ext2_vfsops.c revision 93014
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 93014 2002-03-23 13:10:13Z bde $ 41 */ 42 43#include "opt_quota.h" 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/namei.h> 48#include <sys/proc.h> 49#include <sys/kernel.h> 50#include <sys/vnode.h> 51#include <sys/mount.h> 52#include <sys/bio.h> 53#include <sys/buf.h> 54#include <sys/conf.h> 55#include <sys/fcntl.h> 56#include <sys/malloc.h> 57#include <sys/stat.h> 58#include <sys/mutex.h> 59 60#include <ufs/ufs/extattr.h> 61#include <ufs/ufs/quota.h> 62#include <ufs/ufs/ufsmount.h> 63#include <ufs/ufs/inode.h> 64#include <ufs/ufs/ufs_extern.h> 65 66 67#include <gnu/ext2fs/fs.h> 68#include <gnu/ext2fs/ext2_extern.h> 69#include <gnu/ext2fs/ext2_fs.h> 70#include <gnu/ext2fs/ext2_fs_sb.h> 71 72static int ext2_fhtovp(struct mount *, struct fid *, struct vnode **); 73static int ext2_flushfiles(struct mount *mp, int flags, struct thread *td); 74static int ext2_mount(struct mount *, 75 char *, caddr_t, struct nameidata *, struct thread *); 76static int ext2_mountfs(struct vnode *, struct mount *, struct thread *); 77static int ext2_reload(struct mount *mountp, struct ucred *cred, 78 struct thread *td); 79static int ext2_sbupdate(struct ufsmount *, int); 80static int ext2_statfs(struct mount *, struct statfs *, struct thread *); 81static int ext2_sync(struct mount *, int, struct ucred *, struct thread *); 82static int ext2_unmount(struct mount *, int, struct thread *); 83static int ext2_vget(struct mount *, ino_t, int, struct vnode **); 84static int ext2_vptofh(struct vnode *, struct fid *); 85 86static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part"); 87 88static struct vfsops ext2fs_vfsops = { 89 ext2_mount, 90 ufs_start, /* empty function */ 91 ext2_unmount, 92 ufs_root, /* root inode via vget */ 93 ufs_quotactl, /* does operations associated with quotas */ 94 ext2_statfs, 95 ext2_sync, 96 ext2_vget, 97 ext2_fhtovp, 98 vfs_stdcheckexp, 99 ext2_vptofh, 100 ext2_init, 101 vfs_stduninit, 102 vfs_stdextattrctl, 103}; 104 105VFS_SET(ext2fs_vfsops, ext2fs, 0); 106#define bsd_malloc malloc 107#define bsd_free free 108 109static int ext2fs_inode_hash_lock; 110 111static int ext2_check_sb_compat(struct ext2_super_block *es, dev_t dev, 112 int ronly); 113static int compute_sb_data(struct vnode * devvp, 114 struct ext2_super_block * es, struct ext2_sb_info * fs); 115 116#ifdef notyet 117static int ext2_mountroot(void); 118 119/* 120 * Called by main() when ext2fs is going to be mounted as root. 121 * 122 * Name is updated by mount(8) after booting. 123 */ 124#define ROOTNAME "root_device" 125 126static int 127ext2_mountroot() 128{ 129 register struct ext2_sb_info *fs; 130 register struct mount *mp; 131 struct thread *td = curthread; 132 struct ufsmount *ump; 133 u_int size; 134 int error; 135 136 if ((error = bdevvp(rootdev, &rootvp))) { 137 printf("ext2_mountroot: can't find rootvp\n"); 138 return (error); 139 } 140 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 141 bzero((char *)mp, (u_long)sizeof(struct mount)); 142 TAILQ_INIT(&mp->mnt_nvnodelist); 143 TAILQ_INIT(&mp->mnt_reservedvnlist); 144 mp->mnt_op = &ext2fs_vfsops; 145 mp->mnt_flag = MNT_RDONLY; 146 if (error = ext2_mountfs(rootvp, mp, td)) { 147 bsd_free(mp, M_MOUNT); 148 return (error); 149 } 150 if (error = vfs_lock(mp)) { 151 (void)ext2_unmount(mp, 0, td); 152 bsd_free(mp, M_MOUNT); 153 return (error); 154 } 155 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 156 mp->mnt_flag |= MNT_ROOTFS; 157 mp->mnt_vnodecovered = NULLVP; 158 ump = VFSTOUFS(mp); 159 fs = ump->um_e2fs; 160 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt)); 161 fs->fs_fsmnt[0] = '/'; 162 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 163 MNAMELEN); 164 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 165 &size); 166 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 167 (void)ext2_statfs(mp, &mp->mnt_stat, td); 168 vfs_unlock(mp); 169 inittodr(fs->s_es->s_wtime); /* this helps to set the time */ 170 return (0); 171} 172#endif 173 174/* 175 * VFS Operations. 