ext2_vfsops.c revision 53059
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 53059 1999-11-09 14:15:33Z phk $ 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/buf.h> 53#include <sys/conf.h> 54#include <sys/fcntl.h> 55#include <sys/disklabel.h> 56#include <sys/malloc.h> 57#include <sys/stat.h> 58 59#include <ufs/ufs/quota.h> 60#include <ufs/ufs/ufsmount.h> 61#include <ufs/ufs/inode.h> 62#include <ufs/ufs/ufs_extern.h> 63 64#include <gnu/ext2fs/fs.h> 65#include <gnu/ext2fs/ext2_extern.h> 66#include <gnu/ext2fs/ext2_fs.h> 67#include <gnu/ext2fs/ext2_fs_sb.h> 68 69static int ext2_fhtovp __P((struct mount *, struct fid *, struct vnode **)); 70static int ext2_flushfiles __P((struct mount *mp, int flags, struct proc *p)); 71static int ext2_mount __P((struct mount *, 72 char *, caddr_t, struct nameidata *, struct proc *)); 73static int ext2_mountfs __P((struct vnode *, struct mount *, struct proc *)); 74static int ext2_reload __P((struct mount *mountp, struct ucred *cred, 75 struct proc *p)); 76static int ext2_sbupdate __P((struct ufsmount *, int)); 77static int ext2_statfs __P((struct mount *, struct statfs *, struct proc *)); 78static int ext2_sync __P((struct mount *, int, struct ucred *, struct proc *)); 79static int ext2_unmount __P((struct mount *, int, struct proc *)); 80static int ext2_vget __P((struct mount *, ino_t, struct vnode **)); 81static int ext2_vptofh __P((struct vnode *, struct fid *)); 82 83static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part"); 84 85static struct vfsops ext2fs_vfsops = { 86 ext2_mount, 87 ufs_start, /* empty function */ 88 ext2_unmount, 89 ufs_root, /* root inode via vget */ 90 ufs_quotactl, /* does operations associated with quotas */ 91 ext2_statfs, 92 ext2_sync, 93 ext2_vget, 94 ext2_fhtovp, 95 ufs_check_export, 96 ext2_vptofh, 97 ext2_init, 98}; 99 100VFS_SET(ext2fs_vfsops, ext2fs, 0); 101#define bsd_malloc malloc 102#define bsd_free free 103 104static int ext2fs_inode_hash_lock; 105 106static int compute_sb_data __P((struct vnode * devvp, 107 struct ext2_super_block * es, 108 struct ext2_sb_info * fs)); 109 110#ifdef notyet 111static int ext2_mountroot __P((void)); 112 113/* 114 * Called by main() when ext2fs is going to be mounted as root. 115 * 116 * Name is updated by mount(8) after booting. 117 */ 118#define ROOTNAME "root_device" 119 120static int 121ext2_mountroot() 122{ 123 register struct ext2_sb_info *fs; 124 register struct mount *mp; 125 struct proc *p = curproc; 126 struct ufsmount *ump; 127 u_int size; 128 int error; 129 130 if ((error = bdevvp(rootdev, &rootvp))) { 131 printf("ext2_mountroot: can't find rootvp\n"); 132 return (error); 133 } 134 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 135 bzero((char *)mp, (u_long)sizeof(struct mount)); 136 mp->mnt_op = &ext2fs_vfsops; 137 mp->mnt_flag = MNT_RDONLY; 138 if (error = ext2_mountfs(rootvp, mp, p)) { 139 bsd_free(mp, M_MOUNT); 140 return (error); 141 } 142 if (error = vfs_lock(mp)) { 143 (void)ext2_unmount(mp, 0, p); 144 bsd_free(mp, M_MOUNT); 145 return (error); 146 } 147 CIRCLEQ_INSERT_HEAD(&mountlist, mp, mnt_list); 148 mp->mnt_flag |= MNT_ROOTFS; 149 mp->mnt_vnodecovered = NULLVP; 150 ump = VFSTOUFS(mp); 151 fs = ump->um_e2fs; 152 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt)); 153 fs->fs_fsmnt[0] = '/'; 154 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 155 MNAMELEN); 156 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 157 &size); 158 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 159 (void)ext2_statfs(mp, &mp->mnt_stat, p); 160 vfs_unlock(mp); 161 inittodr(fs->s_es->s_wtime); /* this helps to set the time */ 162 return (0); 163} 164#endif 165 166/* 167 * VFS Operations. 