ext2_vfsops.c revision 96752
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: head/sys/gnu/fs/ext2fs/ext2_vfsops.c 96752 2002-05-16 19:43:28Z iedowse $ 41 */ 42 43#include <sys/param.h> 44#include <sys/systm.h> 45#include <sys/namei.h> 46#include <sys/proc.h> 47#include <sys/kernel.h> 48#include <sys/vnode.h> 49#include <sys/mount.h> 50#include <sys/bio.h> 51#include <sys/buf.h> 52#include <sys/conf.h> 53#include <sys/fcntl.h> 54#include <sys/malloc.h> 55#include <sys/stat.h> 56#include <sys/mutex.h> 57 58#include <gnu/ext2fs/ext2_mount.h> 59#include <gnu/ext2fs/inode.h> 60 61#include <gnu/ext2fs/fs.h> 62#include <gnu/ext2fs/ext2_extern.h> 63#include <gnu/ext2fs/ext2_fs.h> 64#include <gnu/ext2fs/ext2_fs_sb.h> 65 66static int ext2_fhtovp(struct mount *, struct fid *, struct vnode **); 67static int ext2_flushfiles(struct mount *mp, int flags, struct thread *td); 68static int ext2_init(struct vfsconf *); 69static int ext2_mount(struct mount *, 70 char *, caddr_t, struct nameidata *, struct thread *); 71static int ext2_mountfs(struct vnode *, struct mount *, struct thread *); 72static int ext2_reload(struct mount *mountp, struct ucred *cred, 73 struct thread *td); 74static int ext2_root(struct mount *, struct vnode **vpp); 75static int ext2_sbupdate(struct ext2mount *, int); 76static int ext2_statfs(struct mount *, struct statfs *, struct thread *); 77static int ext2_sync(struct mount *, int, struct ucred *, struct thread *); 78static int ext2_unmount(struct mount *, int, struct thread *); 79static int ext2_vget(struct mount *, ino_t, int, struct vnode **); 80static int ext2_vptofh(struct vnode *, struct fid *); 81 82MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part"); 83static MALLOC_DEFINE(M_EXT2MNT, "EXT2 mount", "EXT2 mount structure"); 84 85static struct vfsops ext2fs_vfsops = { 86 ext2_mount, 87 vfs_stdstart, 88 ext2_unmount, 89 ext2_root, /* root inode via vget */ 90 vfs_stdquotactl, 91 ext2_statfs, 92 ext2_sync, 93 ext2_vget, 94 ext2_fhtovp, 95 vfs_stdcheckexp, 96 ext2_vptofh, 97 ext2_init, 98 vfs_stduninit, 99 vfs_stdextattrctl, 100}; 101 102VFS_SET(ext2fs_vfsops, ext2fs, 0); 103#define bsd_malloc malloc 104#define bsd_free free 105 106static int ext2fs_inode_hash_lock; 107 108static int ext2_check_sb_compat(struct ext2_super_block *es, dev_t dev, 109 int ronly); 110static int compute_sb_data(struct vnode * devvp, 111 struct ext2_super_block * es, struct ext2_sb_info * fs); 112 113#ifdef notyet 114static int ext2_mountroot(void); 115 116/* 117 * Called by main() when ext2fs is going to be mounted as root. 118 * 119 * Name is updated by mount(8) after booting. 120 */ 121#define ROOTNAME "root_device" 122 123static int 124ext2_mountroot() 125{ 126 struct ext2_sb_info *fs; 127 struct mount *mp; 128 struct thread *td = curthread; 129 struct ext2mount *ump; 130 u_int size; 131 int error; 132 133 if ((error = bdevvp(rootdev, &rootvp))) { 134 printf("ext2_mountroot: can't find rootvp\n"); 135 return (error); 136 } 137 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 138 bzero((char *)mp, (u_long)sizeof(struct mount)); 139 TAILQ_INIT(&mp->mnt_nvnodelist); 140 TAILQ_INIT(&mp->mnt_reservedvnlist); 141 mp->mnt_op = &ext2fs_vfsops; 142 mp->mnt_flag = MNT_RDONLY; 143 if (error = ext2_mountfs(rootvp, mp, td)) { 144 bsd_free(mp, M_MOUNT); 145 return (error); 146 } 147 if (error = vfs_lock(mp)) { 148 (void)ext2_unmount(mp, 0, td); 149 bsd_free(mp, M_MOUNT); 150 return (error); 151 } 152 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 153 mp->mnt_flag |= MNT_ROOTFS; 154 mp->mnt_vnodecovered = NULLVP; 155 ump = VFSTOEXT2(mp); 156 fs = ump->um_e2fs; 157 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt)); 158 fs->fs_fsmnt[0] = '/'; 159 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 160 MNAMELEN); 161 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 162 &size); 163 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 164 (void)ext2_statfs(mp, &mp->mnt_stat, td); 165 vfs_unlock(mp); 166 inittodr(fs->s_es->s_wtime); /* this helps to set the time */ 167 return (0); 168} 169#endif 170 171/* 172 * VFS Operations. 