176 * 177 * mount system call 178 */ 179static int 180ext2_mount(mp, path, data, ndp, td) 181 register struct mount *mp; 182 char *path; 183 caddr_t data; /* this is actually a (struct ufs_args *) */ 184 struct nameidata *ndp; 185 struct thread *td; 186{ 187 struct vnode *devvp; 188 struct ufs_args args; 189 struct ufsmount *ump = 0; 190 register struct ext2_sb_info *fs; 191 size_t size; 192 int error, flags; 193 mode_t accessmode; 194 195 /* Double-check the length of path.. */ 196 if (strlen(path) >= MAXMNTLEN - 1) 197 return (ENAMETOOLONG); 198 error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args)); 199 if (error != 0) 200 return (error); 201 /* 202 * If updating, check whether changing from read-only to 203 * read/write; if there is no device name, that's all we do. 204 */ 205 if (mp->mnt_flag & MNT_UPDATE) { 206 ump = VFSTOUFS(mp); 207 fs = ump->um_e2fs; 208 error = 0; 209 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) { 210 flags = WRITECLOSE; 211 if (mp->mnt_flag & MNT_FORCE) 212 flags |= FORCECLOSE; 213 if (vfs_busy(mp, LK_NOWAIT, 0, td)) 214 return (EBUSY); 215 error = ext2_flushfiles(mp, flags, td); 216 vfs_unbusy(mp, td); 217 if (!error && fs->s_wasvalid) { 218 fs->s_es->s_state |= EXT2_VALID_FS; 219 ext2_sbupdate(ump, MNT_WAIT); 220 } 221 fs->s_rd_only = 1; 222 } 223 if (!error && (mp->mnt_flag & MNT_RELOAD)) 224 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, td); 225 if (error) 226 return (error); 227 devvp = ump->um_devvp; 228 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev, 229 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0) 230 return (EPERM); 231 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { 232 /* 233 * If upgrade to read-write by non-root, then verify 234 * that user has necessary permissions on the device. 235 */ 236 if (suser_td(td)) { 237 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 238 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE, 239 td->td_ucred, td)) != 0) { 240 VOP_UNLOCK(devvp, 0, td); 241 return (error); 242 } 243 VOP_UNLOCK(devvp, 0, td); 244 } 245 246 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 || 247 (fs->s_es->s_state & EXT2_ERROR_FS)) { 248 if (mp->mnt_flag & MNT_FORCE) { 249 printf( 250"WARNING: %s was not properly dismounted\n", 251 fs->fs_fsmnt); 252 } else { 253 printf( 254"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", 255 fs->fs_fsmnt); 256 return (EPERM); 257 } 258 } 259 fs->s_es->s_state &= ~EXT2_VALID_FS; 260 ext2_sbupdate(ump, MNT_WAIT); 261 fs->s_rd_only = 0; 262 } 263 if (args.fspec == 0) { 264 /* 265 * Process export requests. 266 */ 267 return (vfs_export(mp, &args.export)); 268 } 269 } 270 /* 271 * Not an update, or updating the name: look up the name 272 * and verify that it refers to a sensible block device. 273 */ 274 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, td); 275 if ((error = namei(ndp)) != 0) 276 return (error); 277 NDFREE(ndp, NDF_ONLY_PNBUF); 278 devvp = ndp->ni_vp; 279 280 if (!vn_isdisk(devvp, &error)) { 281 vrele(devvp); 282 return (error); 283 } 284 285 /* 286 * If mount by non-root, then verify that user has necessary 287 * permissions on the device. 288 */ 289 if (suser_td(td)) { 290 accessmode = VREAD; 291 if ((mp->mnt_flag & MNT_RDONLY) == 0) 292 accessmode |= VWRITE; 293 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 294 if ((error = VOP_ACCESS(devvp, accessmode, td->td_ucred, td)) != 0) { 295 vput(devvp); 296 return (error); 297 } 298 VOP_UNLOCK(devvp, 0, td); 299 } 300 301 if ((mp->mnt_flag & MNT_UPDATE) == 0) { 302 error = ext2_mountfs(devvp, mp, td); 303 } else { 304 if (devvp != ump->um_devvp) 305 error = EINVAL; /* needs translation */ 306 else 307 vrele(devvp); 308 } 309 if (error) { 310 vrele(devvp); 311 return (error); 312 } 313 ump = VFSTOUFS(mp); 314 fs = ump->um_e2fs; 315 /* 316 * Note that this strncpy() is ok because of a check at the start 317 * of ext2_mount(). 318 */ 319 strncpy(fs->fs_fsmnt, path, MAXMNTLEN); 320 fs->fs_fsmnt[MAXMNTLEN - 1] = '\0'; 321 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 322 &size); 323 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 324 (void)ext2_statfs(mp, &mp->mnt_stat, td); 325 return (0); 326} 327 328/* 329 * checks that the data in the descriptor blocks make sense 330 * this is taken from ext2/super.c 331 */ 332static int ext2_check_descriptors (struct ext2_sb_info * sb) 333{ 334 int i; 335 int desc_block = 0; 336 unsigned long block = sb->s_es->s_first_data_block; 337 struct ext2_group_desc * gdp = NULL; 338 339 /* ext2_debug ("Checking group descriptors"); */ 340 341 for (i = 0; i < sb->s_groups_count; i++) 342 { 343 /* examine next descriptor block */ 344 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0) 345 gdp = (struct ext2_group_desc *) 346 sb->s_group_desc[desc_block++]->b_data; 347 if (gdp->bg_block_bitmap < block || 348 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 349 { 350 printf ("ext2_check_descriptors: " 351 "Block bitmap for group %d" 352 " not in group (block %lu)!