168 * 169 * mount system call 170 */ 171static int 172ext2_mount(mp, path, data, ndp, p) 173 register struct mount *mp; 174 char *path; 175 caddr_t data; /* this is actually a (struct ufs_args *) */ 176 struct nameidata *ndp; 177 struct proc *p; 178{ 179 struct vnode *devvp; 180 struct ufs_args args; 181 struct ufsmount *ump = 0; 182 register struct ext2_sb_info *fs; 183 u_int size; 184 int error, flags; 185 mode_t accessmode; 186 187 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) != 0) 188 return (error); 189 /* 190 * If updating, check whether changing from read-only to 191 * read/write; if there is no device name, that's all we do. 192 */ 193 if (mp->mnt_flag & MNT_UPDATE) { 194 ump = VFSTOUFS(mp); 195 fs = ump->um_e2fs; 196 error = 0; 197 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) { 198 flags = WRITECLOSE; 199 if (mp->mnt_flag & MNT_FORCE) 200 flags |= FORCECLOSE; 201 if (vfs_busy(mp, LK_NOWAIT, 0, p)) 202 return (EBUSY); 203 error = ext2_flushfiles(mp, flags, p); 204 vfs_unbusy(mp, p); 205 if (!error && fs->s_wasvalid) { 206 fs->s_es->s_state |= EXT2_VALID_FS; 207 ext2_sbupdate(ump, MNT_WAIT); 208 } 209 fs->s_rd_only = 1; 210 } 211 if (!error && (mp->mnt_flag & MNT_RELOAD)) 212 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, p); 213 if (error) 214 return (error); 215 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { 216 /* 217 * If upgrade to read-write by non-root, then verify 218 * that user has necessary permissions on the device. 219 */ 220 if (p->p_ucred->cr_uid != 0) { 221 devvp = ump->um_devvp; 222 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 223 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE, 224 p->p_ucred, p)) != 0) { 225 VOP_UNLOCK(devvp, 0, p); 226 return (error); 227 } 228 VOP_UNLOCK(devvp, 0, p); 229 } 230 231 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 || 232 (fs->s_es->s_state & EXT2_ERROR_FS)) { 233 if (mp->mnt_flag & MNT_FORCE) { 234 printf( 235"WARNING: %s was not properly dismounted\n", 236 fs->fs_fsmnt); 237 } else { 238 printf( 239"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", 240 fs->fs_fsmnt); 241 return (EPERM); 242 } 243 } 244 fs->s_es->s_state &= ~EXT2_VALID_FS; 245 ext2_sbupdate(ump, MNT_WAIT); 246 fs->s_rd_only = 0; 247 } 248 if (args.fspec == 0) { 249 /* 250 * Process export requests. 251 */ 252 return (vfs_export(mp, &ump->um_export, &args.export)); 253 } 254 } 255 /* 256 * Not an update, or updating the name: look up the name 257 * and verify that it refers to a sensible block device. 258 */ 259 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p); 260 if ((error = namei(ndp)) != 0) 261 return (error); 262 devvp = ndp->ni_vp; 263 264 if (!vn_isdisk(devvp)) { 265 vrele(devvp); 266 return (ENOTBLK); 267 } 268 269 /* 270 * If mount by non-root, then verify that user has necessary 271 * permissions on the device. 272 */ 273 if (p->p_ucred->cr_uid != 0) { 274 accessmode = VREAD; 275 if ((mp->mnt_flag & MNT_RDONLY) == 0) 276 accessmode |= VWRITE; 277 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 278 if ((error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p)) != 0) { 279 vput(devvp); 280 return (error); 281 } 282 VOP_UNLOCK(devvp, 0, p); 283 } 284 285 if ((mp->mnt_flag & MNT_UPDATE) == 0) { 286 error = ext2_mountfs(devvp, mp, p); 287 } else { 288 if (devvp != ump->um_devvp) 289 error = EINVAL; /* needs translation */ 290 else 291 vrele(devvp); 292 } 293 if (error) { 294 vrele(devvp); 295 return (error); 296 } 297 ump = VFSTOUFS(mp); 298 fs = ump->um_e2fs; 299 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size); 300 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size); 301 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 302 MNAMELEN); 303 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 304 &size); 305 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 306 (void)ext2_statfs(mp, &mp->mnt_stat, p); 307 return (0); 308} 309 310/* 311 * checks that the data in the descriptor blocks make sense 312 * this is taken from ext2/super.c 313 */ 314static int ext2_check_descriptors (struct ext2_sb_info * sb) 315{ 316 int i; 317 int desc_block = 0; 318 unsigned long block = sb->s_es->s_first_data_block; 319 struct ext2_group_desc * gdp = NULL; 320 321 /* ext2_debug ("Checking group descriptors"); */ 322 323 for (i = 0; i < sb->s_groups_count; i++) 324 { 325 /* examine next descriptor block */ 326 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0) 327 gdp = (struct ext2_group_desc *) 328 sb->s_group_desc[desc_block++]->b_data; 329 if (gdp->bg_block_bitmap < block || 330 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 331 { 332 printf ("ext2_check_descriptors: " 333 "Block bitmap for group %d" 334 " not in group (block %lu)!