromname, |
| 965 | (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); |
| 966 | } |
| 967 | return (0); |
| 968 | } |
| 969 | |
| 970 | /* |
| 971 | * Go through the disk queues to initiate sandbagged IO; |
| 972 | * go through the inodes to write those that have been modified; |
| 973 | * initiate the writing of the super block if it has been modified. |
| 974 | * |
| 975 | * Note: we are always called with the filesystem marked `MPBUSY'. |
| 976 | */ |
| 977 | |
| 978 | static int ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data); |
| 979 | |
| 980 | static int |
| 981 | ext2_sync(struct mount *mp, int waitfor) |
| 982 | { |
| 983 | struct ext2mount *ump = VFSTOEXT2(mp); |
| 984 | struct ext2_sb_info *fs; |
| 985 | struct scaninfo scaninfo; |
| 986 | int error; |
| 987 | |
| 988 | fs = ump->um_e2fs; |
| 989 | if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */ |
| 990 | kprintf("fs = %s\n", fs->fs_fsmnt); |
| 991 | panic("ext2_sync: rofs mod"); |
| 992 | } |
| 993 | |
| 994 | /* |
| 995 | * Write back each (modified) inode. |
| 996 | */ |
| 997 | scaninfo.allerror = 0; |
| 998 | scaninfo.rescan = 1; |
| 999 | scaninfo.waitfor = waitfor; |
| 1000 | while (scaninfo.rescan) { |
| 1001 | scaninfo.rescan = 0; |
| 1002 | vmntvnodescan(mp, VMSC_GETVP|VMSC_NOWAIT, |
| 1003 | NULL, ext2_sync_scan, &scaninfo); |
| 1004 | } |
| 1005 | |
| 1006 | /* |
| 1007 | * Force stale file system control information to be flushed. |
| 1008 | */ |
| 1009 | if (waitfor != MNT_LAZY) { |
| 1010 | vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); |
| 1011 | if ((error = VOP_FSYNC(ump->um_devvp, waitfor)) != 0) |
| 1012 | scaninfo.allerror = error; |
| 1013 | vn_unlock(ump->um_devvp); |
| 1014 | } |
| 1015 | #if QUOTA |
| 1016 | ext2_qsync(mp); |
| 1017 | #endif |
| 1018 | /* |
| 1019 | * Write back modified superblock. |
| 1020 | */ |
| 1021 | if (fs->s_dirt != 0) { |
| 1022 | fs->s_dirt = 0; |
| 1023 | fs->s_es->s_wtime = time_second; |
| 1024 | if ((error = ext2_sbupdate(ump, waitfor)) != 0) |
| 1025 | scaninfo.allerror = error; |
| 1026 | } |
| 1027 | return (scaninfo.allerror); |
| 1028 | } |
| 1029 | |
| 1030 | static int |
| 1031 | ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data) |
| 1032 | { |
| 1033 | struct scaninfo *info = data; |
| 1034 | struct inode *ip; |
| 1035 | int error; |
| 1036 | |
| 1037 | ip = VTOI(vp); |
| 1038 | if (vp->v_type == VNON || |
| 1039 | ((ip->i_flag & |
| 1040 | (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && |
| 1041 | (RB_EMPTY(&vp->v_rbdirty_tree) || info->waitfor == MNT_LAZY))) { |
| 1042 | return(0); |
| 1043 | } |
| 1044 | if ((error = VOP_FSYNC(vp, info->waitfor)) != 0) |
| 1045 | info->allerror = error; |
| 1046 | return(0); |
| 1047 | } |
| 1048 | |
| 1049 | /* |
| 1050 | * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it |
| 1051 | * in from disk. If it is in core, wait for the lock bit to clear, then |
| 1052 | * return the inode locked. Detection and handling of mount points must be |
