ext2_vfsops.c revision 59794
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 59794 2000-04-30 18:52:11Z 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/extattr.h> 60#include <ufs/ufs/quota.h> 61#include <ufs/ufs/ufsmount.h> 62#include <ufs/ufs/inode.h> 63#include <ufs/ufs/ufs_extern.h> 64 65 66#include <gnu/ext2fs/fs.h> 67#include <gnu/ext2fs/ext2_extern.h> 68#include <gnu/ext2fs/ext2_fs.h> 69#include <gnu/ext2fs/ext2_fs_sb.h> 70 71static int ext2_fhtovp __P((struct mount *, struct fid *, struct vnode **)); 72static int ext2_flushfiles __P((struct mount *mp, int flags, struct proc *p)); 73static int ext2_mount __P((struct mount *, 74 char *, caddr_t, struct nameidata *, struct proc *)); 75static int ext2_mountfs __P((struct vnode *, struct mount *, struct proc *)); 76static int ext2_reload __P((struct mount *mountp, struct ucred *cred, 77 struct proc *p)); 78static int ext2_sbupdate __P((struct ufsmount *, int)); 79static int ext2_statfs __P((struct mount *, struct statfs *, struct proc *)); 80static int ext2_sync __P((struct mount *, int, struct ucred *, struct proc *)); 81static int ext2_unmount __P((struct mount *, int, struct proc *)); 82static int ext2_vget __P((struct mount *, ino_t, struct vnode **)); 83static int ext2_vptofh __P((struct vnode *, struct fid *)); 84 85static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part"); 86 87static struct vfsops ext2fs_vfsops = { 88 ext2_mount, 89 ufs_start, /* empty function */ 90 ext2_unmount, 91 ufs_root, /* root inode via vget */ 92 ufs_quotactl, /* does operations associated with quotas */ 93 ext2_statfs, 94 ext2_sync, 95 ext2_vget, 96 ext2_fhtovp, 97 ufs_check_export, 98 ext2_vptofh, 99 ext2_init, 100 vfs_stduninit, 101 vfs_stdextattrctl, 102}; 103 104VFS_SET(ext2fs_vfsops, ext2fs, 0); 105#define bsd_malloc malloc 106#define bsd_free free 107 108static int ext2fs_inode_hash_lock; 109 110static int ext2_check_sb_compat __P((struct ext2_super_block *es, 111 dev_t dev, int ronly)); 112static int compute_sb_data __P((struct vnode * devvp, 113 struct ext2_super_block * es, 114 struct ext2_sb_info * fs)); 115 116#ifdef notyet 117static int ext2_mountroot __P((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 proc *p = curproc; 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 mp->mnt_op = &ext2fs_vfsops; 143 mp->mnt_flag = MNT_RDONLY; 144 if (error = ext2_mountfs(rootvp, mp, p)) { 145 bsd_free(mp, M_MOUNT); 146 return (error); 147 } 148 if (error = vfs_lock(mp)) { 149 (void)ext2_unmount(mp, 0, p); 150 bsd_free(mp, M_MOUNT); 151 return (error); 152 } 153 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list); 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#endif 171 172/* 173 * VFS Operations. 174 * 175 * mount system call 176 */ 177static int 178ext2_mount(mp, path, data, ndp, p) 179 register struct mount *mp; 180 char *path; 181 caddr_t data; /* this is actually a (struct ufs_args *) */ 182 struct nameidata *ndp; 183 struct proc *p; 184{ 185 struct vnode *devvp; 186 struct ufs_args args; 187 struct ufsmount *ump = 0; 188 register struct ext2_sb_info *fs; 189 u_int size; 190 int error, flags; 191 mode_t accessmode; 192 193 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) != 0) 194 return (error); 195 /* 196 * If updating, check whether changing from read-only to 197 * read/write; if there is no device name, that's all we do. 198 */ 199 if (mp->mnt_flag & MNT_UPDATE) { 200 ump = VFSTOUFS(mp); 201 fs = ump->um_e2fs; 202 error = 0; 203 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) { 204 flags = WRITECLOSE; 205 if (mp->mnt_flag & MNT_FORCE) 206 flags |= FORCECLOSE; 207 if (vfs_busy(mp, LK_NOWAIT, 0, p)) 208 return (EBUSY); 209 error = ext2_flushfiles(mp, flags, p); 210 vfs_unbusy(mp, p); 211 if (!error && fs->s_wasvalid) { 212 fs->s_es->s_state |= EXT2_VALID_FS; 213 ext2_sbupdate(ump, MNT_WAIT); 214 } 215 fs->s_rd_only = 1; 216 } 217 if (!error && (mp->mnt_flag & MNT_RELOAD)) 218 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, p); 219 if (error) 220 return (error); 221 devvp = ump->um_devvp; 222 if (ext2_check_sb_compat(fs->s_es, devvp->v_rdev, 223 (mp->mnt_kern_flag & MNTK_WANTRDWR) == 0) != 0) 224 return (EPERM); 225 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { 226 /* 227 * If upgrade to read-write by non-root, then verify 228 * that user has necessary permissions on the device. 