\n", 335 i, (unsigned long) gdp->bg_block_bitmap); 336 return 0; 337 } 338 if (gdp->bg_inode_bitmap < block || 339 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 340 { 341 printf ("ext2_check_descriptors: " 342 "Inode bitmap for group %d" 343 " not in group (block %lu)!\n", 344 i, (unsigned long) gdp->bg_inode_bitmap); 345 return 0; 346 } 347 if (gdp->bg_inode_table < block || 348 gdp->bg_inode_table + sb->s_itb_per_group >= 349 block + EXT2_BLOCKS_PER_GROUP(sb)) 350 { 351 printf ("ext2_check_descriptors: " 352 "Inode table for group %d" 353 " not in group (block %lu)!\n", 354 i, (unsigned long) gdp->bg_inode_table); 355 return 0; 356 } 357 block += EXT2_BLOCKS_PER_GROUP(sb); 358 gdp++; 359 } 360 return 1; 361} 362 363/* 364 * this computes the fields of the ext2_sb_info structure from the 365 * data in the ext2_super_block structure read in 366 */ 367static int compute_sb_data(devvp, es, fs) 368 struct vnode * devvp; 369 struct ext2_super_block * es; 370 struct ext2_sb_info * fs; 371{ 372 int db_count, error; 373 int i, j; 374 int logic_sb_block = 1; /* XXX for now */ 375 376#if 1 377#define V(v) 378#else 379#define V(v) printf(#v"= %d\n", fs->v); 380#endif 381 382 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size; 383 V(s_blocksize) 384 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size; 385 V(s_bshift) 386 fs->s_fsbtodb = es->s_log_block_size + 1; 387 V(s_fsbtodb) 388 fs->s_qbmask = fs->s_blocksize - 1; 389 V(s_bmask) 390 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es); 391 V(s_blocksize_bits) 392 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size; 393 V(s_frag_size) 394 if (fs->s_frag_size) 395 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size; 396 V(s_frags_per_block) 397 fs->s_blocks_per_group = es->s_blocks_per_group; 398 V(s_blocks_per_group) 399 fs->s_frags_per_group = es->s_frags_per_group; 400 V(s_frags_per_group) 401 fs->s_inodes_per_group = es->s_inodes_per_group; 402 V(s_inodes_per_group) 403 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE; 404 V(s_inodes_per_block) 405 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block; 406 V(s_itb_per_group) 407 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc); 408 V(s_desc_per_block) 409 /* s_resuid / s_resgid ? */ 410 fs->s_groups_count = (es->s_blocks_count - 411 es->s_first_data_block + 412 EXT2_BLOCKS_PER_GROUP(fs) - 1) / 413 EXT2_BLOCKS_PER_GROUP(fs); 414 V(s_groups_count) 415 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) / 416 EXT2_DESC_PER_BLOCK(fs); 417 fs->s_db_per_group = db_count; 418 V(s_db_per_group) 419 420 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *), 421 M_UFSMNT, M_WAITOK); 422 423 /* adjust logic_sb_block */ 424 if(fs->s_blocksize > SBSIZE) 425 /* Godmar thinks: if the blocksize is greater than 1024, then 426 the superblock is logically part of block zero. 427 */ 428 logic_sb_block = 0; 429 430 for (i = 0; i < db_count; i++) { 431 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1), 432 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]); 433 if(error) { 434 for (j = 0; j < i; j++) 435 brelse(fs->s_group_desc[j]); 436 bsd_free(fs->s_group_desc, M_UFSMNT); 437 printf("EXT2-fs: unable to read group descriptors (%d)\n", error); 438 return EIO; 439 } 440 /* Set the B_LOCKED flag on the buffer, then brelse() it */ 441 LCK_BUF(fs->s_group_desc[i]) 442 } 443 if(!ext2_check_descriptors(fs)) { 444 for (j = 0; j < db_count; j++) 445 ULCK_BUF(fs->s_group_desc[j]) 446 bsd_free(fs->s_group_desc, M_UFSMNT); 447 printf("EXT2-fs: (ext2_check_descriptors failure) " 448 "unable to read group descriptors\n"); 449 return EIO; 450 } 451 452 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) { 453 fs->s_inode_bitmap_number[i] = 0; 454 fs->s_inode_bitmap[i] = NULL; 455 fs->s_block_bitmap_number[i] = 0; 456 fs->s_block_bitmap[i] = NULL; 457 } 458 fs->s_loaded_inode_bitmaps = 0; 459 fs->s_loaded_block_bitmaps = 0; 460 return 0; 461} 462 463/* 464 * Reload all incore data for a filesystem (used after running fsck on 465 * the root filesystem and finding things to fix). The filesystem must 466 * be mounted read-only. 467 * 468 * Things to do to update the mount: 469 * 1) invalidate all cached meta-data. 470 * 2) re-read superblock from disk. 471 * 3) re-read summary information from disk. 472 * 4) invalidate all inactive vnodes. 