\n", 353 i, (unsigned long) gdp->bg_block_bitmap); 354 return 0; 355 } 356 if (gdp->bg_inode_bitmap < block || 357 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 358 { 359 printf ("ext2_check_descriptors: " 360 "Inode bitmap for group %d" 361 " not in group (block %lu)!\n", 362 i, (unsigned long) gdp->bg_inode_bitmap); 363 return 0; 364 } 365 if (gdp->bg_inode_table < block || 366 gdp->bg_inode_table + sb->s_itb_per_group >= 367 block + EXT2_BLOCKS_PER_GROUP(sb)) 368 { 369 printf ("ext2_check_descriptors: " 370 "Inode table for group %d" 371 " not in group (block %lu)!\n", 372 i, (unsigned long) gdp->bg_inode_table); 373 return 0; 374 } 375 block += EXT2_BLOCKS_PER_GROUP(sb); 376 gdp++; 377 } 378 return 1; 379} 380 381static int 382ext2_check_sb_compat(es, dev, ronly) 383 struct ext2_super_block *es; 384 dev_t dev; 385 int ronly; 386{ 387 388 if (es->s_magic != EXT2_SUPER_MAGIC) { 389 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n", 390 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC); 391 return (1); 392 } 393 if (es->s_rev_level > EXT2_GOOD_OLD_REV) { 394 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) { 395 printf( 396"WARNING: mount of %s denied due to unsupported optional features\n", 397 devtoname(dev)); 398 return (1); 399 } 400 if (!ronly && 401 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) { 402 printf( 403"WARNING: R/W mount of %s denied due to unsupported optional features\n", 404 devtoname(dev)); 405 return (1); 406 } 407 } 408 return (0); 409} 410 411/* 412 * this computes the fields of the ext2_sb_info structure from the 413 * data in the ext2_super_block structure read in 414 */ 415static int compute_sb_data(devvp, es, fs) 416 struct vnode * devvp; 417 struct ext2_super_block * es; 418 struct ext2_sb_info * fs; 419{ 420 int db_count, error; 421 int i, j; 422 int logic_sb_block = 1; /* XXX for now */ 423 424#if 1 425#define V(v) 426#else 427#define V(v) printf(#v"= %d\n", fs->v); 428#endif 429 430 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size; 431 V(s_blocksize) 432 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size; 433 V(s_bshift) 434 fs->s_fsbtodb = es->s_log_block_size + 1; 435 V(s_fsbtodb) 436 fs->s_qbmask = fs->s_blocksize - 1; 437 V(s_bmask) 438 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es); 439 V(s_blocksize_bits) 440 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size; 441 V(s_frag_size) 442 if (fs->s_frag_size) 443 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size; 444 V(s_frags_per_block) 445 fs->s_blocks_per_group = es->s_blocks_per_group; 446 V(s_blocks_per_group) 447 fs->s_frags_per_group = es->s_frags_per_group; 448 V(s_frags_per_group) 449 fs->s_inodes_per_group = es->s_inodes_per_group; 450 V(s_inodes_per_group) 451 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE; 452 V(s_inodes_per_block) 453 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block; 454 V(s_itb_per_group) 455 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc); 456 V(s_desc_per_block) 457 /* s_resuid / s_resgid ? */ 458 fs->s_groups_count = (es->s_blocks_count - 459 es->s_first_data_block + 460 EXT2_BLOCKS_PER_GROUP(fs) - 1) / 461 EXT2_BLOCKS_PER_GROUP(fs); 462 V(s_groups_count) 463 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) / 464 EXT2_DESC_PER_BLOCK(fs); 465 fs->s_db_per_group = db_count; 466 V(s_db_per_group) 467 468 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *), 469 M_UFSMNT, M_WAITOK); 470 471 /* adjust logic_sb_block */ 472 if(fs->s_blocksize > SBSIZE) 473 /* Godmar thinks: if the blocksize is greater than 1024, then 474 the superblock is logically part of block zero. 475 */ 476 logic_sb_block = 0; 477 478 for (i = 0; i < db_count; i++) { 479 