| 1053 | * done by the calling routine. |
| 1054 | */ |
| 1055 | static int |
| 1056 | ext2_vget(struct mount *mp, ino_t ino, struct vnode **vpp) |
| 1057 | { |
| 1058 | struct ext2_sb_info *fs; |
| 1059 | struct inode *ip; |
| 1060 | struct ext2mount *ump; |
| 1061 | struct buf *bp; |
| 1062 | struct vnode *vp; |
| 1063 | cdev_t dev; |
| 1064 | int i, error; |
| 1065 | int used_blocks; |
| 1066 | |
| 1067 | ump = VFSTOEXT2(mp); |
| 1068 | dev = ump->um_dev; |
| 1069 | restart: |
| 1070 | if ((*vpp = ext2_ihashget(dev, ino)) != NULL) |
| 1071 | return (0); |
| 1072 | |
| 1073 | /* |
| 1074 | * Lock out the creation of new entries in the FFS hash table in |
| 1075 | * case getnewvnode() or MALLOC() blocks, otherwise a duplicate |
| 1076 | * may occur! |
| 1077 | */ |
| 1078 | if (ext2fs_inode_hash_lock) { |
| 1079 | while (ext2fs_inode_hash_lock) { |
| 1080 | ext2fs_inode_hash_lock = -1; |
| 1081 | tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0); |
| 1082 | } |
| 1083 | goto restart; |
| 1084 | } |
| 1085 | ext2fs_inode_hash_lock = 1; |
| 1086 | |
| 1087 | /* |
| 1088 | * If this MALLOC() is performed after the getnewvnode() |
| 1089 | * it might block, leaving a vnode with a NULL v_data to be |
| 1090 | * found by ext2_sync() if a sync happens to fire right then, |
| 1091 | * which will cause a panic because ext2_sync() blindly |
| 1092 | * dereferences vp->v_data (as well it should). |
| 1093 | */ |
| 1094 | MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK); |
| 1095 | |
| 1096 | /* Allocate a new vnode/inode. */ |
| 1097 | if ((error = getnewvnode(VT_EXT2FS, mp, &vp, 0, LK_CANRECURSE)) != 0) { |
| 1098 | if (ext2fs_inode_hash_lock < 0) |
| 1099 | wakeup(&ext2fs_inode_hash_lock); |
| 1100 | ext2fs_inode_hash_lock = 0; |
| 1101 | *vpp = NULL; |
| 1102 | FREE(ip, M_EXT2NODE); |
| 1103 | return (error); |
| 1104 | } |
| 1105 | bzero((caddr_t)ip, sizeof(struct inode)); |
| 1106 | vp->v_data = ip; |
| 1107 | ip->i_vnode = vp; |
| 1108 | ip->i_e2fs = fs = ump->um_e2fs; |
| 1109 | ip->i_dev = dev; |
| 1110 | ip->i_number = ino; |
| 1111 | #if QUOTA |
| 1112 | for (i = 0; i < MAXQUOTAS; i++) |
| 1113 | ip->i_dquot[i] = NODQUOT; |
| 1114 | #endif |
| 1115 | /* |
| 1116 | * Put it onto its hash chain. Since our vnode is locked, other |
| 1117 | * requests for this inode will block if they arrive while we are |
| 1118 | * sleeping waiting for old data structures to be purged or for the |
| 1119 | * contents of the disk portion of this inode to be read. |
| 1120 | */ |
| 1121 | ext2_ihashins(ip); |
| 1122 | |
| 1123 | if (ext2fs_inode_hash_lock < 0) |
| 1124 | wakeup(&ext2fs_inode_hash_lock); |
| 1125 | ext2fs_inode_hash_lock = 0; |
| 1126 | |
| 1127 | /* Read in the disk contents for the inode, copy into the inode. */ |
| 1128 | #if 0 |
| 1129 | kprintf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino))); |
| 1130 | #endif |
| 1131 | error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)), |
| 1132 | (int)fs->s_blocksize, &bp); |
| 1133 | if (error) { |
| 1134 | /* |