473 * 5) invalidate all cached file data. 474 * 6) re-read inode data for all active vnodes. 475 */ 476static int 477ext2_reload(mountp, cred, p) 478 register struct mount *mountp; 479 struct ucred *cred; 480 struct proc *p; 481{ 482 register struct vnode *vp, *nvp, *devvp; 483 struct inode *ip; 484 struct buf *bp; 485 struct ext2_super_block * es; 486 struct ext2_sb_info *fs; 487 int error; 488 489 if ((mountp->mnt_flag & MNT_RDONLY) == 0) 490 return (EINVAL); 491 /* 492 * Step 1: invalidate all cached meta-data. 493 */ 494 devvp = VFSTOUFS(mountp)->um_devvp; 495 if (vinvalbuf(devvp, 0, cred, p, 0, 0)) 496 panic("ext2_reload: dirty1"); 497 /* 498 * Step 2: re-read superblock from disk. 499 * constants have been adjusted for ext2 500 */ 501 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 502 return (error); 503 es = (struct ext2_super_block *)bp->b_data; 504 if (es->s_magic != EXT2_SUPER_MAGIC) { 505 if(es->s_magic == EXT2_PRE_02B_MAGIC) 506 printf("This filesystem bears the magic number of a pre " 507 "0.2b version of ext2. This is not supported by " 508 "Lites.\n"); 509 else 510 printf("Wrong magic number: %x (expected %x for ext2 fs\n", 511 es->s_magic, EXT2_SUPER_MAGIC); 512 brelse(bp); 513 return (EIO); /* XXX needs translation */ 514 } 515 fs = VFSTOUFS(mountp)->um_e2fs; 516 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block)); 517 518 if((error = compute_sb_data(devvp, es, fs)) != 0) { 519 brelse(bp); 520 return error; 521 } 522#ifdef UNKLAR 523 if (fs->fs_sbsize < SBSIZE) 524 bp->b_flags |= B_INVAL; 525#endif 526 brelse(bp); 527 528loop: 529 simple_lock(&mntvnode_slock); 530 for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 531 if (vp->v_mount != mountp) { 532 simple_unlock(&mntvnode_slock); 533 goto loop; 534 } 535 nvp = vp->v_mntvnodes.le_next; 536 /* 537 * Step 4: invalidate all inactive vnodes. 538 */ 539 if (vrecycle(vp, &mntvnode_slock, p)) 540 goto loop; 541 /* 542 * Step 5: invalidate all cached file data. 543 */ 544 simple_lock(&vp->v_interlock); 545 simple_unlock(&mntvnode_slock); 546 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 547 goto loop; 548 } 549 if (vinvalbuf(vp, 0, cred, p, 0, 0)) 550 panic("ext2_reload: dirty2"); 551 /* 552 * Step 6: re-read inode data for all active vnodes. 553 */ 554 ip = VTOI(vp); 555 error = 556 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 557 (int)fs->s_blocksize, NOCRED, &bp); 558 if (error) { 559 vput(vp); 560 return (error); 561 } 562 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + 563 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)), 564 &ip->i_din); 565 brelse(bp); 566 vput(vp); 567 simple_lock(&mntvnode_slock); 568 } 569 simple_unlock(&mntvnode_slock); 570 return (0); 571} 572 573/* 574 * Common code for mount and mountroot 575 */ 576static int 577ext2_mountfs(devvp, mp, p) 578 register struct vnode *devvp; 579 struct mount *mp; 580 struct proc *p; 581{ 582 register struct ufsmount *ump; 583 struct buf *bp; 584 register struct ext2_sb_info *fs; 585 struct ext2_super_block * es; 586 dev_t dev = devvp->v_rdev; 587 struct partinfo dpart; 588 int havepart = 0; 589 int error, i, size; 590 int ronly; 591 592 /* 593 * Disallow multiple mounts of the same device. 594 * Disallow mounting of a device that is currently in use 595 * (except for root, which might share swap device for miniroot). 