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1), 480 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]); 481 if(error) { 482 for (j = 0; j < i; j++) 483 brelse(fs->s_group_desc[j]); 484 bsd_free(fs->s_group_desc, M_UFSMNT); 485 printf("EXT2-fs: unable to read group descriptors (%d)\n", error); 486 return EIO; 487 } 488 /* Set the B_LOCKED flag on the buffer, then brelse() it */ 489 LCK_BUF(fs->s_group_desc[i]) 490 } 491 if(!ext2_check_descriptors(fs)) { 492 for (j = 0; j < db_count; j++) 493 ULCK_BUF(fs->s_group_desc[j]) 494 bsd_free(fs->s_group_desc, M_UFSMNT); 495 printf("EXT2-fs: (ext2_check_descriptors failure) " 496 "unable to read group descriptors\n"); 497 return EIO; 498 } 499 500 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) { 501 fs->s_inode_bitmap_number[i] = 0; 502 fs->s_inode_bitmap[i] = NULL; 503 fs->s_block_bitmap_number[i] = 0; 504 fs->s_block_bitmap[i] = NULL; 505 } 506 fs->s_loaded_inode_bitmaps = 0; 507 fs->s_loaded_block_bitmaps = 0; 508 return 0; 509} 510 511/* 512 * Reload all incore data for a filesystem (used after running fsck on 513 * the root filesystem and finding things to fix). The filesystem must 514 * be mounted read-only. 515 * 516 * Things to do to update the mount: 517 * 1) invalidate all cached meta-data. 518 * 2) re-read superblock from disk. 519 * 3) re-read summary information from disk. 520 * 4) invalidate all inactive vnodes. 521 * 5) invalidate all cached file data. 522 * 6) re-read inode data for all active vnodes. 523 */ 524static int 525ext2_reload(mountp, cred, td) 526 register struct mount *mountp; 527 struct ucred *cred; 528 struct thread *td; 529{ 530 register struct vnode *vp, *nvp, *devvp; 531 struct inode *ip; 532 struct buf *bp; 533 struct ext2_super_block * es; 534 struct ext2_sb_info *fs; 535 int error; 536 537 if ((mountp->mnt_flag & MNT_RDONLY) == 0) 538 return (EINVAL); 539 /* 540 * Step 1: invalidate all cached meta-data. 541 */ 542 devvp = VFSTOUFS(mountp)->um_devvp; 543 if (vinvalbuf(devvp, 0, cred, td, 0, 0)) 544 panic("ext2_reload: dirty1"); 545 /* 546 * Step 2: re-read superblock from disk. 547 * constants have been adjusted for ext2 548 */ 549 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 550 return (error); 551 es = (struct ext2_super_block *)bp->b_data; 552 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) { 553 brelse(bp); 554 return (EIO); /* XXX needs translation */ 555 } 556 fs = VFSTOUFS(mountp)->um_e2fs; 557 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block)); 558 559 if((error = compute_sb_data(devvp, es, fs)) != 0) { 560 brelse(bp); 561 return error; 562 } 563#ifdef UNKLAR 564 if (fs->fs_sbsize < SBSIZE) 565 bp->b_flags |= B_INVAL; 566#endif 567 brelse(bp); 568 569loop: 570 mtx_lock(&mntvnode_mtx); 571 for (vp = TAILQ_FIRST(&mountp->mnt_nvnodelist); vp != NULL; vp = nvp) { 572 if (vp->v_mount != mountp) { 573 mtx_unlock(&mntvnode_mtx); 574 goto loop; 575 } 576 nvp = TAILQ_NEXT(vp, v_nmntvnodes); 577 mtx_unlock(&mntvnode_mtx); 578 /* 579 * Step 4: invalidate all inactive vnodes. 580 */ 581 if (vrecycle(vp, NULL, td)) 582 goto loop; 583 /* 584 * Step 5: invalidate all cached file data. 585 */ 586 mtx_lock(&vp->v_interlock); 587 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) { 588 goto loop; 589 } 590 if (vinvalbuf(vp, 0, cred, td, 0, 0)) 591 panic("ext2_reload: dirty2"); 592 /* 593 * Step 6: re-read inode data for all active vnodes. 594 */ 595 ip = VTOI(vp); 596 error = 597 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 598 (int)fs->s_blocksize, NOCRED, &bp); 599 if (error) { 600 vput(vp); 601 return (error); 602 } 603 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + 604 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)), 605 &ip->i_din); 606 brelse(bp); 607 vput(vp); 608 mtx_lock(&mntvnode_mtx); 609 } 610 mtx_unlock(&mntvnode_mtx); 611 return (0); 612} 613 614/* 615 * Common code for mount and mountroot 616 */ 617static int 618ext2_mountfs(devvp, mp, td) 619 register struct vnode *devvp; 620 struct mount *mp; 621 struct thread *td; 622{ 623 register struct ufsmount *ump; 624 struct buf *bp; 625 register struct ext2_sb_info *fs; 626 struct ext2_super_block * es; 627 dev_t dev = devvp->v_rdev; 628 int error, i; 629 int ronly; 630 631 /* 632 * Disallow multiple mounts of the same device. 