ext2_vfsops.c revision 12746
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 */ 41 42#if !defined(__FreeBSD__) 43#include "quota.h" 44#endif 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/namei.h> 49#include <sys/proc.h> 50#include <sys/kernel.h> 51#include <sys/vnode.h> 52#include <sys/socket.h> 53#include <sys/mount.h> 54#include <sys/buf.h> 55#include <sys/mbuf.h> 56#include <sys/file.h> 57#include <sys/disklabel.h> 58#include <sys/ioctl.h> 59#include <sys/errno.h> 60#include <sys/malloc.h> 61#include <sys/stat.h> 62 63#include <miscfs/specfs/specdev.h> 64 65#include <ufs/ufs/quota.h> 66#include <ufs/ufs/ufsmount.h> 67#include <ufs/ufs/inode.h> 68#include <ufs/ufs/ufs_extern.h> 69 70#include <gnu/ext2fs/fs.h> 71#include <gnu/ext2fs/ext2_extern.h> 72#include <gnu/ext2fs/ext2_fs.h> 73#include <gnu/ext2fs/ext2_fs_sb.h> 74 75int ext2_sbupdate __P((struct ufsmount *, int)); 76 77struct vfsops ext2fs_vfsops = { 78 ext2_mount, 79 ufs_start, /* empty function */ 80 ext2_unmount, 81 ufs_root, /* root inode via vget */ 82 ufs_quotactl, /* does operations associated with quotas */ 83 ext2_statfs, 84 ext2_sync, 85 ext2_vget, 86 ext2_fhtovp, 87 ext2_vptofh, 88 ext2_init, 89}; 90 91#if defined(__FreeBSD__) 92VFS_SET(ext2fs_vfsops, ext2fs, MOUNT_EXT2FS, 0); 93#define bsd_malloc malloc 94#define bsd_free free 95#endif 96 97extern u_long nextgennumber; 98#ifdef __FreeBSD__ 99int ext2fs_inode_hash_lock; 100#endif 101 102/* 103 * Called by main() when ufs is going to be mounted as root. 104 * 105 * Name is updated by mount(8) after booting. 106 */ 107#define ROOTNAME "root_device" 108 109static int compute_sb_data __P((struct vnode * devvp, 110 struct ext2_super_block * es, 111 struct ext2_sb_info * fs)); 112 113int 114ext2_mountroot() 115{ 116#if !defined(__FreeBSD__) 117 extern struct vnode *rootvp; 118#endif 119 register struct ext2_sb_info *fs; 120 register struct mount *mp; 121#if defined(__FreeBSD__) 122 struct proc *p = curproc; 123#else 124 struct proc *p = get_proc(); /* XXX */ 125#endif 126 struct ufsmount *ump; 127 u_int size; 128 int error; 129 130 /* 131 * Get vnodes for swapdev and rootdev. 132 */ 133 if (bdevvp(swapdev, &swapdev_vp) || bdevvp(rootdev, &rootvp)) 134 panic("ext2_mountroot: can't setup bdevvp's"); 135 136 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 137 bzero((char *)mp, (u_long)sizeof(struct mount)); 138 mp->mnt_op = &ext2fs_vfsops; 139 mp->mnt_flag = MNT_RDONLY; 140 if (error = ext2_mountfs(rootvp, mp, p)) { 141 bsd_free(mp, M_MOUNT); 142 return (error); 143 } 144 if (error = vfs_lock(mp)) { 145 (void)ext2_unmount(mp, 0, p); 146 bsd_free(mp, M_MOUNT); 147 return (error); 148 } 149#if defined(__FreeBSD__) 150 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 151#else 152 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 153#endif 154 mp->mnt_flag |= MNT_ROOTFS; 155 mp->mnt_vnodecovered = NULLVP; 156 ump = VFSTOUFS(mp); 157 fs = ump->um_e2fs; 158 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt)); 159 fs->fs_fsmnt[0] = '/'; 160 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 161 MNAMELEN); 162 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 163 &size); 164 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 165 (void)ext2_statfs(mp, &mp->mnt_stat, p); 166 vfs_unlock(mp); 167 inittodr(fs->s_es->s_wtime); /* this helps to set the time */ 168 return (0); 169} 170 171/* 172 * VFS Operations. 