173 * 174 * mount system call 175 */ 176static int 177ext2_mount(mp, path, data, ndp, td) 178 struct mount *mp; 179 char *path; 180 caddr_t data; /* this is actually a (struct ext2_args *) */ 181 struct nameidata *ndp; 182 struct thread *td; 183{ 184 struct vnode *devvp; 185 struct ext2_args args; 186 struct ext2mount *ump = 0; 187 struct ext2_sb_info *fs; 188 size_t size; 189 int error, flags; 190 mode_t accessmode; 191 192 /* Double-check the length of path.. */ 193 if (strlen(path) >= MAXMNTLEN - 1) 194 return (ENAMETOOLONG); 195 error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args)); 196 if (error != 0) 197 return (error); 198 /* 199 * If updating, check whether changing from read-only to 200 * read/write; if there is no device name, that's all we do. 201 */ 202 if (mp->mnt_flag & MNT_UPDATE) { 203 ump = VFSTOEXT2(mp); 204 fs = ump->um_e2fs; 205 error = 0; 206 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) { 207 flags = WRITECLOSE; 208 if (mp->mnt_flag & MNT_FORCE) 209 flags |= FORCECLOSE; 210 if (vfs_busy(mp, LK_NOWAIT, 0, td)) 211 return (EBUSY); 212 error = ext2_flushfiles(mp, flags, td); 213 vfs_unbusy(mp, td); 214 if (!error && fs->s_wasvalid) { 215 fs->s_es->s_state |= EXT2_VALID_FS; 216 ext2_sbupdate(ump, MNT_WAIT); 217 } 218 fs->s_rd_only = 1; 219 } 220 if (!error && (mp->mnt_flag & MNT_RELOAD)) 221 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, td); 222 if (error) 223 return (error); 224 devvp = ump->um_devvp; 225 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev, 226 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0) 227 return (EPERM); 228 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { 229 /* 230 * If upgrade to read-write by non-root, then verify 231 * that user has necessary permissions on the device. 232 */ 233 if (suser(td)) { 234 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 235 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE, 236 td->td_ucred, td)) != 0) { 237 VOP_UNLOCK(devvp, 0, td); 238 return (error); 239 } 240 VOP_UNLOCK(devvp, 0, td); 241 } 242 243 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 || 244 (fs->s_es->s_state & EXT2_ERROR_FS)) { 245 if (mp->mnt_flag & MNT_FORCE) { 246 printf( 247"WARNING: %s was not properly dismounted\n", 248 fs->fs_fsmnt); 249 } else { 250 printf( 251"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", 252 fs->fs_fsmnt); 253 return (EPERM); 254 } 255 } 256 fs->s_es->s_state &= ~EXT2_VALID_FS; 257 ext2_sbupdate(ump, MNT_WAIT); 258 fs->s_rd_only = 0; 259 } 260 if (args.fspec == 0) { 261 /* 262 * Process export requests. 263 */ 264 return (vfs_export(mp, &args.export)); 265 } 266 } 267 /* 268 * Not an update, or updating the name: look up the name 269 * and verify that it refers to a sensible block device. 270 */ 271 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, td); 272 if ((error = namei(ndp)) != 0) 273 return (error); 274 NDFREE(ndp, NDF_ONLY_PNBUF); 275 devvp = ndp->ni_vp; 276 277 if (!vn_isdisk(devvp, &error)) { 278 vrele(devvp); 279 return (error); 280 } 281 282 /* 283 * If mount by non-root, then verify that user has necessary 284 * permissions on the device. 285 */ 286 if (suser(td)) { 287 accessmode = VREAD; 288 if ((mp->mnt_flag & MNT_RDONLY) == 0) 289 accessmode |= VWRITE; 290 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 291 if ((error = VOP_ACCESS(devvp, accessmode, td->td_ucred, td)) != 0) { 292 vput(devvp); 293 return (error); 294 } 295 VOP_UNLOCK(devvp, 0, td); 296 } 297 298 if ((mp->mnt_flag & MNT_UPDATE) == 0) { 299 error = ext2_mountfs(devvp, mp, td); 300 } else { 301 if (devvp != ump->um_devvp) 302 error = EINVAL; /* needs translation */ 303 else 304 vrele(devvp); 305 } 306 if (error) { 307 vrele(devvp); 308 return (error); 309 } 310 ump = VFSTOEXT2(mp); 311 fs = ump->um_e2fs; 312 /* 313 * Note that this strncpy() is ok because of a check at the start 314 * of ext2_mount(). 