633 * Disallow mounting of a device that is currently in use 634 * (except for root, which might share swap device for miniroot). 635 * Flush out any old buffers remaining from a previous use. 636 */ 637 if ((error = vfs_mountedon(devvp)) != 0) 638 return (error); 639 if (vcount(devvp) > 1 && devvp != rootvp) 640 return (EBUSY); 641 if ((error = vinvalbuf(devvp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) 642 return (error); 643#ifdef READONLY 644/* turn on this to force it to be read-only */ 645 mp->mnt_flag |= MNT_RDONLY; 646#endif 647 648 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 649 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 650 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, td); 651 VOP_UNLOCK(devvp, 0, td); 652 if (error) 653 return (error); 654 bp = NULL; 655 ump = NULL; 656 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 657 goto out; 658 es = (struct ext2_super_block *)bp->b_data; 659 if (ext2_check_sb_compat(es, dev, ronly) != 0) { 660 error = EINVAL; /* XXX needs translation */ 661 goto out; 662 } 663 if ((es->s_state & EXT2_VALID_FS) == 0 || 664 (es->s_state & EXT2_ERROR_FS)) { 665 if (ronly || (mp->mnt_flag & MNT_FORCE)) { 666 printf( 667"WARNING: Filesystem was not properly dismounted\n"); 668 } else { 669 printf( 670"WARNING: R/W mount denied. Filesystem is not clean - run fsck\n"); 671 error = EPERM; 672 goto out; 673 } 674 } 675 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK); 676 bzero((caddr_t)ump, sizeof *ump); 677 ump->um_malloctype = M_EXT2NODE; 678 ump->um_blkatoff = ext2_blkatoff; 679 ump->um_truncate = ext2_truncate; 680 ump->um_update = ext2_update; 681 ump->um_valloc = ext2_valloc; 682 ump->um_vfree = ext2_vfree; 683 /* I don't know whether this is the right strategy. Note that 684 we dynamically allocate both a ext2_sb_info and a ext2_super_block 685 while Linux keeps the super block in a locked buffer 686 */ 687 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info), 688 M_UFSMNT, M_WAITOK); 689 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block), 690 M_UFSMNT, M_WAITOK); 691 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block)); 692 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs))) 693 goto out; 694 /* 695 * We don't free the group descriptors allocated by compute_sb_data() 696 * until ext2_unmount(). This is OK since the mount will succeed. 697 */ 698 brelse(bp); 699 bp = NULL; 700 fs = ump->um_e2fs; 701 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */ 702 /* if the fs is not mounted read-only, make sure the super block is 703 always written back on a sync() 704 */ 705 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0; 706 if (ronly == 0) { 707 fs->s_dirt = 1; /* mark it modified */ 708 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */ 709 } 710 mp->mnt_data = (qaddr_t)ump; 711 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev); 712 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; 713 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; 714 mp->mnt_flag |= MNT_LOCAL; 715 ump->um_mountp = mp; 716 ump->um_dev = dev; 717 ump->um_devvp = devvp; 718 /* setting those two parameters allows us to use 719 ufs_bmap w/o changse ! 720 */ 721 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); 722 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1; 723 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); 724 for (i = 0; i < MAXQUOTAS; i++) 725 ump->um_quotas[i] = NULLVP; 726 devvp->v_rdev->si_mountpoint = mp; 727 if (ronly == 0) 728 ext2_sbupdate(ump, MNT_WAIT); 729 return (0); 730out: 731 if (bp) 732 brelse(bp); 733 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, td); 734 if (ump) { 735 bsd_free(ump->um_e2fs->s_es, M_UFSMNT); 736 bsd_free(ump->um_e2fs, M_UFSMNT); 737 bsd_free(ump, M_UFSMNT); 738 mp->mnt_data = (qaddr_t)0; 739 } 740 return (error); 741} 742 743/* 744 * unmount system call 745 */ 746static int 747ext2_unmount(mp, mntflags, td) 748 struct mount *mp; 749 int mntflags; 750 struct thread *td; 751{ 752 register struct ufsmount *ump; 753 register struct ext2_sb_info *fs; 754 int error, flags, ronly, i; 755 756 flags = 0; 757 if (mntflags & MNT_FORCE) { 758 if (mp->mnt_flag & MNT_ROOTFS) 759 return (EINVAL); 760 flags |= FORCECLOSE; 