173 * 174 * mount system call 175 */ 176int 177ext2_mount(mp, path, data, ndp, p) 178 register struct mount *mp; 179 char *path; 180 caddr_t data; /* this is actually a (struct ufs_args *) */ 181 struct nameidata *ndp; 182 struct proc *p; 183{ 184 struct vnode *devvp; 185 struct ufs_args args; 186 struct ufsmount *ump = 0; 187 register struct ext2_sb_info *fs; 188 u_int size; 189 int error, flags; 190 191 if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) 192 return (error); 193 /* 194 * If updating, check whether changing from read-only to 195 * read/write; if there is no device name, that's all we do. 196 */ 197 if (mp->mnt_flag & MNT_UPDATE) { 198 ump = VFSTOUFS(mp); 199 fs = ump->um_e2fs; 200 error = 0; 201 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) { 202 flags = WRITECLOSE; 203 if (mp->mnt_flag & MNT_FORCE) 204 flags |= FORCECLOSE; 205 if (vfs_busy(mp)) 206 return (EBUSY); 207 error = ext2_flushfiles(mp, flags, p); 208 vfs_unbusy(mp); 209 } 210 if (!error && (mp->mnt_flag & MNT_RELOAD)) 211 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, p); 212 if (error) 213 return (error); 214 if (fs->s_rd_only && (mp->mnt_flag & MNT_WANTRDWR)) 215 fs->s_rd_only = 0; 216 if (fs->s_rd_only == 0) { 217 /* don't say it's clean */ 218 fs->s_es->s_state &= ~EXT2_VALID_FS; 219 ext2_sbupdate(ump, MNT_WAIT); 220 } 221 if (args.fspec == 0) { 222 /* 223 * Process export requests. 224 */ 225 return (vfs_export(mp, &ump->um_export, &args.export)); 226 } 227 } 228 /* 229 * Not an update, or updating the name: look up the name 230 * and verify that it refers to a sensible block device. 231 */ 232 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p); 233 if (error = namei(ndp)) 234 return (error); 235 devvp = ndp->ni_vp; 236 237 if (devvp->v_type != VBLK) { 238 vrele(devvp); 239 return (ENOTBLK); 240 } 241 if (major(devvp->v_rdev) >= nblkdev) { 242 vrele(devvp); 243 return (ENXIO); 244 } 245 if ((mp->mnt_flag & MNT_UPDATE) == 0) 246 error = ext2_mountfs(devvp, mp, p); 247 else { 248 if (devvp != ump->um_devvp) 249 error = EINVAL; /* needs translation */ 250 else 251 vrele(devvp); 252 } 253 if (error) { 254 vrele(devvp); 255 return (error); 256 } 257 ump = VFSTOUFS(mp); 258 fs = ump->um_e2fs; 259 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size); 260 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size); 261 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 262 MNAMELEN); 263 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 264 &size); 265 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 266 (void)ext2_statfs(mp, &mp->mnt_stat, p); 267 return (0); 268} 269 270/* 271 * checks that the data in the descriptor blocks make sense 272 * this is taken from ext2/super.c 273 */ 274static int ext2_check_descriptors (struct ext2_sb_info * sb) 275{ 276 int i; 277 int desc_block = 0; 278 unsigned long block = sb->s_es->s_first_data_block; 279 struct ext2_group_desc * gdp = NULL; 280 281 /* ext2_debug ("Checking group descriptors"); */ 282 283 for (i = 0; i < sb->s_groups_count; i++) 284 { 285 /* examine next descriptor block */ 286 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0) 287 gdp = (struct ext2_group_desc *) 288 sb->s_group_desc[desc_block++]->b_data; 289 if (gdp->bg_block_bitmap < block || 290 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 291 { 292 printf ("ext2_check_descriptors: " 293 "Block bitmap for group %d" 294 " not in group (block %lu)!", 295 i, (unsigned long) gdp->bg_block_bitmap); 296 return 0; 297 } 298 if (gdp->bg_inode_bitmap < block || 299 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 300 { 301 printf ("ext2_check_descriptors: " 302 "Inode bitmap for group %d" 303 " not in group (block %lu)!", 304 i, (unsigned long) gdp->bg_inode_bitmap); 305 return 0; 306 } 307 if (gdp->bg_inode_table < block || 308 gdp->bg_inode_table + sb->s_itb_per_group >= 309 block + EXT2_BLOCKS_PER_GROUP(sb)) 310 { 311 printf ("ext2_check_descriptors: " 312 "Inode table for group %d" 313 " not in group (block %lu)!", 314 i, (unsigned long) gdp->bg_inode_table); 315 return 0; 316 } 317 block += EXT2_BLOCKS_PER_GROUP(sb); 318 gdp++; 319 } 320 return 1; 321} 322 323/* 324 * this computes the fields of the ext2_sb_info structure from the 325 * data in the