229 */ 230 if (p->p_ucred->cr_uid != 0) { 231 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 232 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE, 233 p->p_ucred, p)) != 0) { 234 VOP_UNLOCK(devvp, 0, p); 235 return (error); 236 } 237 VOP_UNLOCK(devvp, 0, p); 238 } 239 240 if ((fs->s_es->s_state & EXT2_VALID_FS) == 0 || 241 (fs->s_es->s_state & EXT2_ERROR_FS)) { 242 if (mp->mnt_flag & MNT_FORCE) { 243 printf( 244"WARNING: %s was not properly dismounted\n", 245 fs->fs_fsmnt); 246 } else { 247 printf( 248"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", 249 fs->fs_fsmnt); 250 return (EPERM); 251 } 252 } 253 fs->s_es->s_state &= ~EXT2_VALID_FS; 254 ext2_sbupdate(ump, MNT_WAIT); 255 fs->s_rd_only = 0; 256 } 257 if (args.fspec == 0) { 258 /* 259 * Process export requests. 260 */ 261 return (vfs_export(mp, &ump->um_export, &args.export)); 262 } 263 } 264 /* 265 * Not an update, or updating the name: look up the name 266 * and verify that it refers to a sensible block device. 267 */ 268 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p); 269 if ((error = namei(ndp)) != 0) 270 return (error); 271 NDFREE(ndp, NDF_ONLY_PNBUF); 272 devvp = ndp->ni_vp; 273 274 if (!vn_isdisk(devvp, &error)) { 275 vrele(devvp); 276 return (error); 277 } 278 279 /* 280 * If mount by non-root, then verify that user has necessary 281 * permissions on the device. 282 */ 283 if (p->p_ucred->cr_uid != 0) { 284 accessmode = VREAD; 285 if ((mp->mnt_flag & MNT_RDONLY) == 0) 286 accessmode |= VWRITE; 287 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 288 if ((error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p)) != 0) { 289 vput(devvp); 290 return (error); 291 } 292 VOP_UNLOCK(devvp, 0, p); 293 } 294 295 if ((mp->mnt_flag & MNT_UPDATE) == 0) { 296 error = ext2_mountfs(devvp, mp, p); 297 } else { 298 if (devvp != ump->um_devvp) 299 error = EINVAL; /* needs translation */ 300 else 301 vrele(devvp); 302 } 303 if (error) { 304 vrele(devvp); 305 return (error); 306 } 307 ump = VFSTOUFS(mp); 308 fs = ump->um_e2fs; 309 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size); 310 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size); 311 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 312 MNAMELEN); 313 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 314 &size); 315 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 316 (void)ext2_statfs(mp, &mp->mnt_stat, p); 317 return (0); 318} 319 320/* 321 * checks that the data in the descriptor blocks make sense 322 * this is taken from ext2/super.c 323 */ 324static int ext2_check_descriptors (struct ext2_sb_info * sb) 325{ 326 int i; 327 int desc_block = 0; 328 unsigned long block = sb->s_es->s_first_data_block; 329 struct ext2_group_desc * gdp = NULL; 330 331 /* ext2_debug ("Checking group descriptors"); */ 332 333 for (i = 0; i < sb->s_groups_count; i++) 334 { 335 /* examine next descriptor block */ 336 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0) 337 gdp = (struct ext2_group_desc *) 338 sb->s_group_desc[desc_block++]->b_data; 339 if (gdp->bg_block_bitmap < block || 340 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 341 { 342 printf ("ext2_check_descriptors: " 343 "Block bitmap for group %d" 344 " not in group (block %lu)!\n", 345 i, (unsigned long) gdp->bg_block_bitmap); 346 return 0; 347 } 348 if (gdp->bg_inode_bitmap < block || 349 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 350 { 351 printf ("ext2_check_descriptors: " 352 "Inode bitmap for group %d" 353 " not in group (block %lu)!\n", 354 i, (unsigned long) gdp->bg_inode_bitmap); 355 return 0; 356 } 357 if (gdp->bg_inode_table < block || 358 gdp->bg_inode_table + sb->s_itb_per_group >= 359 block + EXT2_BLOCKS_PER_GROUP(sb)) 360 { 361 printf ("ext2_check_descriptors: " 362 "Inode table for group %d" 363 " not in group (block %lu)!