761 } 762 if ((error = ext2_flushfiles(mp, flags, td)) != 0) 763 return (error); 764 ump = VFSTOUFS(mp); 765 fs = ump->um_e2fs; 766 ronly = fs->s_rd_only; 767 if (ronly == 0) { 768 if (fs->s_wasvalid) 769 fs->s_es->s_state |= EXT2_VALID_FS; 770 ext2_sbupdate(ump, MNT_WAIT); 771 } 772 773 /* release buffers containing group descriptors */ 774 for(i = 0; i < fs->s_db_per_group; i++) 775 ULCK_BUF(fs->s_group_desc[i]) 776 bsd_free(fs->s_group_desc, M_UFSMNT); 777 778 /* release cached inode/block bitmaps */ 779 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 780 if (fs->s_inode_bitmap[i]) 781 ULCK_BUF(fs->s_inode_bitmap[i]) 782 783 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 784 if (fs->s_block_bitmap[i]) 785 ULCK_BUF(fs->s_block_bitmap[i]) 786 787 ump->um_devvp->v_rdev->si_mountpoint = NULL; 788 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, 789 NOCRED, td); 790 vrele(ump->um_devvp); 791 bsd_free(fs->s_es, M_UFSMNT); 792 bsd_free(fs, M_UFSMNT); 793 bsd_free(ump, M_UFSMNT); 794 mp->mnt_data = (qaddr_t)0; 795 mp->mnt_flag &= ~MNT_LOCAL; 796 return (error); 797} 798 799/* 800 * Flush out all the files in a filesystem. 801 */ 802static int 803ext2_flushfiles(mp, flags, td) 804 register struct mount *mp; 805 int flags; 806 struct thread *td; 807{ 808 register struct ufsmount *ump; 809 int error; 810#if QUOTA 811 int i; 812#endif 813 814 ump = VFSTOUFS(mp); 815#if QUOTA 816 if (mp->mnt_flag & MNT_QUOTA) { 817 if ((error = vflush(mp, 0, SKIPSYSTEM|flags)) != 0) 818 return (error); 819 for (i = 0; i < MAXQUOTAS; i++) { 820 if (ump->um_quotas[i] == NULLVP) 821 continue; 822 quotaoff(td, mp, i); 823 } 824 /* 825 * Here we fall through to vflush again to ensure 826 * that we have gotten rid of all the system vnodes. 827 */ 828 } 829#endif 830 error = vflush(mp, 0, flags); 831 return (error); 832} 833 834/* 835 * Get file system statistics. 836 * taken from ext2/super.c ext2_statfs 837 */ 838static int 839ext2_statfs(mp, sbp, td) 840 struct mount *mp; 841 register struct statfs *sbp; 842 struct thread *td; 843{ 844 unsigned long overhead; 845 register struct ufsmount *ump; 846 register struct ext2_sb_info *fs; 847 register struct ext2_super_block *es; 848 int i, nsb; 849 850 ump = VFSTOUFS(mp); 851 fs = ump->um_e2fs; 852 es = fs->s_es; 853 854 if (es->s_magic != EXT2_SUPER_MAGIC) 855 panic("ext2_statfs - magic number spoiled"); 856 857 /* 858 * Compute the overhead (FS structures) 859 */ 860 if (es->s_feature_ro_compat & EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER) { 861 nsb = 0; 862 for (i = 0 ; i < fs->s_groups_count; i++) 863 if (ext2_group_sparse(i)) 864 nsb++; 865 } else 866 nsb = fs->s_groups_count; 867 overhead = es->s_first_data_block + 868 /* Superblocks and block group descriptors: */ 869 nsb * (1 + fs->s_db_per_group) + 870 /* Inode bitmap, block bitmap, and inode table: */ 871 fs->s_groups_count * (1 + 1 + fs->s_itb_per_group); 872 873 sbp->f_bsize = EXT2_FRAG_SIZE(fs); 874 sbp->f_iosize = EXT2_BLOCK_SIZE(fs); 875 sbp->f_blocks = es->s_blocks_count - overhead; 876 sbp->f_bfree = es->s_free_blocks_count; 877 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count; 878 sbp->f_files = es->s_inodes_count; 879 sbp->f_ffree = es->s_free_inodes_count; 880 if (sbp != &mp->mnt_stat) { 881 sbp->f_type = mp->mnt_vfc->vfc_typenum; 882 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 883 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 884 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 885 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 886 } 887 return (0); 888} 889 890/* 891 * Go through the disk queues to initiate sandbagged IO; 892 * go through the inodes to write those that have been modified; 893 * initiate the writing of the super block if it has been modified. 894 * 895 * Note: we are always called with the filesystem marked `MPBUSY'. 896 */ 897static int 898ext2_sync(mp, waitfor, cred, td) 899 struct mount *mp; 900 int waitfor; 901 struct ucred *cred; 902 struct thread *td; 903{ 904 struct vnode *nvp, *vp; 905 struct inode *ip; 906 struct ufsmount *ump = VFSTOUFS(mp); 907 struct ext2_sb_info *fs; 908 int error, allerror = 0; 909 910 fs = ump->um_e2fs; 911 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */ 912 printf("fs = %s\n", fs->fs_fsmnt); 913 panic("ext2_sync: rofs mod"); 914 } 915 /* 916 * Write back each (modified) inode. 