ext2_vfsops.c revision 31315
1219089Spjd/* 2219089Spjd * modified for EXT2FS support in Lites 1.1 3219089Spjd * 4219089Spjd * Aug 1995, Godmar Back (gback@cs.utah.edu) 5219089Spjd * University of Utah, Department of Computer Science 6219089Spjd */ 7219089Spjd/* 8219089Spjd * Copyright (c) 1989, 1991, 1993, 1994 9219089Spjd * The Regents of the University of California. All rights reserved. 10219089Spjd * 11219089Spjd * Redistribution and use in source and binary forms, with or without 12219089Spjd * modification, are permitted provided that the following conditions 13219089Spjd * are met: 14219089Spjd * 1. Redistributions of source code must retain the above copyright 15219089Spjd * notice, this list of conditions and the following disclaimer. 16219089Spjd * 2. Redistributions in binary form must reproduce the above copyright 17219089Spjd * notice, this list of conditions and the following disclaimer in the 18219089Spjd * documentation and/or other materials provided with the distribution. 19219089Spjd * 3. All advertising materials mentioning features or use of this software 20219089Spjd * must display the following acknowledgement: 21219089Spjd * This product includes software developed by the University of 22219089Spjd * California, Berkeley and its contributors. 23219089Spjd * 4. Neither the name of the University nor the names of its contributors 24219089Spjd * may be used to endorse or promote products derived from this software 25289422Smav * without specific prior written permission. 26289422Smav * 27289422Smav * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28219089Spjd * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29219089Spjd * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30219089Spjd * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31219089Spjd * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32219089Spjd * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33219089Spjd * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34219089Spjd * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35219089Spjd * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36219089Spjd * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37219089Spjd * SUCH DAMAGE. 38219089Spjd * 39219089Spjd * @(#)ffs_vfsops.c 8.8 (Berkeley) 4/18/94 40219089Spjd */ 41219089Spjd 42219089Spjd#if !defined(__FreeBSD__) 43289422Smav#include "quota.h" 44289422Smav#else 45289422Smav#include "opt_quota.h" 46289422Smav#endif 47289422Smav 48289422Smav#include <sys/param.h> 49219089Spjd#include <sys/systm.h> 50219089Spjd#include <sys/namei.h> 51219089Spjd#include <sys/proc.h> 52219089Spjd#include <sys/kernel.h> 53219089Spjd#include <sys/vnode.h> 54219089Spjd#include <sys/mount.h> 55#include <sys/buf.h> 56#include <sys/conf.h> 57#include <sys/fcntl.h> 58#include <sys/disklabel.h> 59#include <sys/malloc.h> 60#include <sys/stat.h> 61 62#include <miscfs/specfs/specdev.h> 63 64#include <ufs/ufs/quota.h> 65#include <ufs/ufs/ufsmount.h> 66#include <ufs/ufs/inode.h> 67#include <ufs/ufs/ufs_extern.h> 68 69#include <gnu/ext2fs/fs.h> 70#include <gnu/ext2fs/ext2_extern.h> 71#include <gnu/ext2fs/ext2_fs.h> 72#include <gnu/ext2fs/ext2_fs_sb.h> 73 74static int ext2_fhtovp __P((struct mount *, struct fid *, struct sockaddr *, 75 struct vnode **, int *, struct ucred **)); 76static int ext2_flushfiles __P((struct mount *mp, int flags, struct proc *p)); 77static int ext2_mount __P((struct mount *, 78 char *, caddr_t, struct nameidata *, struct proc *)); 79static int ext2_mountfs __P((struct vnode *, struct mount *, struct proc *)); 80static int ext2_reload __P((struct mount *mountp, struct ucred *cred, 81 struct proc *p)); 82static int ext2_sbupdate __P((struct ufsmount *, int)); 83static int ext2_statfs __P((struct mount *, struct statfs *, struct proc *)); 84static int ext2_sync __P((struct mount *, int, struct ucred *, struct proc *)); 85static int ext2_unmount __P((struct mount *, int, struct proc *)); 86static