315 */ 316 strncpy(fs->fs_fsmnt, path, MAXMNTLEN); 317 fs->fs_fsmnt[MAXMNTLEN - 1] = '\0'; 318 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 319 &size); 320 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 321 (void)ext2_statfs(mp, &mp->mnt_stat, td); 322 return (0); 323} 324 325/* 326 * checks that the data in the descriptor blocks make sense 327 * this is taken from ext2/super.c 328 */ 329static int ext2_check_descriptors (struct ext2_sb_info * sb) 330{ 331 int i; 332 int desc_block = 0; 333 unsigned long block = sb->s_es->s_first_data_block; 334 struct ext2_group_desc * gdp = NULL; 335 336 /* ext2_debug ("Checking group descriptors"); */ 337 338 for (i = 0; i < sb->s_groups_count; i++) 339 { 340 /* examine next descriptor block */ 341 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0) 342 gdp = (struct ext2_group_desc *) 343 sb->s_group_desc[desc_block++]->b_data; 344 if (gdp->bg_block_bitmap < block || 345 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 346 { 347 printf ("ext2_check_descriptors: " 348 "Block bitmap for group %d" 349 " not in group (block %lu)!\n", 350 i, (unsigned long) gdp->bg_block_bitmap); 351 return 0; 352 } 353 if (gdp->bg_inode_bitmap < block || 354 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 355 { 356 printf ("ext2_check_descriptors: " 357 "Inode bitmap for group %d" 358 " not in group (block %lu)!\n", 359 i, (unsigned long) gdp->bg_inode_bitmap); 360 return 0; 361 } 362 if (gdp->bg_inode_table < block || 363 gdp->bg_inode_table + sb->s_itb_per_group >= 364 block + EXT2_BLOCKS_PER_GROUP(sb)) 365 { 366 printf ("ext2_check_descriptors: " 367 "Inode table for group %d" 368 " not in group (block %lu)!\n", 369 i, (unsigned long) gdp->bg_inode_table); 370 return 0; 371 } 372 block += EXT2_BLOCKS_PER_GROUP(sb); 373 gdp++; 374 } 375 return 1; 376} 377 378static int 379ext2_check_sb_compat(es, dev, ronly) 380 struct ext2_super_block *es; 381 dev_t dev; 382 int ronly; 383{ 384 385 if (es->s_magic != EXT2_SUPER_MAGIC) { 386 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n", 387 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC); 388 return (1); 389 } 390 if (es->s_rev_level > EXT2_GOOD_OLD_REV) { 391 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) { 392 printf( 393"WARNING: mount of %s denied due to unsupported optional features\n", 394 devtoname(dev)); 395 return (1); 396 } 397 if (!ronly && 398 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) { 399 printf( 400"WARNING: R/W mount of %s denied due to unsupported optional features\n", 401 devtoname(dev)); 402 return (1); 403 } 404 } 405 return (0); 406} 407 408/* 409 * this computes the fields of the ext2_sb_info structure from the 410 * data in the ext2_super_block structure read in 411 */ 412static int compute_sb_data(devvp, es, fs) 413 struct vnode * devvp; 414 struct ext2_super_block * es; 415 struct ext2_sb_info * fs; 416{ 417 int db_count, error; 418 int i, j; 419 int logic_sb_block = 1; /* XXX for now */ 420 421#if 1 422#define V(v) 423#else 424#define V(v) printf(#v"= %d\n", fs->v); 425#endif 426 427 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size; 428 V(s_blocksize) 429 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size; 430 V(s_bshift) 431 fs->s_fsbtodb = es->s_log_block_size + 1; 432 V(s_fsbtodb) 433 fs->s_qbmask = fs->s_blocksize - 1; 434 V(s_bmask) 435 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es); 436 V(s_blocksize_bits) 437 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size; 438 V(s_frag_size) 439 if (fs->s_frag_size) 440 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size; 441 V(s_frags_per_block) 442 fs->s_blocks_per_group = es->s_blocks_per_group; 443 V(s_blocks_per_group) 444 fs->s_frags_per_group = es->s_frags_per_group; 445 V(s_frags_per_group) 446 fs->s_inodes_per_group = es->s_inodes_per_group; 447 V(s_inodes_per_group) 448 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE; 449 V(s_inodes_per_block) 450 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block; 451 V(s_itb_per_group) 452 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc); 453 V(s_desc_per_block) 454 /* s_resuid / s_resgid ? */ 455 fs->s_groups_count = (es->s_blocks_count - 456 es->s_first_data_block + 457 EXT2_BLOCKS_PER_GROUP(fs) - 1) / 458 EXT2_BLOCKS_PER_GROUP(fs); 459 V(s_groups_count) 460 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) / 461 EXT2_DESC_PER_BLOCK(fs); 462 fs->s_db_per_group = db_count; 463 V(s_db_per_group) 464 465 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *), 466 M_EXT2MNT, M_WAITOK); 467 468 /* adjust logic_sb_block */ 469 if(fs->s_blocksize > SBSIZE) 470 /* Godmar thinks: if the blocksize is greater than 1024, then 471 the superblock is logically part of block zero. 472 */ 473 logic_sb_block = 0; 474 475 for (i = 0; i < db_count; i++) { 476 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1), 477 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]); 478 if(error) { 479 for (j = 0; j < i; j++) 480 brelse(fs->s_group_desc[j]); 481 bsd_free(fs->s_group_desc, M_EXT2MNT); 482 printf("EXT2-fs: unable to read group descriptors (%d)\n", error); 483 return EIO; 484 } 485 /* Set the B_LOCKED flag on the buffer, then brelse() it */ 486 LCK_BUF(fs->s_group_desc[i]) 487 } 488 if(!ext2_check_descriptors(fs)) { 489 for (j = 0; j < db_count; j++) 490 ULCK_BUF(fs->s_group_desc[j]) 491 bsd_free(fs->s_group_desc, M_EXT2MNT); 492 printf("EXT2-fs: (ext2_check_descriptors failure) " 493 "unable to read group descriptors\n"); 494 return EIO; 495 } 496 497 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) { 498 fs->s_inode_bitmap_number[i] = 0; 499 fs->s_inode_bitmap[i] = NULL; 500 fs->s_block_bitmap_number[i] = 0; 501 fs->s_block_bitmap[i] = NULL; 502 } 503 fs->s_loaded_inode_bitmaps = 0; 504 fs->s_loaded_block_bitmaps = 0; 505 return 0; 506} 507 508/* 509 * Reload all incore data for a filesystem (used after running fsck on 510 * the root filesystem and finding things to fix). The filesystem must 511 * be mounted read-only. 512 * 513 * Things to do to update the mount: 514 * 1) invalidate all cached meta-data. 515 * 2) re-read superblock from disk. 516 * 3) re-read summary information from disk. 517 * 4) invalidate all inactive vnodes. 518 * 5) invalidate all cached file data. 519 * 6) re-read inode data for all active vnodes. 520 */ 521static int 522ext2_reload(mountp, cred, td) 523 struct mount *mountp; 524 struct ucred *cred; 525 struct thread *td; 526{ 527 struct vnode *vp, *nvp, *devvp; 528 struct inode *ip; 529 struct buf *bp; 530 struct ext2_super_block * es; 531 struct ext2_sb_info *fs; 532 int error; 533 534 if ((mountp->mnt_flag & MNT_RDONLY) == 0) 535 return (EINVAL); 536 /* 537 * Step 1: invalidate all cached meta-data. 538 */ 539 devvp = VFSTOEXT2(mountp)->um_devvp; 540 if (vinvalbuf(devvp, 0, cred, td, 0, 0)) 541 panic("ext2_reload: dirty1"); 542 /* 543 * Step 2: re-read superblock from disk. 544 * constants have been adjusted for ext2 545 */ 546 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 547 return (error); 548 es = (struct ext2_super_block *)bp->b_data; 549 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) { 550 brelse(bp); 551 return (EIO); /* XXX needs translation */ 552 } 553 fs = VFSTOEXT2(mountp)->um_e2fs; 554 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block)); 555 556 if((error = compute_sb_data(devvp, es, fs)) != 0) { 557 brelse(bp); 558 return error; 559 } 560#ifdef UNKLAR 561 if (fs->fs_sbsize < SBSIZE) 562 bp->b_flags |= B_INVAL; 563#endif 564 brelse(bp); 565 566loop: 567 mtx_lock(&mntvnode_mtx); 568 for (vp = TAILQ_FIRST(&mountp->mnt_nvnodelist); vp != NULL; vp = nvp) { 569 if (vp->v_mount != mountp) { 570 mtx_unlock(&mntvnode_mtx); 571 goto loop; 572 } 573 nvp = TAILQ_NEXT(vp, v_nmntvnodes); 574 mtx_unlock(&mntvnode_mtx); 575 /* 576 * Step 4: invalidate all inactive vnodes. 577 */ 578 if (vrecycle(vp, NULL, td)) 579 goto loop; 580 /* 581 * Step 5: invalidate all cached file data. 582 */ 583 mtx_lock(&vp->v_interlock); 584 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) { 585 goto loop; 586 } 587 if (vinvalbuf(vp, 0, cred, td, 0, 0)) 588 panic("ext2_reload: dirty2"); 589 /* 590 * Step 6: re-read inode data for all active vnodes. 591 */ 592 ip = VTOI(vp); 593 error = 594 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 595 (int)fs->s_blocksize, NOCRED, &bp); 596 if (error) { 597 vput(vp); 598 return (error); 599 } 600 ext2_ei2i((struct ext2_inode *) ((char *)bp->b_data + 601 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)), ip); 602 brelse(bp); 603 vput(vp); 604 mtx_lock(&mntvnode_mtx); 605 } 606 mtx_unlock(&mntvnode_mtx); 607 return (0); 608} 609 610/* 611 * Common code for mount and mountroot 612 */ 613static int 614ext2_mountfs(devvp, mp, td) 615 struct vnode *devvp; 616 struct mount *mp; 617 struct thread *td; 618{ 619 struct ext2mount *ump; 620 struct buf *bp; 621 struct ext2_sb_info *fs; 622 struct ext2_super_block * es; 623 dev_t dev = devvp->v_rdev; 624 int error; 625 int ronly; 626 627 /* 628 * Disallow multiple mounts of the same device. 