917 */ 918 mtx_lock(&mntvnode_mtx); 919loop: 920 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) { 921 /* 922 * If the vnode that we are about to sync is no longer 923 * associated with this mount point, start over. 924 */ 925 if (vp->v_mount != mp) 926 goto loop; 927 nvp = TAILQ_NEXT(vp, v_nmntvnodes); 928 mtx_unlock(&mntvnode_mtx); 929 mtx_lock(&vp->v_interlock); 930 ip = VTOI(vp); 931 if (vp->v_type == VNON || 932 ((ip->i_flag & 933 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 934 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || waitfor == MNT_LAZY))) { 935 mtx_unlock(&vp->v_interlock); 936 mtx_lock(&mntvnode_mtx); 937 continue; 938 } 939 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td); 940 if (error) { 941 mtx_lock(&mntvnode_mtx); 942 if (error == ENOENT) 943 goto loop; 944 continue; 945 } 946 if ((error = VOP_FSYNC(vp, cred, waitfor, td)) != 0) 947 allerror = error; 948 VOP_UNLOCK(vp, 0, td); 949 vrele(vp); 950 mtx_lock(&mntvnode_mtx); 951 } 952 mtx_unlock(&mntvnode_mtx); 953 /* 954 * Force stale file system control information to be flushed. 955 */ 956 if (waitfor != MNT_LAZY) { 957 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td); 958 if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, td)) != 0) 959 allerror = error; 960 VOP_UNLOCK(ump->um_devvp, 0, td); 961 } 962#if QUOTA 963 qsync(mp); 964#endif 965 /* 966 * Write back modified superblock. 967 */ 968 if (fs->s_dirt != 0) { 969 fs->s_dirt = 0; 970 fs->s_es->s_wtime = time_second; 971 if ((error = ext2_sbupdate(ump, waitfor)) != 0) 972 allerror = error; 973 } 974 return (allerror); 975} 976 977/* 978 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it 979 * in from disk. If it is in core, wait for the lock bit to clear, then 980 * return the inode locked. Detection and handling of mount points must be 981 * done by the calling routine. 982 */ 983static int 984ext2_vget(mp, ino, flags, vpp) 985 struct mount *mp; 986 ino_t ino; 987 int flags; 988 struct vnode **vpp; 989{ 990 register struct ext2_sb_info *fs; 991 register struct inode *ip; 992 struct ufsmount *ump; 993 struct buf *bp; 994 struct vnode *vp; 995 dev_t dev; 996 int i, error; 997 int used_blocks; 998 999 ump = VFSTOUFS(mp); 1000 dev = ump->um_dev; 1001restart: 1002 if ((error = ufs_ihashget(dev, ino, flags, vpp)) != 0) 1003 return (error); 1004 if (*vpp != NULL) 1005 return (0); 1006 1007 /* 1008 * Lock out the creation of new entries in the FFS hash table in 1009 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 1010 * may occur! 1011 */ 1012 if (ext2fs_inode_hash_lock) { 1013 while (ext2fs_inode_hash_lock) { 1014 ext2fs_inode_hash_lock = -1; 1015 tsleep(&ext2fs_inode_hash_lock, PVM, "e2vget", 0); 1016 } 1017 goto restart; 1018 } 1019 ext2fs_inode_hash_lock = 1; 1020 1021 /* 1022 * If this MALLOC() is performed after the getnewvnode() 1023 * it might block, leaving a vnode with a NULL v_data to be 1024 * found by ext2_sync() if a sync happens to fire right then, 1025 * which will cause a panic because ext2_sync() blindly 1026 * dereferences vp->v_data (as well it should). 1027 */ 1028 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK); 1029 1030 /* Allocate a new vnode/inode. */ 1031 if ((error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) != 0) { 1032 if (ext2fs_inode_hash_lock < 0) 1033 wakeup(&ext2fs_inode_hash_lock); 1034 ext2fs_inode_hash_lock = 0; 1035 *vpp = NULL; 1036 FREE(ip, M_EXT2NODE); 1037 return (error); 1038 } 1039 bzero((caddr_t)ip, sizeof(struct inode)); 1040 lockinit(&vp->v_lock, PINOD, "ext2in", 0, 0); 1041 vp->v_data = ip; 1042 ip->i_vnode = vp; 1043 ip->i_e2fs = fs = ump->um_e2fs; 1044 ip->i_dev = dev; 1045 ip->i_number = ino; 1046#if QUOTA 1047 for (i = 0; i < MAXQUOTAS; i++) 1048 ip->i_dquot[i] = NODQUOT; 1049#endif 1050 /* 1051 * Put it onto its hash chain and lock it so that other requests for 1052 * this inode will block if they arrive while we are sleeping waiting 1053 * for old data structures to be purged or for the contents of the 1054 * disk portion of this inode to be read. 1055 */ 1056 ufs_ihashins(ip); 1057 1058 if (ext2fs_inode_hash_lock < 0) 1059 wakeup(&ext2fs_inode_hash_lock); 1060 ext2fs_inode_hash_lock = 0; 1061 1062 /* Read in the disk contents for the inode, copy into the inode. */ 1063#if 0 1064printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino))); 1065#endif 1066 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 1067 (int)fs->s_blocksize, NOCRED, &bp)) != 0) { 1068 /* 1069 * The inode does not contain anything useful, so it would 1070 * be misleading to leave it on its hash chain. With mode 1071 * still zero, it will be unlinked and returned to the free 1072 * list by vput(). 