\n", 364 i, (unsigned long) gdp->bg_inode_table); 365 return 0; 366 } 367 block += EXT2_BLOCKS_PER_GROUP(sb); 368 gdp++; 369 } 370 return 1; 371} 372 373static int 374ext2_check_sb_compat(es, dev, ronly) 375 struct ext2_super_block *es; 376 dev_t dev; 377 int ronly; 378{ 379 380 if (es->s_magic != EXT2_SUPER_MAGIC) { 381 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n", 382 devtoname(dev), es->s_magic, EXT2_SUPER_MAGIC); 383 return (1); 384 } 385 if (es->s_rev_level > EXT2_GOOD_OLD_REV) { 386 if (es->s_feature_incompat & ~EXT2_FEATURE_INCOMPAT_SUPP) { 387 printf( 388"WARNING: mount of %s denied due to unsupported optional features\n", 389 devtoname(dev)); 390 return (1); 391 } 392 if (!ronly && 393 (es->s_feature_ro_compat & ~EXT2_FEATURE_RO_COMPAT_SUPP)) { 394 printf( 395"WARNING: R/W mount of %s denied due to unsupported optional features\n", 396 devtoname(dev)); 397 return (1); 398 } 399 } 400 return (0); 401} 402 403/* 404 * this computes the fields of the ext2_sb_info structure from the 405 * data in the ext2_super_block structure read in 406 */ 407static int compute_sb_data(devvp, es, fs) 408 struct vnode * devvp; 409 struct ext2_super_block * es; 410 struct ext2_sb_info * fs; 411{ 412 int db_count, error; 413 int i, j; 414 int logic_sb_block = 1; /* XXX for now */ 415 416#if 1 417#define V(v) 418#else 419#define V(v) printf(#v"= %d\n", fs->v); 420#endif 421 422 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size; 423 V(s_blocksize) 424 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size; 425 V(s_bshift) 426 fs->s_fsbtodb = es->s_log_block_size + 1; 427 V(s_fsbtodb) 428 fs->s_qbmask = fs->s_blocksize - 1; 429 V(s_bmask) 430 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es); 431 V(s_blocksize_bits) 432 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size; 433 V(s_frag_size) 434 if (fs->s_frag_size) 435 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size; 436 V(s_frags_per_block) 437 fs->s_blocks_per_group = es->s_blocks_per_group; 438 V(s_blocks_per_group) 439 fs->s_frags_per_group = es->s_frags_per_group; 440 V(s_frags_per_group) 441 fs->s_inodes_per_group = es->s_inodes_per_group; 442 V(s_inodes_per_group) 443 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE; 444 V(s_inodes_per_block) 445 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block; 446 V(s_itb_per_group) 447 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc); 448 V(s_desc_per_block) 449 /* s_resuid / s_resgid ? */ 450 fs->s_groups_count = (es->s_blocks_count - 451 es->s_first_data_block + 452 EXT2_BLOCKS_PER_GROUP(fs) - 1) / 453 EXT2_BLOCKS_PER_GROUP(fs); 454 V(s_groups_count) 455 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) / 456 EXT2_DESC_PER_BLOCK(fs); 457 fs->s_db_per_group = db_count; 458 V(s_db_per_group) 459 460 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *), 461 M_UFSMNT, M_WAITOK); 462 463 /* adjust logic_sb_block */ 464 if(fs->s_blocksize > SBSIZE) 465 /* Godmar thinks: if the blocksize is greater than 1024, then 466 the superblock is logically part of block zero. 467 */ 468 logic_sb_block = 0; 469 470 for (i = 0; i < db_count; i++) { 471 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1), 472 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]); 473 if(error) { 474 for (j = 0; j < i; j++) 475 brelse(fs->s_group_desc[j]); 476 bsd_free(fs->s_group_desc, M_UFSMNT); 477 printf("EXT2-fs: unable to read group descriptors (%d)\n", error); 478 return EIO; 479 } 480 /* Set the B_LOCKED flag on the buffer, then brelse() it */ 481 LCK_BUF(fs->s_group_desc[i]) 482 } 483 if(!ext2_check_descriptors(fs)) { 484 for (j = 0; j < db_count; j++) 485 ULCK_BUF(fs->s_group_desc[j]) 486 bsd_free(fs->s_group_desc, M_UFSMNT); 487 printf("EXT2-fs: (ext2_check_descriptors failure) " 488 "unable to read group descriptors\n"); 489 return EIO; 490 } 491 492 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) { 493 fs->s_inode_bitmap_number[i] = 0; 494 fs->s_inode_bitmap[i] = NULL; 495 fs->s_block_bitmap_number[i] = 0; 496 fs->s_block_bitmap[i] = NULL; 497 } 498 fs->s_loaded_inode_bitmaps = 0; 499 fs->s_loaded_block_bitmaps = 0; 500 return 0; 501} 502 503/* 504 * Reload all incore data for a filesystem (used after running fsck on 505 * the root filesystem and finding things to fix). The filesystem must 506 * be mounted read-only. 507 * 508 * Things to do to update the mount: 509 * 1) invalidate all cached meta-data. 510 * 2) re-read superblock from disk. 511 * 3) re-read summary information from disk. 512 * 4) invalidate all inactive vnodes. 