596 * Flush out any old buffers remaining from a previous use. 597 */ 598 if ((error = vfs_mountedon(devvp)) != 0) 599 return (error); 600 if (vcount(devvp) > 1 && devvp != rootvp) 601 return (EBUSY); 602 if ((error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) 603 return (error); 604#ifdef READONLY 605/* turn on this to force it to be read-only */ 606 mp->mnt_flag |= MNT_RDONLY; 607#endif 608 609 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 610 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 611 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p); 612 VOP_UNLOCK(devvp, 0, p); 613 if (error) 614 return (error); 615 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0) 616 size = DEV_BSIZE; 617 else { 618 havepart = 1; 619 size = dpart.disklab->d_secsize; 620 } 621 622 bp = NULL; 623 ump = NULL; 624 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 625 goto out; 626 es = (struct ext2_super_block *)bp->b_data; 627 if (es->s_magic != EXT2_SUPER_MAGIC) { 628 if(es->s_magic == EXT2_PRE_02B_MAGIC) 629 printf("This filesystem bears the magic number of a pre " 630 "0.2b version of ext2. This is not supported by " 631 "Lites.\n"); 632 else 633 printf("Wrong magic number: %x (expected %x for EXT2FS)\n", 634 es->s_magic, EXT2_SUPER_MAGIC); 635 error = EINVAL; /* XXX needs translation */ 636 goto out; 637 } 638 if ((es->s_state & EXT2_VALID_FS) == 0 || 639 (es->s_state & EXT2_ERROR_FS)) { 640 if (ronly || (mp->mnt_flag & MNT_FORCE)) { 641 printf( 642"WARNING: Filesystem was not properly dismounted\n"); 643 } else { 644 printf( 645"WARNING: R/W mount denied. Filesystem is not clean - run fsck\n"); 646 error = EPERM; 647 goto out; 648 } 649 } 650 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK); 651 bzero((caddr_t)ump, sizeof *ump); 652 ump->um_malloctype = M_EXT2NODE; 653 ump->um_blkatoff = ext2_blkatoff; 654 ump->um_truncate = ext2_truncate; 655 ump->um_update = ext2_update; 656 ump->um_valloc = ext2_valloc; 657 ump->um_vfree = ext2_vfree; 658 /* I don't know whether this is the right strategy. Note that 659 we dynamically allocate both a ext2_sb_info and a ext2_super_block 660 while Linux keeps the super block in a locked buffer 661 */ 662 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info), 663 M_UFSMNT, M_WAITOK); 664 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block), 665 M_UFSMNT, M_WAITOK); 666 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block)); 667 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs))) 668 goto out; 669 /* 670 * We don't free the group descriptors allocated by compute_sb_data() 671 * until ext2_unmount(). This is OK since the mount will succeed. 672 */ 673 brelse(bp); 674 bp = NULL; 675 fs = ump->um_e2fs; 676 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */ 677 /* if the fs is not mounted read-only, make sure the super block is 678 always written back on a sync() 679 */ 680 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0; 681 if (ronly == 0) { 682 fs->s_dirt = 1; /* mark it modified */ 683 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */ 684 } 685 mp->mnt_data = (qaddr_t)ump; 686 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev); 687 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; 688 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; 689 mp->mnt_flag |= MNT_LOCAL; 690 ump->um_mountp = mp; 691 ump->um_dev = dev; 692 ump->um_devvp = devvp; 693 /* setting those two parameters allows us to use 694 ufs_bmap w/o changse ! 695 */ 696 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); 697 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1; 698 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); 699 for (i = 0; i < MAXQUOTAS; i++) 700 ump->um_quotas[i] = NULLVP; 701 devvp->v_specmountpoint = mp; 702 if (ronly == 0) 703 ext2_sbupdate(ump, MNT_WAIT); 704 return (0); 705out: 706 if (bp) 707 brelse(bp); 708 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p); 709 if (ump) { 710 bsd_free(ump->um_e2fs->s_es, M_UFSMNT); 711 bsd_free(ump->um_e2fs, M_UFSMNT); 712 bsd_free(ump, M_UFSMNT); 713 mp->mnt_data = (qaddr_t)0; 714 } 715 return (error); 716} 717 718/* 719 * unmount system call 720 */ 721static int 722ext2_unmount(mp, mntflags, p) 723 struct mount *mp; 724 int mntflags; 725 struct proc *p; 726{ 