| 1135 | * The inode does not contain anything useful, so it would |
| 1136 | * be misleading to leave it on its hash chain. With mode |
| 1137 | * still zero, it will be unlinked and returned to the free |
| 1138 | * list by vput(). |
| 1139 | */ |
| 1140 | vx_put(vp); |
| 1141 | brelse(bp); |
| 1142 | *vpp = NULL; |
| 1143 | return (error); |
| 1144 | } |
| 1145 | /* convert ext2 inode to dinode */ |
| 1146 | ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE * |
| 1147 | ino_to_fsbo(fs, ino)), &ip->i_din); |
| 1148 | ip->i_block_group = ino_to_cg(fs, ino); |
| 1149 | ip->i_next_alloc_block = 0; |
| 1150 | ip->i_next_alloc_goal = 0; |
| 1151 | ip->i_prealloc_count = 0; |
| 1152 | ip->i_prealloc_block = 0; |
| 1153 | /* now we want to make sure that block pointers for unused |
| 1154 | blocks are zeroed out - ext2_balloc depends on this |
| 1155 | although for regular files and directories only |
| 1156 | */ |
| 1157 | if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) { |
| 1158 | used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize; |
| 1159 | for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) |
| 1160 | ip->i_db[i] = 0; |
| 1161 | } |
| 1162 | #if 0 |
| 1163 | ext2_print_inode(ip); |
| 1164 | #endif |
| 1165 | brelse(bp); |
| 1166 | |
| 1167 | /* |
| 1168 | * Initialize the vnode from the inode, check for aliases. |
| 1169 | * Note that the underlying vnode may have changed. |
| 1170 | */ |
| 1171 | if ((error = ext2_vinit(mp, &vp)) != 0) { |
| 1172 | vx_put(vp); |
| 1173 | *vpp = NULL; |
| 1174 | return (error); |
| 1175 | } |
| 1176 | |
| 1177 | /* |
| 1178 | * Finish inode initialization now that aliasing has been resolved. |
| 1179 | */ |
| 1180 | ip->i_devvp = ump->um_devvp; |
| 1181 | vref(ip->i_devvp); |
| 1182 | /* |
| 1183 | * Set up a generation number for this inode if it does not |
| 1184 | * already have one. This should only happen on old filesystems. |
| 1185 | */ |
| 1186 | if (ip->i_gen == 0) { |
| 1187 | ip->i_gen = krandom() / 2 + 1; |
| 1188 | if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) |
| 1189 | ip->i_flag |= IN_MODIFIED; |
| 1190 | } |
| 1191 | /* |
| 1192 | * Return the locked and refd vnode. |
| 1193 | */ |
| 1194 | *vpp = vp; |
| 1195 | return (0); |
| 1196 | } |
| 1197 | |
| 1198 | /* |
| 1199 | * File handle to vnode |
| 1200 | * |
| 1201 | * Have to be really careful about stale file handles: |
| 1202 | * - check that the inode number is valid |
| 1203 | * - call ext2_vget() to get the locked inode |
| 1204 | * - check for an unallocated inode (i_mode == 0) |
| 1205 | * - check that the given client host has export rights and return |
| 1206 | * those rights via. exflagsp and credanonp |
| 1207 | */ |
| 1208 | static int |
| 1209 | ext2_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp) |
| 1210 | { |
| 1211 | struct ufid *ufhp; |
| 1212 | struct ext2_sb_info *fs; |
| 1213 | struct inode *ip; |
| 1214 | struct vnode *nvp; |
| 1215 | int error; |
| 1216 | |
| 1217 | ufhp = (struct ufid *)fhp; |
| 1218 | fs = VFSTOEXT2(mp)->um_e2fs; |