513 * 5) invalidate all cached file data. 514 * 6) re-read inode data for all active vnodes. 515 */ 516static int 517ext2_reload(mountp, cred, p) 518 register struct mount *mountp; 519 struct ucred *cred; 520 struct proc *p; 521{ 522 register struct vnode *vp, *nvp, *devvp; 523 struct inode *ip; 524 struct buf *bp; 525 struct ext2_super_block * es; 526 struct ext2_sb_info *fs; 527 int error; 528 529 if ((mountp->mnt_flag & MNT_RDONLY) == 0) 530 return (EINVAL); 531 /* 532 * Step 1: invalidate all cached meta-data. 533 */ 534 devvp = VFSTOUFS(mountp)->um_devvp; 535 if (vinvalbuf(devvp, 0, cred, p, 0, 0)) 536 panic("ext2_reload: dirty1"); 537 /* 538 * Step 2: re-read superblock from disk. 539 * constants have been adjusted for ext2 540 */ 541 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 542 return (error); 543 es = (struct ext2_super_block *)bp->b_data; 544 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) { 545 brelse(bp); 546 return (EIO); /* XXX needs translation */ 547 } 548 fs = VFSTOUFS(mountp)->um_e2fs; 549 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block)); 550 551 if((error = compute_sb_data(devvp, es, fs)) != 0) { 552 brelse(bp); 553 return error; 554 } 555#ifdef UNKLAR 556 if (fs->fs_sbsize < SBSIZE) 557 bp->b_flags |= B_INVAL; 558#endif 559 brelse(bp); 560 561loop: 562 simple_lock(&mntvnode_slock); 563 for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 564 if (vp->v_mount != mountp) { 565 simple_unlock(&mntvnode_slock); 566 goto loop; 567 } 568 nvp = vp->v_mntvnodes.le_next; 569 /* 570 * Step 4: invalidate all inactive vnodes. 571 */ 572 if (vrecycle(vp, &mntvnode_slock, p)) 573 goto loop; 574 /* 575 * Step 5: invalidate all cached file data. 576 */ 577 simple_lock(&vp->v_interlock); 578 simple_unlock(&mntvnode_slock); 579 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) { 580 goto loop; 581 } 582 if (vinvalbuf(vp, 0, cred, p, 0, 0)) 583 panic("ext2_reload: dirty2"); 584 /* 585 * Step 6: re-read inode data for all active vnodes. 586 */ 587 ip = VTOI(vp); 588 error = 589 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 590 (int)fs->s_blocksize, NOCRED, &bp); 591 if (error) { 592 vput(vp); 593 return (error); 594 } 595 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + 596 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)), 597 &ip->i_din); 598 brelse(bp); 599 vput(vp); 600 simple_lock(&mntvnode_slock); 601 } 602 simple_unlock(&mntvnode_slock); 603 return (0); 604} 605 606/* 607 * Common code for mount and mountroot 608 */ 609static int 610ext2_mountfs(devvp, mp, p) 611 register struct vnode *devvp; 612 struct mount *mp; 613 struct proc *p; 614{ 615 register struct ufsmount *ump; 616 struct buf *bp; 617 register struct ext2_sb_info *fs; 618 struct ext2_super_block * es; 619 dev_t dev = devvp->v_rdev; 620 struct partinfo dpart; 621 int havepart = 0; 622 int error, i, size; 623 int ronly; 624 625 /* 626 * Disallow multiple mounts of the same device. 627 * Disallow mounting of a device that is currently in use 628 * (except for root, which might share swap device for miniroot). 