727 register struct ufsmount *ump; 728 register struct ext2_sb_info *fs; 729 int error, flags, ronly, i; 730 731 flags = 0; 732 if (mntflags & MNT_FORCE) { 733 if (mp->mnt_flag & MNT_ROOTFS) 734 return (EINVAL); 735 flags |= FORCECLOSE; 736 } 737 if ((error = ext2_flushfiles(mp, flags, p)) != 0) 738 return (error); 739 ump = VFSTOUFS(mp); 740 fs = ump->um_e2fs; 741 ronly = fs->s_rd_only; 742 if (ronly == 0) { 743 if (fs->s_wasvalid) 744 fs->s_es->s_state |= EXT2_VALID_FS; 745 ext2_sbupdate(ump, MNT_WAIT); 746 } 747 748 /* release buffers containing group descriptors */ 749 for(i = 0; i < fs->s_db_per_group; i++) 750 ULCK_BUF(fs->s_group_desc[i]) 751 bsd_free(fs->s_group_desc, M_UFSMNT); 752 753 /* release cached inode/block bitmaps */ 754 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 755 if (fs->s_inode_bitmap[i]) 756 ULCK_BUF(fs->s_inode_bitmap[i]) 757 758 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 759 if (fs->s_block_bitmap[i]) 760 ULCK_BUF(fs->s_block_bitmap[i]) 761 762 ump->um_devvp->v_specmountpoint = NULL; 763 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, 764 NOCRED, p); 765 vrele(ump->um_devvp); 766 bsd_free(fs->s_es, M_UFSMNT); 767 bsd_free(fs, M_UFSMNT); 768 bsd_free(ump, M_UFSMNT); 769 mp->mnt_data = (qaddr_t)0; 770 mp->mnt_flag &= ~MNT_LOCAL; 771 return (error); 772} 773 774/* 775 * Flush out all the files in a filesystem. 776 */ 777static int 778ext2_flushfiles(mp, flags, p) 779 register struct mount *mp; 780 int flags; 781 struct proc *p; 782{ 783 register struct ufsmount *ump; 784 int error; 785#if QUOTA 786 int i; 787#endif 788 789 ump = VFSTOUFS(mp); 790#if QUOTA 791 if (mp->mnt_flag & MNT_QUOTA) { 792 if ((error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) != 0) 793 return (error); 794 for (i = 0; i < MAXQUOTAS; i++) { 795 if (ump->um_quotas[i] == NULLVP) 796 continue; 797 quotaoff(p, mp, i); 798 } 799 /* 800 * Here we fall through to vflush again to ensure 801 * that we have gotten rid of all the system vnodes. 802 */ 803 } 804#endif 805 error = vflush(mp, NULLVP, flags); 806 return (error); 807} 808 809/* 810 * Get file system statistics. 811 * taken from ext2/super.c ext2_statfs 812 */ 813static int 814ext2_statfs(mp, sbp, p) 815 struct mount *mp; 816 register struct statfs *sbp; 817 struct proc *p; 818{ 819 unsigned long overhead; 820 unsigned long overhead_per_group; 821 822 register struct ufsmount *ump; 823 register struct ext2_sb_info *fs; 824 register struct ext2_super_block *es; 825 826 ump = VFSTOUFS(mp); 827 fs = ump->um_e2fs; 828 es = fs->s_es; 829 830 if (es->s_magic != EXT2_SUPER_MAGIC) 831 panic("ext2_statfs - magic number spoiled"); 832 833 /* 834 * Compute the overhead (FS structures) 835 */ 836 overhead_per_group = 1 /* super block */ + 837 fs->s_db_per_group + 838 1 /* block bitmap */ + 839 1 /* inode bitmap */ + 840 fs->s_itb_per_group; 841 overhead = es->s_first_data_block + 842 fs->s_groups_count * overhead_per_group; 843 844 sbp->f_bsize = EXT2_FRAG_SIZE(fs); 845 sbp->f_iosize = EXT2_BLOCK_SIZE(fs); 846 sbp->f_blocks = es->s_blocks_count - overhead; 847 sbp->f_bfree = es->s_free_blocks_count; 848 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count; 849 sbp->f_files = es->s_inodes_count; 850 sbp->f_ffree = es->s_free_inodes_count; 851 if (sbp != &mp->mnt_stat) { 852 sbp->f_type = mp->mnt_vfc->vfc_typenum; 853 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 854 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 855 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 856 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 857 } 858 return (0); 859} 860 861/* 862 * Go through the disk queues to initiate sandbagged IO; 863 * go through the inodes to write those that have been modified; 864 * initiate the writing of the super block if it has been modified. 865 * 866 * Note: we are always called with the filesystem marked `MPBUSY'. 