ext2_super_block structure read in 326 */ 327static int compute_sb_data(devvp, es, fs) 328 struct vnode * devvp; 329 struct ext2_super_block * es; 330 struct ext2_sb_info * fs; 331{ 332 int db_count, error; 333 int i, j; 334 int logic_sb_block = 1; /* XXX for now */ 335 336#if 1 337#define V(v) 338#else 339#define V(v) printf(#v"= %d\n", fs->v); 340#endif 341 342 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size; 343 V(s_blocksize) 344 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size; 345 V(s_bshift) 346 fs->s_fsbtodb = es->s_log_block_size + 1; 347 V(s_fsbtodb) 348 fs->s_qbmask = fs->s_blocksize - 1; 349 V(s_bmask) 350 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es); 351 V(s_blocksize_bits) 352 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size; 353 V(s_frag_size) 354 if (fs->s_frag_size) 355 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size; 356 V(s_frags_per_block) 357 fs->s_blocks_per_group = es->s_blocks_per_group; 358 V(s_blocks_per_group) 359 fs->s_frags_per_group = es->s_frags_per_group; 360 V(s_frags_per_group) 361 fs->s_inodes_per_group = es->s_inodes_per_group; 362 V(s_inodes_per_group) 363 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE; 364 V(s_inodes_per_block) 365 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block; 366 V(s_itb_per_group) 367 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc); 368 V(s_desc_per_block) 369 /* s_resuid / s_resgid ? */ 370 fs->s_groups_count = (es->s_blocks_count - 371 es->s_first_data_block + 372 EXT2_BLOCKS_PER_GROUP(fs) - 1) / 373 EXT2_BLOCKS_PER_GROUP(fs); 374 V(s_groups_count) 375 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) / 376 EXT2_DESC_PER_BLOCK(fs); 377 fs->s_db_per_group = db_count; 378 V(s_db_per_group) 379 380 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *), 381 M_UFSMNT, M_WAITOK); 382 383 /* adjust logic_sb_block */ 384 if(fs->s_blocksize > SBSIZE) 385 /* Godmar thinks: if the blocksize is greater than 1024, then 386 the superblock is logically part of block zero. 387 */ 388 logic_sb_block = 0; 389 390 for (i = 0; i < db_count; i++) { 391 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1), 392 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]); 393 if(error) { 394 for (j = 0; j < i; j++) 395 brelse(fs->s_group_desc[j]); 396 bsd_free(fs->s_group_desc, M_UFSMNT); 397 printf("EXT2-fs: unable to read group descriptors (%d)\n", error); 398 return EIO; 399 } 400 } 401 if(!ext2_check_descriptors(fs)) { 402 for (j = 0; j < db_count; j++) 403 brelse(fs->s_group_desc[j]); 404 bsd_free(fs->s_group_desc, M_UFSMNT); 405 printf("EXT2-fs: (ext2_check_descriptors failure) " 406 "unable to read group descriptors\n"); 407 return EIO; 408 } 409 410 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) { 411 fs->s_inode_bitmap_number[i] = 0; 412 fs->s_inode_bitmap[i] = NULL; 413 fs->s_block_bitmap_number[i] = 0; 414 fs->s_block_bitmap[i] = NULL; 415 } 416 fs->s_loaded_inode_bitmaps = 0; 417 fs->s_loaded_block_bitmaps = 0; 418 return 0; 419} 420 421/* 422 * Reload all incore data for a filesystem (used after running fsck on 423 * the root filesystem and finding things to fix). The filesystem must 424 * be mounted read-only. 425 * 426 * Things to do to update the mount: 427 * 1) invalidate all cached meta-data. 428 * 2) re-read superblock from disk. 429 * 3) re-read summary information from disk. 430 * 4) invalidate all inactive vnodes. 431 * 5) invalidate all cached file data. 432 * 6) re-read inode data for all active vnodes. 