629 * Disallow mounting of a device that is currently in use 630 * (except for root, which might share swap device for miniroot). 631 * Flush out any old buffers remaining from a previous use. 632 */ 633 if ((error = vfs_mountedon(devvp)) != 0) 634 return (error); 635 if (vcount(devvp) > 1 && devvp != rootvp) 636 return (EBUSY); 637 if ((error = vinvalbuf(devvp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) 638 return (error); 639#ifdef READONLY 640/* turn on this to force it to be read-only */ 641 mp->mnt_flag |= MNT_RDONLY; 642#endif 643 644 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 645 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 646 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, td); 647 VOP_UNLOCK(devvp, 0, td); 648 if (error) 649 return (error); 650 if (devvp->v_rdev->si_iosize_max != 0) 651 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max; 652 if (mp->mnt_iosize_max > MAXPHYS) 653 mp->mnt_iosize_max = MAXPHYS; 654 655 bp = NULL; 656 ump = NULL; 657 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 658 goto out; 659 es = (struct ext2_super_block *)bp->b_data; 660 if (ext2_check_sb_compat(es, dev, ronly) != 0) { 661 error = EINVAL; /* XXX needs translation */ 662 goto out; 663 } 664 if ((es->s_state & EXT2_VALID_FS) == 0 || 665 (es->s_state & EXT2_ERROR_FS)) { 666 if (ronly || (mp->mnt_flag & MNT_FORCE)) { 667 printf( 668"WARNING: Filesystem was not properly dismounted\n"); 669 } else { 670 printf( 671"WARNING: R/W mount denied. Filesystem is not clean - run fsck\n"); 672 error = EPERM; 673 goto out; 674 } 675 } 676 ump = bsd_malloc(sizeof *ump, M_EXT2MNT, M_WAITOK); 677 bzero((caddr_t)ump, sizeof *ump); 678 /* I don't know whether this is the right strategy. Note that 679 we dynamically allocate both a ext2_sb_info and a ext2_super_block 680 while Linux keeps the super block in a locked buffer 681 */ 682 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info), 683 M_EXT2MNT, M_WAITOK); 684 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block), 685 M_EXT2MNT, M_WAITOK); 686 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block)); 687 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs))) 688 goto out; 689 /* 690 * We don't free the group descriptors allocated by compute_sb_data() 691 * until ext2_unmount(). This is OK since the mount will succeed. 692 */ 693 brelse(bp); 694 bp = NULL; 695 fs = ump->um_e2fs; 696 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */ 697 /* if the fs is not mounted read-only, make sure the super block is 698 always written back on a sync() 699 */ 700 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0; 701 if (ronly == 0) { 702 fs->s_dirt = 1; /* mark it modified */ 703 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */ 704 } 705 mp->mnt_data = (qaddr_t)ump; 706 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev); 707 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; 708 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; 709 mp->mnt_flag |= MNT_LOCAL; 710 ump->um_mountp = mp; 711 ump->um_dev = dev; 712 ump->um_devvp = devvp; 713 /* setting those two parameters allowed us to use 714 ufs_bmap w/o changse ! 715 */ 716 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); 717 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1; 718 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); 719 devvp->v_rdev->si_mountpoint = mp; 720 if (ronly == 0) 721 ext2_sbupdate(ump, MNT_WAIT); 722 return (0); 723out: 724 if (bp) 725 brelse(bp); 726 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, td); 727 if (ump) { 728 bsd_free(ump->um_e2fs->s_es, M_EXT2MNT); 729 bsd_free(ump->um_e2fs, M_EXT2MNT); 730 bsd_free(ump, M_EXT2MNT); 731 mp->mnt_data = (qaddr_t)0; 732 } 733 return (error); 734} 735 736/* 737 * unmount system call 738 */ 739static int 740ext2_unmount(mp, mntflags, td) 741 struct mount *mp; 742 int mntflags; 743 struct thread *td; 744{ 745 struct ext2mount *ump; 746 struct ext2_sb_info *fs; 747 int error, flags, ronly, i; 748 749 flags = 0; 750 if (mntflags & MNT_FORCE) { 751 if (mp->mnt_flag & MNT_ROOTFS) 752 return (EINVAL); 753 flags |= FORCECLOSE; 754 } 755 if ((error = ext2_flushfiles(mp, flags, td)) != 0) 756 return (error); 757 ump = VFSTOEXT2(mp); 758 fs = ump->um_e2fs; 759 ronly = fs->s_rd_only; 760 if (ronly == 0) { 761 if (fs->s_wasvalid) 762 fs->s_es->s_state |= EXT2_VALID_FS; 763 ext2_sbupdate(ump, MNT_WAIT); 764 } 765 766 /* release buffers containing group descriptors */ 767 for(i = 0; i < fs->s_db_per_group; i++) 768 ULCK_BUF(fs->s_group_desc[i]) 769 bsd_free(fs->s_group_desc, M_EXT2MNT); 770 771 /* release cached inode/block bitmaps */ 772 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 773 if (fs->s_inode_bitmap[i]) 774 ULCK_BUF(fs->s_inode_bitmap[i]) 775 776 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 777 if (fs->s_block_bitmap[i]) 778 ULCK_BUF(fs->s_block_bitmap[i]) 779 780 ump->um_devvp->v_rdev->si_mountpoint = NULL; 781 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, 782 NOCRED, td); 783 vrele(ump->um_devvp); 784 bsd_free(fs->s_es, M_EXT2MNT); 785 bsd_free(fs, M_EXT2MNT); 786 bsd_free(ump, M_EXT2MNT); 787 mp->mnt_data = (qaddr_t)0; 788 mp->mnt_flag &= ~MNT_LOCAL; 789 return (error); 790} 791 792/* 793 * Flush out all the files in a filesystem. 794 */ 795static int 796ext2_flushfiles(mp, flags, td) 797 struct mount *mp; 798 int flags; 799 struct thread *td; 800{ 801 int error; 802 803 error = vflush(mp, 0, flags); 804 return (error); 805} 806 807/* 808 * Get file system statistics. 809 * taken from ext2/super.c ext2_statfs 810 */ 811static int 812ext2_statfs(mp, sbp, td) 813 struct mount *mp; 814 struct statfs *sbp; 815 struct thread *td; 816{ 817 unsigned long overhead; 818 struct ext2mount *ump; 819 struct ext2_sb_info *fs; 820 struct ext2_super_block *es; 821 int i, nsb; 822 823 ump = VFSTOEXT2(mp); 824 fs = ump->um_e2fs; 825 es = fs->s_es; 826 827 if (es->s_magic != EXT2_SUPER_MAGIC) 828 panic("ext2_statfs - magic number spoiled"); 829 830 /* 831 * Compute the overhead (FS structures) 832 */ 833 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) { 834 nsb = 0; 835 for (i = 0 ; i < fs->s_groups_count; i++) 836 if (ext2_group_sparse(i)) 837 nsb++; 838 } else 839 nsb = fs->s_groups_count; 840 overhead = es->s_first_data_block + 841 /* Superblocks and block group descriptors: */ 842 nsb * (1 + fs->s_db_per_group) + 843 /* Inode bitmap, block bitmap, and inode table: */ 844 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group); 845 846 sbp->f_bsize = EXT2_FRAG_SIZE(fs); 847 sbp->f_iosize = EXT2_BLOCK_SIZE(fs); 848 sbp->f_blocks = es->s_blocks_count - overhead; 849 sbp->f_bfree = es->s_free_blocks_count; 850 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count; 851 sbp->f_files = es->s_inodes_count; 852 sbp->f_ffree = es->s_free_inodes_count; 853 if (sbp != &mp->mnt_stat) { 854 sbp->f_type = mp->mnt_vfc->vfc_typenum; 855 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 856 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 857 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 858 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 859 } 860 return (0); 861} 862 863/* 864 * Go through the disk queues to initiate sandbagged IO; 865 * go through the inodes to write those that have been modified; 866 * initiate the writing of the super block if it has been modified. 867 * 868 * Note: we are always called with the filesystem marked `MPBUSY'. 869 */ 870static int 871ext2_sync(mp, waitfor, cred, td) 872 struct mount *mp; 873 int waitfor; 874 struct ucred *cred; 875 struct thread *td; 876{ 877 struct vnode *nvp, *vp; 878 struct inode *ip; 879 struct ext2mount *ump = VFSTOEXT2(mp); 880 struct ext2_sb_info *fs; 881 int error, allerror = 0; 882 883 fs = ump->um_e2fs; 884 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */ 885 printf("fs = %s\n", fs->fs_fsmnt); 886 panic("ext2_sync: rofs mod"); 887 } 888 /* 889 * Write back each (modified) inode. 890 */ 891 mtx_lock(&mntvnode_mtx); 892loop: 893 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) { 894 /* 895 * If the vnode that we are about to sync is no longer 896 * associated with this mount point, start over. 897 */ 898 if (vp->v_mount != mp) 899 goto loop; 900 nvp = TAILQ_NEXT(vp, v_nmntvnodes); 901 mtx_unlock(&mntvnode_mtx); 902 mtx_lock(&vp->v_interlock); 903 ip = VTOI(vp); 904 if (vp->v_type == VNON || 905 ((ip->i_flag & 906 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 907 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || waitfor == MNT_LAZY))) { 908 mtx_unlock(&vp->v_interlock); 909 mtx_lock(&mntvnode_mtx); 910 continue; 911 } 912 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td); 913 if (error) { 914 mtx_lock(&mntvnode_mtx); 915 if (error == ENOENT) 916 goto loop; 917 continue; 918 } 919 if ((error = VOP_FSYNC(vp, cred, waitfor, td)) != 0) 920 allerror = error; 921 VOP_UNLOCK(vp, 0, td); 922 vrele(vp); 923 mtx_lock(&mntvnode_mtx); 924 } 925 mtx_unlock(&mntvnode_mtx); 926 /* 927 * Force stale file system control information to be flushed. 