| 1219 | if (ufhp->ufid_ino < ROOTINO || |
| 1220 | ufhp->ufid_ino > fs->s_groups_count * fs->s_es->s_inodes_per_group) |
| 1221 | return (ESTALE); |
| 1222 | |
| 1223 | error = VFS_VGET(mp, ufhp->ufid_ino, &nvp); |
| 1224 | if (error) { |
| 1225 | *vpp = NULLVP; |
| 1226 | return (error); |
| 1227 | } |
| 1228 | ip = VTOI(nvp); |
| 1229 | if (ip->i_mode == 0 || |
| 1230 | ip->i_gen != ufhp->ufid_gen || |
| 1231 | (VFSTOEXT2(mp)->um_i_effnlink_valid ? ip->i_effnlink : |
| 1232 | ip->i_nlink) <= 0) { |
| 1233 | vput(nvp); |
| 1234 | *vpp = NULLVP; |
| 1235 | return (ESTALE); |
| 1236 | } |
| 1237 | *vpp = nvp; |
| 1238 | return (0); |
| 1239 | } |
| 1240 | |
| 1241 | /* |
| 1242 | * Vnode pointer to File handle |
| 1243 | */ |
| 1244 | /* ARGSUSED */ |
| 1245 | static int |
| 1246 | ext2_vptofh(struct vnode *vp, struct fid *fhp) |
| 1247 | { |
| 1248 | struct inode *ip; |
| 1249 | struct ufid *ufhp; |
| 1250 | |
| 1251 | ip = VTOI(vp); |
| 1252 | ufhp = (struct ufid *)fhp; |
| 1253 | ufhp->ufid_len = sizeof(struct ufid); |
| 1254 | ufhp->ufid_ino = ip->i_number; |
| 1255 | ufhp->ufid_gen = ip->i_gen; |
| 1256 | return (0); |
| 1257 | } |
| 1258 | |
| 1259 | /* |
| 1260 | * This is the generic part of fhtovp called after the underlying |
| 1261 | * filesystem has validated the file handle. |
| 1262 | * |
| 1263 | * Verify that a host should have access to a filesystem. |
| 1264 | */ |
| 1265 | int |
| 1266 | ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp, |
| 1267 | struct ucred **credanonp) |
| 1268 | { |
| 1269 | struct netcred *np; |
| 1270 | struct ext2mount *ump; |
| 1271 | |
| 1272 | ump = VFSTOEXT2(mp); |
| 1273 | /* |
| 1274 | * Get the export permission structure for this <mp, client> tuple. |
| 1275 | */ |
| 1276 | np = vfs_export_lookup(mp, &ump->um_export, nam); |
| 1277 | if (np == NULL) |
| 1278 | return (EACCES); |
| 1279 | |
| 1280 | *exflagsp = np->netc_exflags; |
| 1281 | *credanonp = &np->netc_anon; |
| 1282 | return (0); |
| 1283 | } |
| 1284 | |
| 1285 | /* |
| 1286 | * Write a superblock and associated information back to disk. |
| 1287 | */ |
| 1288 | static int |
| 1289 | ext2_sbupdate(struct ext2mount *mp, int waitfor) |
| 1290 | { |
| 1291 | struct ext2_sb_info *fs = mp->um_e2fs; |
| 1292 | struct ext2_super_block *es = fs->s_es; |
| 1293 | struct buf *bp; |
| 1294 | int error = 0; |
| 1295 | /* |
| 1296 | kprintf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no"); |
| 1297 | */ |
| 1298 | bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0); |
| 1299 | bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block)); |
| 1300 | if (waitfor == MNT_WAIT) |
| 1301 | error = bwrite(bp); |
| 1302 | else |
| 1303 | bawrite(bp); |
| 1304 | |
| 1305 | /* |
| 1306 | * The buffers for group descriptors, inode bitmaps and block bitmaps |
| 1307 | * are not busy at this point and are (hopefully) written by the |
| 1308 | * usual sync mechanism. No need to write them here |
| 1309 | */ |
| 1310 | |
| 1311 | return (error); |
| 1312 | } |