433 */ 434int 435ext2_reload(mountp, cred, p) 436 register struct mount *mountp; 437 struct ucred *cred; 438 struct proc *p; 439{ 440 register struct vnode *vp, *nvp, *devvp; 441 struct inode *ip; 442 struct buf *bp; 443 struct ext2_super_block * es; 444 struct ext2_sb_info *fs; 445 int error; 446 447 if ((mountp->mnt_flag & MNT_RDONLY) == 0) 448 return (EINVAL); 449 /* 450 * Step 1: invalidate all cached meta-data. 451 */ 452 devvp = VFSTOUFS(mountp)->um_devvp; 453 if (vinvalbuf(devvp, 0, cred, p, 0, 0)) 454 panic("ext2_reload: dirty1"); 455 /* 456 * Step 2: re-read superblock from disk. 457 * constants have been adjusted for ext2 458 */ 459 if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) 460 return (error); 461 es = (struct ext2_super_block *)bp->b_data; 462 if (es->s_magic != EXT2_SUPER_MAGIC) { 463 if(es->s_magic == EXT2_PRE_02B_MAGIC) 464 printf("This filesystem bears the magic number of a pre " 465 "0.2b version of ext2. This is not supported by " 466 "Lites.\n"); 467 else 468 printf("Wrong magic number: %x (expected %x for ext2 fs\n", 469 es->s_magic, EXT2_SUPER_MAGIC); 470 brelse(bp); 471 return (EIO); /* XXX needs translation */ 472 } 473 fs = VFSTOUFS(mountp)->um_e2fs; 474 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block)); 475 476 if(error = compute_sb_data(devvp, es, fs)) { 477 brelse(bp); 478 return error; 479 } 480#ifdef UNKLAR 481 if (fs->fs_sbsize < SBSIZE) 482 bp->b_flags |= B_INVAL; 483#endif 484 brelse(bp); 485 486loop: 487 for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 488 nvp = vp->v_mntvnodes.le_next; 489 /* 490 * Step 4: invalidate all inactive vnodes. 491 */ 492 if (vp->v_usecount == 0) { 493 vgone(vp); 494 continue; 495 } 496 /* 497 * Step 5: invalidate all cached file data. 498 */ 499 if (vget(vp, 1)) 500 goto loop; 501 if (vinvalbuf(vp, 0, cred, p, 0, 0)) 502 panic("ext2_reload: dirty2"); 503 /* 504 * Step 6: re-read inode data for all active vnodes. 505 */ 506 ip = VTOI(vp); 507 if (error = 508 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 509 (int)fs->s_blocksize, NOCRED, &bp)) { 510 vput(vp); 511 return (error); 512 } 513 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + 514 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)), 515 &ip->i_din); 516 brelse(bp); 517 vput(vp); 518 if (vp->v_mount != mountp) 519 goto loop; 520 } 521 return (0); 522} 523 524/* 525 * Common code for mount and mountroot 526 */ 527int 528ext2_mountfs(devvp, mp, p) 529 register struct vnode *devvp; 530 struct mount *mp; 531 struct proc *p; 532{ 533 register struct ufsmount *ump; 534 struct buf *bp; 535 register struct ext2_sb_info *fs; 536 struct ext2_super_block * es; 537 dev_t dev = devvp->v_rdev; 538 struct partinfo dpart; 539 int havepart = 0; 540 int error, i, size; 541 int ronly; 542#if !defined(__FreeBSD__) 543 extern struct vnode *rootvp; 544#endif 545 546 /* 547 * Disallow multiple mounts of the same device. 548 * Disallow mounting of a device that is currently in use 549 * (except for root, which might share swap device for miniroot). 550 * Flush out any old buffers remaining from a previous use. 551 */ 552 if (error = vfs_mountedon(devvp)) 553 return (error); 554 if (vcount(devvp) > 1 && devvp != rootvp) 555 return (EBUSY); 556 if (error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0)) 557 return (error); 558#ifdef READONLY 559/* turn on this to force it to be read-only */ 560 mp->mnt_flag |= MNT_RDONLY; 561#endif 562 563 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 564 if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p)) 565 return (error); 566 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0) 567 size = DEV_BSIZE; 568 else { 569 havepart = 1; 570 size = dpart.disklab->d_secsize; 571 } 572 573 bp = NULL; 574 ump = NULL; 575 if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) 576 goto out; 577 es = (struct ext2_super_block *)bp->b_data; 578 if (es->s_magic != EXT2_SUPER_MAGIC) { 579 if(es->s_magic == EXT2_PRE_02B_MAGIC) 580 printf("This filesystem bears the magic number of a pre " 581 "0.2b version of ext2. This is not supported by " 582 "Lites.