928 */ 929 if (waitfor != MNT_LAZY) { 930 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td); 931 if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, td)) != 0) 932 allerror = error; 933 VOP_UNLOCK(ump->um_devvp, 0, td); 934 } 935 /* 936 * Write back modified superblock. 937 */ 938 if (fs->s_dirt != 0) { 939 fs->s_dirt = 0; 940 fs->s_es->s_wtime = time_second; 941 if ((error = ext2_sbupdate(ump, waitfor)) != 0) 942 allerror = error; 943 } 944 return (allerror); 945} 946 947/* 948 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it 949 * in from disk. If it is in core, wait for the lock bit to clear, then 950 * return the inode locked. Detection and handling of mount points must be 951 * done by the calling routine. 952 */ 953static int 954ext2_vget(mp, ino, flags, vpp) 955 struct mount *mp; 956 ino_t ino; 957 int flags; 958 struct vnode **vpp; 959{ 960 struct ext2_sb_info *fs; 961 struct inode *ip; 962 struct ext2mount *ump; 963 struct buf *bp; 964 struct vnode *vp; 965 dev_t dev; 966 int i, error; 967 int used_blocks; 968 969 ump = VFSTOEXT2(mp); 970 dev = ump->um_dev; 971restart: 972 if ((error = ext2_ihashget(dev, ino, flags, vpp)) != 0) 973 return (error); 974 if (*vpp != NULL) 975 return (0); 976 977 /* 978 * Lock out the creation of new entries in the FFS hash table in 979 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 980 * may occur! 981 */ 982 if (ext2fs_inode_hash_lock) { 983 while (ext2fs_inode_hash_lock) { 984 ext2fs_inode_hash_lock = -1; 985 tsleep(&ext2fs_inode_hash_lock, PVM, "e2vget", 0); 986 } 987 goto restart; 988 } 989 ext2fs_inode_hash_lock = 1; 990 991 /* 992 * If this MALLOC() is performed after the getnewvnode() 993 * it might block, leaving a vnode with a NULL v_data to be 994 * found by ext2_sync() if a sync happens to fire right then, 995 * which will cause a panic because ext2_sync() blindly 996 * dereferences vp->v_data (as well it should). 997 */ 998 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK); 999 1000 /* Allocate a new vnode/inode. */ 1001 if ((error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) != 0) { 1002 if (ext2fs_inode_hash_lock < 0) 1003 wakeup(&ext2fs_inode_hash_lock); 1004 ext2fs_inode_hash_lock = 0; 1005 *vpp = NULL; 1006 FREE(ip, M_EXT2NODE); 1007 return (error); 1008 } 1009 bzero((caddr_t)ip, sizeof(struct inode)); 1010 lockinit(&vp->v_lock, PINOD, "ext2in", 0, 0); 1011 vp->v_data = ip; 1012 ip->i_vnode = vp; 1013 ip->i_e2fs = fs = ump->um_e2fs; 1014 ip->i_dev = dev; 1015 ip->i_number = ino; 1016 /* 1017 * Put it onto its hash chain and lock it so that other requests for 1018 * this inode will block if they arrive while we are sleeping waiting 1019 * for old data structures to be purged or for the contents of the 1020 * disk portion of this inode to be read. 1021 */ 1022 ext2_ihashins(ip); 1023 1024 if (ext2fs_inode_hash_lock < 0) 1025 wakeup(&ext2fs_inode_hash_lock); 1026 ext2fs_inode_hash_lock = 0; 1027 1028 /* Read in the disk contents for the inode, copy into the inode. */ 1029#if 0 1030printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino))); 1031#endif 1032 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 1033 (int)fs->s_blocksize, NOCRED, &bp)) != 0) { 1034 /* 1035 * The inode does not contain anything useful, so it would 1036 * be misleading to leave it on its hash chain. With mode 1037 * still zero, it will be unlinked and returned to the free 1038 * list by vput(). 