629 * Flush out any old buffers remaining from a previous use. 630 */ 631 if ((error = vfs_mountedon(devvp)) != 0) 632 return (error); 633 if (vcount(devvp) > 1 && devvp != rootvp) 634 return (EBUSY); 635 if ((error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) 636 return (error); 637#ifdef READONLY 638/* turn on this to force it to be read-only */ 639 mp->mnt_flag |= MNT_RDONLY; 640#endif 641 642 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 643 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p); 644 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p); 645 VOP_UNLOCK(devvp, 0, p); 646 if (error) 647 return (error); 648 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0) 649 size = DEV_BSIZE; 650 else { 651 havepart = 1; 652 size = dpart.disklab->d_secsize; 653 } 654 655 bp = NULL; 656 ump = NULL; 657 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0) 658 goto out; 659 es = (struct ext2_super_block *)bp->b_data; 660 if (ext2_check_sb_compat(es, dev, ronly) != 0) { 661 error = EINVAL; /* XXX needs translation */ 662 goto out; 663 } 664 if ((es->s_state & EXT2_VALID_FS) == 0 || 665 (es->s_state & EXT2_ERROR_FS)) { 666 if (ronly || (mp->mnt_flag & MNT_FORCE)) { 667 printf( 668"WARNING: Filesystem was not properly dismounted\n"); 669 } else { 670 printf( 671"WARNING: R/W mount denied. Filesystem is not clean - run fsck\n"); 672 error = EPERM; 673 goto out; 674 } 675 } 676 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK); 677 bzero((caddr_t)ump, sizeof *ump); 678 ump->um_malloctype = M_EXT2NODE; 679 ump->um_blkatoff = ext2_blkatoff; 680 ump->um_truncate = ext2_truncate; 681 ump->um_update = ext2_update; 682 ump->um_valloc = ext2_valloc; 683 ump->um_vfree = ext2_vfree; 684 /* I don't know whether this is the right strategy. Note that 685 we dynamically allocate both a ext2_sb_info and a ext2_super_block 686 while Linux keeps the super block in a locked buffer 687 */ 688 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info), 689 M_UFSMNT, M_WAITOK); 690 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block), 691 M_UFSMNT, M_WAITOK); 692 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block)); 693 if ((error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs))) 694 goto out; 695 /* 696 * We don't free the group descriptors allocated by compute_sb_data() 697 * until ext2_unmount(). This is OK since the mount will succeed. 698 */ 699 brelse(bp); 700 bp = NULL; 701 fs = ump->um_e2fs; 702 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */ 703 /* if the fs is not mounted read-only, make sure the super block is 704 always written back on a sync() 705 */ 706 fs->s_wasvalid = fs->s_es->s_state & EXT2_VALID_FS ? 1 : 0; 707 if (ronly == 0) { 708 fs->s_dirt = 1; /* mark it modified */ 709 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */ 710 } 711 mp->mnt_data = (qaddr_t)ump; 712 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev); 713 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum; 714 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; 715 mp->mnt_flag |= MNT_LOCAL; 716 ump->um_mountp = mp; 717 ump->um_dev = dev; 718 ump->um_devvp = devvp; 719 /* setting those two parameters allows us to use 720 ufs_bmap w/o changse ! 721 */ 722 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); 723 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1; 724 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); 725 for (i = 0; i < MAXQUOTAS; i++) 726 ump->um_quotas[i] = NULLVP; 727 devvp->v_specmountpoint = mp; 728 if (ronly == 0) 729 ext2_sbupdate(ump, MNT_WAIT); 730 return (0); 731out: 732 if (bp) 733 brelse(bp); 734 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p); 735 if (ump) { 736 bsd_free(ump->um_e2fs->s_es, M_UFSMNT); 737 bsd_free(ump->um_e2fs, M_UFSMNT); 738 bsd_free(ump, M_UFSMNT); 739 mp->mnt_data = (qaddr_t)0; 740 } 741 return (error); 742} 743 744/* 745 * unmount system call 746 */ 747static int 748ext2_unmount(mp, mntflags, p) 749 struct mount *mp; 750 int mntflags; 751 struct proc *p; 752{ 753 register struct ufsmount *ump; 754 register struct ext2_sb_info *fs; 755 int error, flags, ronly, i; 756 757 flags = 0; 758 if (mntflags & MNT_FORCE) { 759 if (mp->mnt_flag & MNT_ROOTFS) 760 return (EINVAL); 761 flags |= FORCECLOSE; 762 } 763 if ((error = ext2_flushfiles(mp, flags, p)) != 0) 764 return (error); 765 ump = VFSTOUFS(mp); 766 fs = ump->um_e2fs; 767 ronly = fs->s_rd_only; 768 if (ronly == 0) { 769 if (fs->s_wasvalid) 770 fs->s_es->s_state |= EXT2_VALID_FS; 771 ext2_sbupdate(ump, MNT_WAIT); 772 } 773 774 /* release buffers containing group descriptors */ 775 for(i = 0; i < fs->s_db_per_group; i++) 776 ULCK_BUF(fs->s_group_desc[i]) 777 bsd_free(fs->s_group_desc, M_UFSMNT); 778 779 /* release cached inode/block bitmaps */ 780 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 781 if (fs->s_inode_bitmap[i]) 782 ULCK_BUF(fs->s_inode_bitmap[i]) 783 784 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 785 if (fs->s_block_bitmap[i]) 786 ULCK_BUF(fs->s_block_bitmap[i]) 787 788 ump->um_devvp->v_specmountpoint = NULL; 789 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, 790 NOCRED, p); 791 vrele(ump->um_devvp); 792 bsd_free(fs->s_es, M_UFSMNT); 793 bsd_free(fs, M_UFSMNT); 794 bsd_free(ump, M_UFSMNT); 795 mp->mnt_data = (qaddr_t)0; 796 mp->mnt_flag &= ~MNT_LOCAL; 797 return (error); 798} 799 800/* 801 * Flush out all the files in a filesystem. 802 */ 803static int 804ext2_flushfiles(mp, flags, p) 805 register struct mount *mp; 806 int flags; 807 struct proc *p; 808{ 809 register struct ufsmount *ump; 810 int error; 811#if QUOTA 812 int i; 813#endif 814 815 ump = VFSTOUFS(mp); 816#if QUOTA 817 if (mp->mnt_flag & MNT_QUOTA) { 818 if ((error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) != 0) 819 return (error); 820 for (i = 0; i < MAXQUOTAS; i++) { 821 if (ump->um_quotas[i] == NULLVP) 822 continue; 823 quotaoff(p, mp, i); 824 } 825 /* 826 * Here we fall through to vflush again to ensure 827 * that we have gotten rid of all the system vnodes. 828 */ 829 } 830#endif 831 error = vflush(mp, NULLVP, flags); 832 return (error); 833} 834 835/* 836 * Get file system statistics. 837 * taken from ext2/super.c ext2_statfs 838 */ 839static int 840ext2_statfs(mp, sbp, p) 841 struct mount *mp; 842 register struct statfs *sbp; 843 struct proc *p; 844{ 845 unsigned long overhead; 846 unsigned long overhead_per_group; 847 848 register struct ufsmount *ump; 849 register struct ext2_sb_info *fs; 850 register struct ext2_super_block *es; 851 852 ump = VFSTOUFS(mp); 853 fs = ump->um_e2fs; 854 es = fs->s_es; 855 856 if (es->s_magic != EXT2_SUPER_MAGIC) 857 panic("ext2_statfs - magic number spoiled"); 858 859 /* 860 * Compute the overhead (FS structures) 861 */ 862 overhead_per_group = 1 /* super block */ + 863 fs->s_db_per_group + 864 1 /* block bitmap */ + 865 1 /* inode bitmap */ + 866 fs->s_itb_per_group; 867 overhead = es->s_first_data_block + 868 fs->s_groups_count * overhead_per_group; 869 870 sbp->f_bsize = EXT2_FRAG_SIZE(fs); 871 sbp->f_iosize = EXT2_BLOCK_SIZE(fs); 872 sbp->f_blocks = es->s_blocks_count - overhead; 873 sbp->f_bfree = es->s_free_blocks_count; 874 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count; 875 sbp->f_files = es->s_inodes_count; 876 sbp->f_ffree = es->s_free_inodes_count; 877 if (sbp != &mp->mnt_stat) { 878 sbp->f_type = mp->mnt_vfc->vfc_typenum; 879 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 880 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 881 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 882 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 883 } 884 return (0); 885} 886 887/* 888 * Go through the disk queues to initiate sandbagged IO; 889 * go through the inodes to write those that have been modified; 890 * initiate the writing of the super block if it has been modified. 891 * 892 * Note: we are always called with the filesystem marked `MPBUSY'. 893 */ 894static int 895ext2_sync(mp, waitfor, cred, p) 896 struct mount *mp; 897 int waitfor; 898 struct ucred *cred; 899 struct proc *p; 900{ 901 struct vnode *nvp, *vp; 902 struct inode *ip; 903 struct ufsmount *ump = VFSTOUFS(mp); 904 struct ext2_sb_info *fs; 905 int error, allerror = 0; 906 907 fs = ump->um_e2fs; 908 if (fs->s_dirt != 0 && fs->s_rd_only != 0) { /* XXX */ 909 printf("fs = %s\n", fs->fs_fsmnt); 910 panic("ext2_sync: rofs mod"); 911 } 912 /* 913 * Write back each (modified) inode. 