867 */ 868static int 869ext2_sync(mp, waitfor, cred, p) 870 struct mount *mp; 871 int waitfor; 872 struct ucred *cred; 873 struct proc *p; 874{ 875 struct vnode *nvp, *vp; 876 struct inode *ip; 877 struct ufsmount *ump = VFSTOUFS(mp); 878 struct ext2_sb_info *fs; 879 int error, allerror = 0; 880 881 fs = ump->um_e2fs; 882 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */ 883 printf("fs = %s\n", fs->fs_fsmnt); 884 panic("ext2_sync: rofs mod"); 885 } 886 /* 887 * Write back each (modified) inode. 888 */ 889 simple_lock(&mntvnode_slock); 890loop: 891 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 892 /* 893 * If the vnode that we are about to sync is no longer 894 * associated with this mount point, start over. 895 */ 896 if (vp->v_mount != mp) 897 goto loop; 898 simple_lock(&vp->v_interlock); 899 nvp = vp->v_mntvnodes.le_next; 900 ip = VTOI(vp); 901 if (vp->v_type == VNON || 902 ((ip->i_flag & 903 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 904 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || waitfor == MNT_LAZY))) { 905 simple_unlock(&vp->v_interlock); 906 continue; 907 } 908 simple_unlock(&mntvnode_slock); 909 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, p); 910 if (error) { 911 simple_lock(&mntvnode_slock); 912 if (error == ENOENT) 913 goto loop; 914 continue; 915 } 916 if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0) 917 allerror = error; 918 VOP_UNLOCK(vp, 0, p); 919 vrele(vp); 920 simple_lock(&mntvnode_slock); 921 } 922 simple_unlock(&mntvnode_slock); 923 /* 924 * Force stale file system control information to be flushed. 925 */ 926 if (waitfor != MNT_LAZY) { 927 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p); 928 if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0) 929 allerror = error; 930 VOP_UNLOCK(ump->um_devvp, 0, p); 931 } 932#if QUOTA 933 qsync(mp); 934#endif 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, vpp) 955 struct mount *mp; 956 ino_t ino; 957 struct vnode **vpp; 958{ 959 register struct ext2_sb_info *fs; 960 register struct inode *ip; 961 struct ufsmount *ump; 962 struct buf *bp; 963 struct vnode *vp; 964 dev_t dev; 965 int i, error; 966 int used_blocks; 967 968 ump = VFSTOUFS(mp); 969 dev = ump->um_dev; 970restart: 971 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) 972 return (0); 973 974 /* 975 * Lock out the creation of new entries in the FFS hash table in 976 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 977 * may occur! 978 */ 979 if (ext2fs_inode_hash_lock) { 980 while (ext2fs_inode_hash_lock) { 981 ext2fs_inode_hash_lock = -1; 982 tsleep(&ext2fs_inode_hash_lock, PVM, "e2vget", 0); 983 } 984 goto restart; 985 } 986 ext2fs_inode_hash_lock = 1; 987 988 /* 989 * If this MALLOC() is performed after the getnewvnode() 990 * it might block, leaving a vnode with a NULL v_data to be 991 * found by ext2_sync() if a sync happens to fire right then, 992 * which will cause a panic because ext2_sync() blindly 993 * dereferences vp->v_data (as well it should). 994 */ 995 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK); 996 997 /* Allocate a new vnode/inode. */ 998 if ((error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) != 0) { 999 if (ext2fs_inode_hash_lock < 0) 1000 wakeup(&ext2fs_inode_hash_lock); 1001 ext2fs_inode_hash_lock = 0; 1002 *vpp = NULL; 1003 FREE(ip, M_EXT2NODE); 1004 return (error); 1005 } 1006 bzero((caddr_t)ip, sizeof(struct inode)); 1007 lockinit(&ip->i_lock, PINOD, "ext2in", 0, 0); 1008 vp->v_data = ip; 1009 ip->i_vnode = vp; 1010 ip->i_e2fs = fs = ump->um_e2fs; 1011 ip->i_dev = dev; 1012 ip->i_number = ino; 1013#if QUOTA 1014 for (i = 0; i < MAXQUOTAS; i++) 1015 ip->i_dquot[i] = NODQUOT; 1016#endif 1017 /* 1018 * Put it onto its hash chain and lock it so that other requests for 1019 * this inode will block if they arrive while we are sleeping waiting 1020 * for old data structures to be purged or for the contents of the 1021 * disk portion of this inode to be read. 1022 */ 1023 ufs_ihashins(ip); 1024 1025 if (ext2fs_inode_hash_lock < 0) 1026 wakeup(&ext2fs_inode_hash_lock); 1027 ext2fs_inode_hash_lock = 0; 1028 1029 /* Read in the disk contents for the inode, copy into the inode. */ 1030#if 0 1031printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino))); 1032#endif 1033 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 1034 (int)fs->s_blocksize, NOCRED, &bp)) != 0) { 1035 /* 1036 * The inode does not contain anything useful, so it would 1037 * be misleading to leave it on its hash chain. With mode 1038 * still zero, it will be unlinked and returned to the free 1039 * list by vput(). 