1039 */ 1040 vput(vp); 1041 brelse(bp); 1042 *vpp = NULL; 1043 return (error); 1044 } 1045 /* convert ext2 inode to dinode */ 1046 ext2_ei2i((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE * 1047 ino_to_fsbo(fs, ino)), ip); 1048 ip->i_block_group = ino_to_cg(fs, ino); 1049 ip->i_next_alloc_block = 0; 1050 ip->i_next_alloc_goal = 0; 1051 ip->i_prealloc_count = 0; 1052 ip->i_prealloc_block = 0; 1053 /* now we want to make sure that block pointers for unused 1054 blocks are zeroed out - ext2_balloc depends on this 1055 although for regular files and directories only 1056 */ 1057 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) { 1058 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize; 1059 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) 1060 ip->i_db[i] = 0; 1061 } 1062/* 1063 ext2_print_inode(ip); 1064*/ 1065 brelse(bp); 1066 1067 /* 1068 * Initialize the vnode from the inode, check for aliases. 1069 * Note that the underlying vnode may have changed. 1070 */ 1071 if ((error = ext2_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) != 0) { 1072 vput(vp); 1073 *vpp = NULL; 1074 return (error); 1075 } 1076 /* 1077 * Finish inode initialization now that aliasing has been resolved. 1078 */ 1079 ip->i_devvp = ump->um_devvp; 1080 VREF(ip->i_devvp); 1081 /* 1082 * Set up a generation number for this inode if it does not 1083 * already have one. This should only happen on old filesystems. 1084 */ 1085 if (ip->i_gen == 0) { 1086 ip->i_gen = random() / 2 + 1; 1087 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1088 ip->i_flag |= IN_MODIFIED; 1089 } 1090 *vpp = vp; 1091 return (0); 1092} 1093 1094/* 1095 * File handle to vnode 1096 * 1097 * Have to be really careful about stale file handles: 1098 * - check that the inode number is valid 1099 * - call ext2_vget() to get the locked inode 1100 * - check for an unallocated inode (i_mode == 0) 1101 * - check that the given client host has export rights and return 1102 * those rights via. exflagsp and credanonp 1103 */ 1104static int 1105ext2_fhtovp(mp, fhp, vpp) 1106 struct mount *mp; 1107 struct fid *fhp; 1108 struct vnode **vpp; 1109{ 1110 struct inode *ip; 1111 struct ufid *ufhp; 1112 struct vnode *nvp; 1113 struct ext2_sb_info *fs; 1114 int error; 1115 1116 ufhp = (struct ufid *)fhp; 1117 fs = VFSTOEXT2(mp)->um_e2fs; 1118 if (ufhp->ufid_ino < ROOTINO || 1119 ufhp->ufid_ino >= fs->s_groups_count * fs->s_es->s_inodes_per_group) 1120 return (ESTALE); 1121 1122 error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp); 1123 if (error) { 1124 *vpp = NULLVP; 1125 return (error); 1126 } 1127 ip = VTOI(nvp); 1128 if (ip->i_mode == 0 || 1129 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) { 1130 vput(nvp); 1131 *vpp = NULLVP; 1132 return (ESTALE); 1133 } 1134 *vpp = nvp; 1135 return (0); 1136} 1137 1138/* 1139 * Vnode pointer to File handle 1140 */ 1141/* ARGSUSED */ 1142static int 1143ext2_vptofh(vp, fhp) 1144 struct vnode *vp; 1145 struct fid *fhp; 1146{ 1147 struct inode *ip; 1148 struct ufid *ufhp; 1149 1150 ip = VTOI(vp); 1151 ufhp = (struct ufid *)fhp; 1152 ufhp->ufid_len = sizeof(struct ufid); 1153 ufhp->ufid_ino = ip->i_number; 1154 ufhp->ufid_gen = ip->i_gen; 1155 return (0); 1156} 1157 1158/* 1159 * Write a superblock and associated information back to disk. 1160 */ 1161static int 1162ext2_sbupdate(mp, waitfor) 1163 struct ext2mount *mp; 1164 int waitfor; 1165{ 1166 struct ext2_sb_info *fs = mp->um_e2fs; 1167 struct ext2_super_block *es = fs->s_es; 1168 struct buf *bp; 1169 int error = 0; 1170/* 1171printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no"); 1172*/ 1173 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0); 1174 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block)); 1175 if (waitfor == MNT_WAIT) 1176 error = bwrite(bp); 1177 else 1178 bawrite(bp); 1179 1180 /* 1181 * The buffers for group descriptors, inode bitmaps and block bitmaps 1182 * are not busy at this point and are (hopefully) written by the 1183 * usual sync mechanism. No need to write them here 1184 */ 1185 1186 return (error); 1187} 1188 1189/* 1190 * Return the root of a filesystem. 1191 */ 1192static int 1193ext2_root(mp, vpp) 1194 struct mount *mp; 1195 struct vnode **vpp; 1196{ 1197 struct vnode *nvp; 1198 int error; 1199 1200 error = VFS_VGET(mp, (ino_t)ROOTINO, LK_EXCLUSIVE, &nvp); 1201 if (error) 1202 return (error); 1203 *vpp = nvp; 1204 return (0); 1205} 1206 1207static int 1208ext2_init(struct vfsconf *vfsp) 1209{ 1210 static int done; 1211 1212 if (done) 1213 return (0); 1214 done = 1; 1215 ext2_ihashinit(); 1216 1217 return (0); 1218} 1219