\n"); 583 else 584 printf("Wrong magic number: %x (expected %x for EXT2FS)\n", 585 es->s_magic, EXT2_SUPER_MAGIC); 586 error = EINVAL; /* XXX needs translation */ 587 goto out; 588 } 589 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK); 590 bzero((caddr_t)ump, sizeof *ump); 591 /* I don't know whether this is the right strategy. Note that 592 we dynamically allocate both a ext2_sb_info and a ext2_super_block 593 while Linux keeps the super block in a locked buffer 594 */ 595 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info), 596 M_UFSMNT, M_WAITOK); 597 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block), 598 M_UFSMNT, M_WAITOK); 599 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block)); 600 if(error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)) { 601 brelse(bp); 602 return error; 603 } 604 brelse(bp); 605 bp = NULL; 606 fs = ump->um_e2fs; 607 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */ 608 if (!(fs->s_es->s_state & EXT2_VALID_FS)) { 609 printf("WARNING: %s was not properly dismounted\n", 610 fs->fs_fsmnt); 611 } 612 /* if the fs is not mounted read-only, make sure the super block is 613 always written back on a sync() 614 */ 615 if (ronly == 0) { 616 fs->s_dirt = 1; /* mark it modified */ 617 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */ 618 } 619 mp->mnt_data = (qaddr_t)ump; 620 mp->mnt_stat.f_fsid.val[0] = (long)dev; 621 mp->mnt_stat.f_fsid.val[1] = MOUNT_EXT2FS; 622 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; 623 mp->mnt_flag |= MNT_LOCAL; 624 ump->um_mountp = mp; 625 ump->um_dev = dev; 626 ump->um_devvp = devvp; 627 /* setting those two parameters allows us to use 628 ufs_bmap w/o changse ! 629 */ 630 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); 631 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1; 632 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); 633 for (i = 0; i < MAXQUOTAS; i++) 634 ump->um_quotas[i] = NULLVP; 635 devvp->v_specflags |= SI_MOUNTEDON; 636 if (ronly == 0) 637 ext2_sbupdate(ump, MNT_WAIT); 638 return (0); 639out: 640 if (bp) 641 brelse(bp); 642 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p); 643 if (ump) { 644 bsd_free(ump->um_fs, M_UFSMNT); 645 bsd_free(ump, M_UFSMNT); 646 mp->mnt_data = (qaddr_t)0; 647 } 648 return (error); 649} 650 651/* 652 * unmount system call 653 */ 654int 655ext2_unmount(mp, mntflags, p) 656 struct mount *mp; 657 int mntflags; 658 struct proc *p; 659{ 660 register struct ufsmount *ump; 661 register struct ext2_sb_info *fs; 662 int error, flags, ronly, i; 663 664 flags = 0; 665 if (mntflags & MNT_FORCE) { 666 if (mp->mnt_flag & MNT_ROOTFS) 667 return (EINVAL); 668 flags |= FORCECLOSE; 669 } 670 if (error = ext2_flushfiles(mp, flags, p)) 671 return (error); 672 ump = VFSTOUFS(mp); 673 fs = ump->um_e2fs; 674 ronly = fs->s_rd_only; 675 if (!ronly) { 676 fs->s_es->s_state |= EXT2_VALID_FS; /* was fs_clean = 1 */ 677 ext2_sbupdate(ump, MNT_WAIT); 678 } 679 /* release buffers containing group descriptors */ 680 for(i = 0; i < fs->s_db_per_group; i++) 681 brelse(fs->s_group_desc[i]); 682 /* release cached inode/block bitmaps */ 683 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 684 if (fs->s_inode_bitmap[i]) 685 brelse (fs->s_inode_bitmap[i]); 686 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 687 if (fs->s_block_bitmap[i]) 688 brelse (fs->s_block_bitmap[i]); 689 690 ump->um_devvp->v_specflags &= ~SI_MOUNTEDON; 691 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, 692 NOCRED, p); 693 vrele(ump->um_devvp); 694 bsd_free(fs->s_es, M_UFSMNT); 695 bsd_free(fs, M_UFSMNT); 696 bsd_free(ump, M_UFSMNT); 697 mp->mnt_data = (qaddr_t)0; 698 mp->mnt_flag &= ~MNT_LOCAL; 699 return (error); 700} 701 702/* 703 * Flush out all the files in a filesystem. 