914 */ 915 simple_lock(&mntvnode_slock); 916loop: 917 for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 918 /* 919 * If the vnode that we are about to sync is no longer 920 * associated with this mount point, start over. 921 */ 922 if (vp->v_mount != mp) 923 goto loop; 924 simple_lock(&vp->v_interlock); 925 nvp = vp->v_mntvnodes.le_next; 926 ip = VTOI(vp); 927 if (vp->v_type == VNON || 928 ((ip->i_flag & 929 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 930 (TAILQ_EMPTY(&vp->v_dirtyblkhd) || waitfor == MNT_LAZY))) { 931 simple_unlock(&vp->v_interlock); 932 continue; 933 } 934 simple_unlock(&mntvnode_slock); 935 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, p); 936 if (error) { 937 simple_lock(&mntvnode_slock); 938 if (error == ENOENT) 939 goto loop; 940 continue; 941 } 942 if ((error = VOP_FSYNC(vp, cred, waitfor, p)) != 0) 943 allerror = error; 944 VOP_UNLOCK(vp, 0, p); 945 vrele(vp); 946 simple_lock(&mntvnode_slock); 947 } 948 simple_unlock(&mntvnode_slock); 949 /* 950 * Force stale file system control information to be flushed. 951 */ 952 if (waitfor != MNT_LAZY) { 953 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, p); 954 if ((error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) != 0) 955 allerror = error; 956 VOP_UNLOCK(ump->um_devvp, 0, p); 957 } 958#if QUOTA 959 qsync(mp); 960#endif 961 /* 962 * Write back modified superblock. 963 */ 964 if (fs->s_dirt != 0) { 965 fs->s_dirt = 0; 966 fs->s_es->s_wtime = time_second; 967 if ((error = ext2_sbupdate(ump, waitfor)) != 0) 968 allerror = error; 969 } 970 return (allerror); 971} 972 973/* 974 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it 975 * in from disk. If it is in core, wait for the lock bit to clear, then 976 * return the inode locked. Detection and handling of mount points must be 977 * done by the calling routine. 978 */ 979static int 980ext2_vget(mp, ino, vpp) 981 struct mount *mp; 982 ino_t ino; 983 struct vnode **vpp; 984{ 985 register struct ext2_sb_info *fs; 986 register struct inode *ip; 987 struct ufsmount *ump; 988 struct buf *bp; 989 struct vnode *vp; 990 dev_t dev; 991 int i, error; 992 int used_blocks; 993 994 ump = VFSTOUFS(mp); 995 dev = ump->um_dev; 996restart: 997 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) 998 return (0); 999 1000 /* 1001 * Lock out the creation of new entries in the FFS hash table in 1002 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 1003 * may occur! 1004 */ 1005 if (ext2fs_inode_hash_lock) { 1006 while (ext2fs_inode_hash_lock) { 1007 ext2fs_inode_hash_lock = -1; 1008 tsleep(&ext2fs_inode_hash_lock, PVM, "e2vget", 0); 1009 } 1010 goto restart; 1011 } 1012 ext2fs_inode_hash_lock = 1; 1013 1014 /* 1015 * If this MALLOC() is performed after the getnewvnode() 1016 * it might block, leaving a vnode with a NULL v_data to be 1017 * found by ext2_sync() if a sync happens to fire right then, 1018 * which will cause a panic because ext2_sync() blindly 1019 * dereferences vp->v_data (as well it should). 1020 */ 1021 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK); 1022 1023 /* Allocate a new vnode/inode. */ 1024 if ((error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) != 0) { 1025 if (ext2fs_inode_hash_lock < 0) 1026 wakeup(&ext2fs_inode_hash_lock); 1027 ext2fs_inode_hash_lock = 0; 1028 *vpp = NULL; 1029 FREE(ip, M_EXT2NODE); 1030 return (error); 1031 } 1032 bzero((caddr_t)ip, sizeof(struct inode)); 1033 lockinit(&ip->i_lock, PINOD, "ext2in", 0, 0); 1034 vp->v_data = ip; 1035 ip->i_vnode = vp; 1036 ip->i_e2fs = fs = ump->um_e2fs; 1037 ip->i_dev = dev; 1038 ip->i_number = ino; 1039#if QUOTA 1040 for (i = 0; i < MAXQUOTAS; i++) 1041 ip->i_dquot[i] = NODQUOT; 1042#endif 1043 /* 1044 * Put it onto its hash chain and lock it so that other requests for 1045 * this inode will block if they arrive while we are sleeping waiting 1046 * for old data structures to be purged or for the contents of the 1047 * disk portion of this inode to be read. 1048 */ 1049 ufs_ihashins(ip); 1050 1051 if (ext2fs_inode_hash_lock < 0) 1052 wakeup(&ext2fs_inode_hash_lock); 1053 ext2fs_inode_hash_lock = 0; 1054 1055 /* Read in the disk contents for the inode, copy into the inode. */ 1056#if 0 1057printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino))); 1058#endif 1059 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 1060 (int)fs->s_blocksize, NOCRED, &bp)) != 0) { 1061 /* 1062 * The inode does not contain anything useful, so it would 1063 * be misleading to leave it on its hash chain. With mode 1064 * still zero, it will be unlinked and returned to the free 1065 * list by vput(). 