1073 */ 1074 vput(vp); 1075 brelse(bp); 1076 *vpp = NULL; 1077 return (error); 1078 } 1079 /* convert ext2 inode to dinode */ 1080 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE * 1081 ino_to_fsbo(fs, ino)), &ip->i_din); 1082 ip->i_block_group = ino_to_cg(fs, ino); 1083 ip->i_next_alloc_block = 0; 1084 ip->i_next_alloc_goal = 0; 1085 ip->i_prealloc_count = 0; 1086 ip->i_prealloc_block = 0; 1087 /* now we want to make sure that block pointers for unused 1088 blocks are zeroed out - ext2_balloc depends on this 1089 although for regular files and directories only 1090 */ 1091 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) { 1092 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize; 1093 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) 1094 ip->i_db[i] = 0; 1095 } 1096/* 1097 ext2_print_inode(ip); 1098*/ 1099 brelse(bp); 1100 1101 /* 1102 * Initialize the vnode from the inode, check for aliases. 1103 * Note that the underlying vnode may have changed. 1104 */ 1105 if ((error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) != 0) { 1106 vput(vp); 1107 *vpp = NULL; 1108 return (error); 1109 } 1110 /* 1111 * Finish inode initialization now that aliasing has been resolved. 1112 */ 1113 ip->i_devvp = ump->um_devvp; 1114 VREF(ip->i_devvp); 1115 /* 1116 * Set up a generation number for this inode if it does not 1117 * already have one. This should only happen on old filesystems. 1118 */ 1119 if (ip->i_gen == 0) { 1120 ip->i_gen = random() / 2 + 1; 1121 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1122 ip->i_flag |= IN_MODIFIED; 1123 } 1124 *vpp = vp; 1125 return (0); 1126} 1127 1128/* 1129 * File handle to vnode 1130 * 1131 * Have to be really careful about stale file handles: 1132 * - check that the inode number is valid 1133 * - call ext2_vget() to get the locked inode 1134 * - check for an unallocated inode (i_mode == 0) 1135 * - check that the given client host has export rights and return 1136 * those rights via. exflagsp and credanonp 1137 */ 1138static int 1139ext2_fhtovp(mp, fhp, vpp) 1140 register struct mount *mp; 1141 struct fid *fhp; 1142 struct vnode **vpp; 1143{ 1144 register struct ufid *ufhp; 1145 struct ext2_sb_info *fs; 1146 1147 ufhp = (struct ufid *)fhp; 1148 fs = VFSTOUFS(mp)->um_e2fs; 1149 if (ufhp->ufid_ino < ROOTINO || 1150 ufhp->ufid_ino >= fs->s_groups_count * fs->s_es->s_inodes_per_group) 1151 return (ESTALE); 1152 return (ufs_fhtovp(mp, ufhp, vpp)); 1153} 1154 1155/* 1156 * Vnode pointer to File handle 1157 */ 1158/* ARGSUSED */ 1159static int 1160ext2_vptofh(vp, fhp) 1161 struct vnode *vp; 1162 struct fid *fhp; 1163{ 1164 register struct inode *ip; 1165 register struct ufid *ufhp; 1166 1167 ip = VTOI(vp); 1168 ufhp = (struct ufid *)fhp; 1169 ufhp->ufid_len = sizeof(struct ufid); 1170 ufhp->ufid_ino = ip->i_number; 1171 ufhp->ufid_gen = ip->i_gen; 1172 return (0); 1173} 1174 1175/* 1176 * Write a superblock and associated information back to disk. 1177 */ 1178static int 1179ext2_sbupdate(mp, waitfor) 1180 struct ufsmount *mp; 1181 int waitfor; 1182{ 1183 register struct ext2_sb_info *fs = mp->um_e2fs; 1184 register struct ext2_super_block *es = fs->s_es; 1185 register struct buf *bp; 1186 int error = 0; 1187/* 1188printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no"); 1189*/ 1190 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0); 1191 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block)); 1192 if (waitfor == MNT_WAIT) 1193 error = bwrite(bp); 1194 else 1195 bawrite(bp); 1196 1197 /* 1198 * The buffers for group descriptors, inode bitmaps and block bitmaps 1199 * are not busy at this point and are (hopefully) written by the 1200 * usual sync mechanism. No need to write them here 1201 */ 1202 1203 return (error); 1204} 1205