704 */ 705int 706ext2_flushfiles(mp, flags, p) 707 register struct mount *mp; 708 int flags; 709 struct proc *p; 710{ 711#if !defined(__FreeBSD__) 712 extern int doforce; 713#endif 714 register struct ufsmount *ump; 715 int error; 716#if QUOTA 717 int i; 718#endif 719 720 if (!doforce) 721 flags &= ~FORCECLOSE; 722 ump = VFSTOUFS(mp); 723#if QUOTA 724 if (mp->mnt_flag & MNT_QUOTA) { 725 if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) 726 return (error); 727 for (i = 0; i < MAXQUOTAS; i++) { 728 if (ump->um_quotas[i] == NULLVP) 729 continue; 730 quotaoff(p, mp, i); 731 } 732 /* 733 * Here we fall through to vflush again to ensure 734 * that we have gotten rid of all the system vnodes. 735 */ 736 } 737#endif 738 error = vflush(mp, NULLVP, flags); 739 return (error); 740} 741 742/* 743 * Get file system statistics. 744 * taken from ext2/super.c ext2_statfs 745 */ 746int 747ext2_statfs(mp, sbp, p) 748 struct mount *mp; 749 register struct statfs *sbp; 750 struct proc *p; 751{ 752 unsigned long overhead; 753 unsigned long overhead_per_group; 754 755 register struct ufsmount *ump; 756 register struct ext2_sb_info *fs; 757 register struct ext2_super_block *es; 758 759 ump = VFSTOUFS(mp); 760 fs = ump->um_e2fs; 761 es = fs->s_es; 762 763 if (es->s_magic != EXT2_SUPER_MAGIC) 764 panic("ext2_statfs - magic number spoiled"); 765 766 /* 767 * Compute the overhead (FS structures) 768 */ 769 overhead_per_group = 1 /* super block */ + 770 fs->s_db_per_group + 771 1 /* block bitmap */ + 772 1 /* inode bitmap */ + 773 fs->s_itb_per_group; 774 overhead = es->s_first_data_block + 775 fs->s_groups_count * overhead_per_group; 776 777 sbp->f_type = MOUNT_EXT2FS; 778 sbp->f_bsize = EXT2_FRAG_SIZE(fs); 779 sbp->f_iosize = EXT2_BLOCK_SIZE(fs); 780 sbp->f_blocks = es->s_blocks_count - overhead; 781 sbp->f_bfree = es->s_free_blocks_count; 782 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count; 783 sbp->f_files = es->s_inodes_count; 784 sbp->f_ffree = es->s_free_inodes_count; 785 if (sbp != &mp->mnt_stat) { 786 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 787 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 788 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 789 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 790 } 791 return (0); 792} 793 794/* 795 * Go through the disk queues to initiate sandbagged IO; 796 * go through the inodes to write those that have been modified; 797 * initiate the writing of the super block if it has been modified. 798 * 799 * Note: we are always called with the filesystem marked `MPBUSY'. 800 */ 801int 802ext2_sync(mp, waitfor, cred, p) 803 struct mount *mp; 804 int waitfor; 805 struct ucred *cred; 806 struct proc *p; 807{ 808 register struct vnode *vp; 809 register struct inode *ip; 810 register struct ufsmount *ump = VFSTOUFS(mp); 811 register struct ext2_sb_info *fs; 812 int error, allerror = 0; 813 814 fs = ump->um_e2fs; 815 /* 816 * Write back modified superblock. 817 * Consistency check that the superblock 818 * is still in the buffer cache. 819 */ 820 if (fs->s_dirt) { 821#if !defined(__FreeBSD__) 822 struct timeval time; 823#endif 824 825 if (fs->s_rd_only != 0) { /* XXX */ 826 printf("fs = %s\n", fs->fs_fsmnt); 827 panic("update: rofs mod"); 828 } 829 fs->s_dirt = 0; 830#if !defined(__FreeBSD__) 831 get_time(&time); 832#endif 833 fs->s_es->s_wtime = time.tv_sec; 834 allerror = ext2_sbupdate(ump, waitfor); 835 } 836 /* 837 * Write back each (modified) inode. 838 */ 839loop: 840 for (vp = mp->mnt_vnodelist.lh_first; 841 vp != NULL; 842 vp = vp->v_mntvnodes.le_next) { 843 /* 844 * If the vnode that we are about to sync is no longer 845 * associated with this mount point, start over. 846 */ 847 if (vp->v_mount != mp) 848 goto loop; 849 if (VOP_ISLOCKED(vp)) 850 continue; 851 ip = VTOI(vp); 852 if ((ip->i_flag & 853 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 854 vp->v_dirtyblkhd.lh_first == NULL) 855 continue; 856 if (vget(vp, 1)) 857 goto loop; 858 if (error = VOP_FSYNC(vp, cred, waitfor, p)) 859 allerror = error; 860 vput(vp); 861 } 862 /* 863 * Force stale file system control information to be flushed. 