1040 */ 1041 vput(vp); 1042 brelse(bp); 1043 *vpp = NULL; 1044 return (error); 1045 } 1046 /* convert ext2 inode to dinode */ 1047 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE * 1048 ino_to_fsbo(fs, ino)), &ip->i_din); 1049 ip->i_block_group = ino_to_cg(fs, ino); 1050 ip->i_next_alloc_block = 0; 1051 ip->i_next_alloc_goal = 0; 1052 ip->i_prealloc_count = 0; 1053 ip->i_prealloc_block = 0; 1054 /* now we want to make sure that block pointers for unused 1055 blocks are zeroed out - ext2_balloc depends on this 1056 although for regular files and directories only 1057 */ 1058 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) { 1059 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize; 1060 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) 1061 ip->i_db[i] = 0; 1062 } 1063/* 1064 ext2_print_inode(ip); 1065*/ 1066 brelse(bp); 1067 1068 /* 1069 * Initialize the vnode from the inode, check for aliases. 1070 * Note that the underlying vnode may have changed. 1071 */ 1072 if ((error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) != 0) { 1073 vput(vp); 1074 *vpp = NULL; 1075 return (error); 1076 } 1077 /* 1078 * Finish inode initialization now that aliasing has been resolved. 1079 */ 1080 ip->i_devvp = ump->um_devvp; 1081 VREF(ip->i_devvp); 1082 /* 1083 * Set up a generation number for this inode if it does not 1084 * already have one. This should only happen on old filesystems. 1085 */ 1086 if (ip->i_gen == 0) { 1087 ip->i_gen = random() / 2 + 1; 1088 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1089 ip->i_flag |= IN_MODIFIED; 1090 } 1091 *vpp = vp; 1092 return (0); 1093} 1094 1095/* 1096 * File handle to vnode 1097 * 1098 * Have to be really careful about stale file handles: 1099 * - check that the inode number is valid 1100 * - call ext2_vget() to get the locked inode 1101 * - check for an unallocated inode (i_mode == 0) 1102 * - check that the given client host has export rights and return 1103 * those rights via. exflagsp and credanonp 1104 */ 1105static int 1106ext2_fhtovp(mp, fhp, vpp) 1107 register struct mount *mp; 1108 struct fid *fhp; 1109 struct vnode **vpp; 1110{ 1111 register struct ufid *ufhp; 1112 struct ext2_sb_info *fs; 1113 1114 ufhp = (struct ufid *)fhp; 1115 fs = VFSTOUFS(mp)->um_e2fs; 1116 if (ufhp->ufid_ino < ROOTINO || 1117 ufhp->ufid_ino >= fs->s_groups_count * fs->s_es->s_inodes_per_group) 1118 return (ESTALE); 1119 return (ufs_fhtovp(mp, ufhp, vpp)); 1120} 1121 1122/* 1123 * Vnode pointer to File handle 1124 */ 1125/* ARGSUSED */ 1126static int 1127ext2_vptofh(vp, fhp) 1128 struct vnode *vp; 1129 struct fid *fhp; 1130{ 1131 register struct inode *ip; 1132 register struct ufid *ufhp; 1133 1134 ip = VTOI(vp); 1135 ufhp = (struct ufid *)fhp; 1136 ufhp->ufid_len = sizeof(struct ufid); 1137 ufhp->ufid_ino = ip->i_number; 1138 ufhp->ufid_gen = ip->i_gen; 1139 return (0); 1140} 1141 1142/* 1143 * Write a superblock and associated information back to disk. 1144 */ 1145static int 1146ext2_sbupdate(mp, waitfor) 1147 struct ufsmount *mp; 1148 int waitfor; 1149{ 1150 register struct ext2_sb_info *fs = mp->um_e2fs; 1151 register struct ext2_super_block *es = fs->s_es; 1152 register struct buf *bp; 1153 int error = 0; 1154/* 1155printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no"); 1156*/ 1157 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0); 1158 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block)); 1159 if (waitfor == MNT_WAIT) 1160 error = bwrite(bp); 1161 else 1162 bawrite(bp); 1163 1164 /* 1165 * The buffers for group descriptors, inode bitmaps and block bitmaps 1166 * are not busy at this point and are (hopefully) written by the 1167 * usual sync mechanism. No need to write them here 1168 */ 1169 1170 return (error); 1171} 1172