1066 */ 1067 vput(vp); 1068 brelse(bp); 1069 *vpp = NULL; 1070 return (error); 1071 } 1072 /* convert ext2 inode to dinode */ 1073 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE * 1074 ino_to_fsbo(fs, ino)), &ip->i_din); 1075 ip->i_block_group = ino_to_cg(fs, ino); 1076 ip->i_next_alloc_block = 0; 1077 ip->i_next_alloc_goal = 0; 1078 ip->i_prealloc_count = 0; 1079 ip->i_prealloc_block = 0; 1080 /* now we want to make sure that block pointers for unused 1081 blocks are zeroed out - ext2_balloc depends on this 1082 although for regular files and directories only 1083 */ 1084 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) { 1085 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize; 1086 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) 1087 ip->i_db[i] = 0; 1088 } 1089/* 1090 ext2_print_inode(ip); 1091*/ 1092 brelse(bp); 1093 1094 /* 1095 * Initialize the vnode from the inode, check for aliases. 1096 * Note that the underlying vnode may have changed. 1097 */ 1098 if ((error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) != 0) { 1099 vput(vp); 1100 *vpp = NULL; 1101 return (error); 1102 } 1103 /* 1104 * Finish inode initialization now that aliasing has been resolved. 1105 */ 1106 ip->i_devvp = ump->um_devvp; 1107 VREF(ip->i_devvp); 1108 /* 1109 * Set up a generation number for this inode if it does not 1110 * already have one. This should only happen on old filesystems. 1111 */ 1112 if (ip->i_gen == 0) { 1113 ip->i_gen = random() / 2 + 1; 1114 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1115 ip->i_flag |= IN_MODIFIED; 1116 } 1117 *vpp = vp; 1118 return (0); 1119} 1120 1121/* 1122 * File handle to vnode 1123 * 1124 * Have to be really careful about stale file handles: 1125 * - check that the inode number is valid 1126 * - call ext2_vget() to get the locked inode 1127 * - check for an unallocated inode (i_mode == 0) 1128 * - check that the given client host has export rights and return 1129 * those rights via. exflagsp and credanonp 1130 */ 1131static int 1132ext2_fhtovp(mp, fhp, vpp) 1133 register struct mount *mp; 1134 struct fid *fhp; 1135 struct vnode **vpp; 1136{ 1137 register struct ufid *ufhp; 1138 struct ext2_sb_info *fs; 1139 1140 ufhp = (struct ufid *)fhp; 1141 fs = VFSTOUFS(mp)->um_e2fs; 1142 if (ufhp->ufid_ino < ROOTINO || 1143 ufhp->ufid_ino >= fs->s_groups_count * fs->s_es->s_inodes_per_group) 1144 return (ESTALE); 1145 return (ufs_fhtovp(mp, ufhp, vpp)); 1146} 1147 1148/* 1149 * Vnode pointer to File handle 1150 */ 1151/* ARGSUSED */ 1152static int 1153ext2_vptofh(vp, fhp) 1154 struct vnode *vp; 1155 struct fid *fhp; 1156{ 1157 register struct inode *ip; 1158 register struct ufid *ufhp; 1159 1160 ip = VTOI(vp); 1161 ufhp = (struct ufid *)fhp; 1162 ufhp->ufid_len = sizeof(struct ufid); 1163 ufhp->ufid_ino = ip->i_number; 1164 ufhp->ufid_gen = ip->i_gen; 1165 return (0); 1166} 1167 1168/* 1169 * Write a superblock and associated information back to disk. 1170 */ 1171static int 1172ext2_sbupdate(mp, waitfor) 1173 struct ufsmount *mp; 1174 int waitfor; 1175{ 1176 register struct ext2_sb_info *fs = mp->um_e2fs; 1177 register struct ext2_super_block *es = fs->s_es; 1178 register struct buf *bp; 1179 int error = 0; 1180/* 1181printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no"); 1182*/ 1183 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0); 1184 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block)); 1185 if (waitfor == MNT_WAIT) 1186 error = bwrite(bp); 1187 else 1188 bawrite(bp); 1189 1190 /* 1191 * The buffers for group descriptors, inode bitmaps and block bitmaps 1192 * are not busy at this point and are (hopefully) written by the 1193 * usual sync mechanism. No need to write them here 1194 */ 1195 1196 return (error); 1197} 1198