864 */ 865 if (error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) 866 allerror = error; 867#if QUOTA 868 qsync(mp); 869#endif 870 return (allerror); 871} 872 873/* 874 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it 875 * in from disk. If it is in core, wait for the lock bit to clear, then 876 * return the inode locked. Detection and handling of mount points must be 877 * done by the calling routine. 878 */ 879int 880ext2_vget(mp, ino, vpp) 881 struct mount *mp; 882 ino_t ino; 883 struct vnode **vpp; 884{ 885 register struct ext2_sb_info *fs; 886 register struct inode *ip; 887 struct ufsmount *ump; 888 struct buf *bp; 889 struct vnode *vp; 890 dev_t dev; 891 int i, type, error; 892 int used_blocks; 893 894 ump = VFSTOUFS(mp); 895 dev = ump->um_dev; 896restart: 897 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) 898 return (0); 899 900#ifdef __FreeBSD__ 901 /* 902 * Lock out the creation of new entries in the FFS hash table in 903 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 904 * may occur! 905 */ 906 if (ext2fs_inode_hash_lock) { 907 while (ext2fs_inode_hash_lock) { 908 ext2fs_inode_hash_lock = -1; 909 tsleep(&ext2fs_inode_hash_lock, PVM, "ffsvgt", 0); 910 } 911 goto restart; 912 } 913 ext2fs_inode_hash_lock = 1; 914#endif 915 916 /* Allocate a new vnode/inode. */ 917 if (error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) { 918 *vpp = NULL; 919 return (error); 920 } 921 /* I don't really know what this 'type' does. I suppose it's some kind 922 * of memory accounting. Let's just book this memory on FFS's account 923 * If I'm not mistaken, this stuff isn't implemented anyway in Lites 924 */ 925 type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */ 926 MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK); 927#ifndef __FreeBSD__ 928 insmntque(vp, mp); 929#endif 930 bzero((caddr_t)ip, sizeof(struct inode)); 931 vp->v_data = ip; 932 ip->i_vnode = vp; 933 ip->i_e2fs = fs = ump->um_e2fs; 934 ip->i_dev = dev; 935 ip->i_number = ino; 936#if QUOTA 937 for (i = 0; i < MAXQUOTAS; i++) 938 ip->i_dquot[i] = NODQUOT; 939#endif 940 /* 941 * Put it onto its hash chain and lock it so that other requests for 942 * this inode will block if they arrive while we are sleeping waiting 943 * for old data structures to be purged or for the contents of the 944 * disk portion of this inode to be read. 945 */ 946 ufs_ihashins(ip); 947 948#ifdef __FreeBSD__ 949 if (ext2fs_inode_hash_lock < 0) 950 wakeup(&ext2fs_inode_hash_lock); 951 ext2fs_inode_hash_lock = 0; 952#endif 953 954 /* Read in the disk contents for the inode, copy into the inode. */ 955 /* Read in the disk contents for the inode, copy into the inode. */ 956#if 0 957printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino))); 958#endif 959 if (error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 960 (int)fs->s_blocksize, NOCRED, &bp)) { 961 /* 962 * The inode does not contain anything useful, so it would 963 * be misleading to leave it on its hash chain. With mode 964 * still zero, it will be unlinked and returned to the free 965 * list by vput(). 966 */ 967 vput(vp); 968 brelse(bp); 969 *vpp = NULL; 970 return (error); 971 } 972 /* convert ext2 inode to dinode */ 973 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE * 974 ino_to_fsbo(fs, ino)), &ip->i_din); 975 ip->i_block_group = ino_to_cg(fs, ino); 976 ip->i_next_alloc_block = 0; 977 ip->i_next_alloc_goal = 0; 978 ip->i_prealloc_count = 0; 979 ip->i_prealloc_block = 0; 980 /* now we want to make sure that block pointers for unused 981 blocks are zeroed out - ext2_balloc depends on this 982 although for regular files and directories only 983 */ 984 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) { 985 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize; 986 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) 987 ip->i_db[i] = 0; 988 } 989/* 990 ext2_print_inode(ip); 991*/ 992 brelse(bp); 993 994 /* 995 * Initialize the vnode from the inode, check for aliases. 996 * Note that the underlying vnode may have changed. 997 */ 998 if (error = ufs_vinit(mp, ext2_specop_p, EXT2_FIFOOPS, &vp)) { 999 vput(vp); 1000 *vpp = NULL; 1001 return (error); 1002 } 1003 /* 1004 * Finish inode initialization now that aliasing has been resolved. 1005 */ 1006 ip->i_devvp = ump->um_devvp; 1007 VREF(ip->i_devvp); 1008 /* 1009 * Set up a generation number for this inode if it does not 1010 * already have one. This should only happen on old filesystems. 1011 */ 1012 if (ip->i_gen == 0) { 1013#if !defined(__FreeBSD__) 1014 struct timeval time; 1015 get_time(&time); 1016#endif 1017 if (++nextgennumber < (u_long)time.tv_sec) 1018 nextgennumber = time.tv_sec; 1019 ip->i_gen = nextgennumber; 1020 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1021 ip->i_flag |= IN_MODIFIED; 1022 } 1023 *vpp = vp; 1024 return (0); 1025} 1026 1027/* 1028 * File handle to vnode 1029 * 1030 * Have to be really careful about stale file handles: 1031 * - check that the inode number is valid 1032 * - call ext2_vget() to get the locked inode 1033 * - check for an unallocated inode (i_mode == 0) 1034 * - check that the given client host has export rights and return 1035 * those rights via. exflagsp and credanonp 1036 */ 1037int 1038ext2_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) 1039 register struct mount *mp; 1040 struct fid *fhp; 1041 struct mbuf *nam; 1042 struct vnode **vpp; 1043 int *exflagsp; 1044 struct ucred **credanonp; 1045{ 1046 register struct ufid *ufhp; 1047 struct ext2_sb_info *fs; 1048 1049 ufhp = (struct ufid *)fhp; 1050 fs = VFSTOUFS(mp)->um_e2fs; 1051 if (ufhp->ufid_ino < ROOTINO || 1052 ufhp->ufid_ino >= fs->s_groups_count * fs->s_es->s_inodes_per_group) 1053 return (ESTALE); 1054 return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp)); 1055} 1056 1057/* 1058 * Vnode pointer to File handle 1059 */ 1060/* ARGSUSED */ 1061int 1062ext2_vptofh(vp, fhp) 1063 struct vnode *vp; 1064 struct fid *fhp; 1065{ 1066 register struct inode *ip; 1067 register struct ufid *ufhp; 1068 1069 ip = VTOI(vp); 1070 ufhp = (struct ufid *)fhp; 1071 ufhp->ufid_len = sizeof(struct ufid); 1072 ufhp->ufid_ino = ip->i_number; 1073 ufhp->ufid_gen = ip->i_gen; 1074 return (0); 1075} 1076 1077/* 1078 * Write a superblock and associated information back to disk. 1079 */ 1080int 1081ext2_sbupdate(mp, waitfor) 1082 struct ufsmount *mp; 1083 int waitfor; 1084{ 1085 register struct ext2_sb_info *fs = mp->um_e2fs; 1086 register struct ext2_super_block *es = fs->s_es; 1087 register struct buf *bp; 1088 int i, error = 0; 1089/* 1090printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no"); 1091*/ 1092 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0); 1093 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block)); 1094 if (waitfor == MNT_WAIT) 1095 error = bwrite(bp); 1096 else 1097 bawrite(bp); 1098 1099 /* write group descriptors back on disk */ 1100 for(i = 0; i < fs->s_db_per_group; i++) 1101 /* Godmar thinks: we must avoid using any of the b*write 1102 * functions here: we want to keep the buffer locked 1103 * so we use my 'housemade' write routine: 1104 */ 1105 error |= ll_w_block(fs->s_group_desc[i], waitfor == MNT_WAIT); 1106 1107 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 1108 if (fs->s_inode_bitmap[i]) 1109 ll_w_block (fs->s_inode_bitmap[i], 1); 1110 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 1111 if (fs->s_block_bitmap[i]) 1112 ll_w_block (fs->s_block_bitmap[i], 1); 1113 1114 return (error); 1115} 1116