int ext2_vget __P((struct mount *, ino_t, struct vnode **)); 87static int ext2_vptofh __P((struct vnode *, struct fid *)); 88 89static MALLOC_DEFINE(M_EXT2NODE, "EXT2 node", "EXT2 vnode private part"); 90 91static struct vfsops ext2fs_vfsops = { 92 ext2_mount, 93 ufs_start, /* empty function */ 94 ext2_unmount, 95 ufs_root, /* root inode via vget */ 96 ufs_quotactl, /* does operations associated with quotas */ 97 ext2_statfs, 98 ext2_sync, 99 ext2_vget, 100 ext2_fhtovp, 101 ext2_vptofh, 102 ext2_init, 103}; 104 105#if defined(__FreeBSD__) 106VFS_SET(ext2fs_vfsops, ext2fs, MOUNT_EXT2FS, 0); 107#define bsd_malloc malloc 108#define bsd_free free 109#endif 110 111extern u_long nextgennumber; 112#ifdef __FreeBSD__ 113static int ext2fs_inode_hash_lock; 114#endif 115 116static int compute_sb_data __P((struct vnode * devvp, 117 struct ext2_super_block * es, 118 struct ext2_sb_info * fs)); 119 120#ifdef notyet 121static int ext2_mountroot __P((void)); 122 123/* 124 * Called by main() when ext2fs is going to be mounted as root. 125 * 126 * Name is updated by mount(8) after booting. 127 */ 128#define ROOTNAME "root_device" 129 130static int 131ext2_mountroot() 132{ 133#if !defined(__FreeBSD__) 134 extern struct vnode *rootvp; 135#endif 136 register struct ext2_sb_info *fs; 137 register struct mount *mp; 138#if defined(__FreeBSD__) 139 struct proc *p = curproc; 140#else 141 struct proc *p = get_proc(); /* XXX */ 142#endif 143 struct ufsmount *ump; 144 u_int size; 145 int error; 146 147 if ((error = bdevvp(rootdev, &rootvp))) { 148 printf("ext2_mountroot: can't find rootvp"); 149 return (error); 150 } 151 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 152 bzero((char *)mp, (u_long)sizeof(struct mount)); 153 mp->mnt_op = &ext2fs_vfsops; 154 mp->mnt_flag = MNT_RDONLY; 155 if (bdevsw[major(rootdev)]->d_flags & D_NOCLUSTERR) 156 mp->mnt_flag |= MNT_NOCLUSTERR; 157 if (bdevsw[major(rootdev)]->d_flags & D_NOCLUSTERW) 158 mp->mnt_flag |= MNT_NOCLUSTERW; 159 if (error = ext2_mountfs(rootvp, mp, p)) { 160 bsd_free(mp, M_MOUNT); 161 return (error); 162 } 163 if (error = vfs_lock(mp)) { 164 (void)ext2_unmount(mp, 0, p); 165 bsd_free(mp, M_MOUNT); 166 return (error); 167 } 168#if defined(__FreeBSD__) 169 CIRCLEQ_INSERT_HEAD(&mountlist, mp, mnt_list); 170#else 171 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 172#endif 173 mp->mnt_flag |= MNT_ROOTFS; 174 mp->mnt_vnodecovered = NULLVP; 175 ump = VFSTOUFS(mp); 176 fs = ump->um_e2fs; 177 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt)); 178 fs->fs_fsmnt[0] = '/'; 179 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 180 MNAMELEN); 181 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 182 &size); 183 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 184 (void)ext2_statfs(mp, &mp->mnt_stat, p); 185 vfs_unlock(mp); 186 inittodr(fs->s_es->s_wtime); /* this helps to set the time */ 187 return (0); 188} 189#endif 190 191/* 192 * VFS Operations. 193 * 194 * mount system call 195 */ 196static int 197ext2_mount(mp, path, data, ndp, p) 198 register struct mount *mp; 199 char *path; 200 caddr_t data; /* this is actually a (struct ufs_args *) */ 201 struct nameidata *ndp; 202 struct proc *p; 203{ 204 struct vnode *devvp; 205 struct ufs_args args; 206 struct ufsmount *ump = 0; 207 register struct ext2_sb_info *fs; 208 u_int size; 209 int error, flags; 210 211 if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) 212 return (error); 213 /* 214 * If updating, check whether changing from read-only to 215 * read/write; if there is no device name, that's all we do. 216 * Disallow clearing MNT_NOCLUSTERR and MNT_NOCLUSTERW flags, 217 * if block device requests. 218 */ 219 if (mp->mnt_flag & MNT_UPDATE) { 220 ump = VFSTOUFS(mp); 221 fs = ump->um_e2fs; 222 error = 0; 223 if (bdevsw[major(ump->um_dev)]->d_flags & D_NOCLUSTERR) 224 mp->mnt_flag |= MNT_NOCLUSTERR; 225 if (bdevsw[major(ump->um_dev)]->d_flags & D_NOCLUSTERW) 226 mp->mnt_flag |= MNT_NOCLUSTERW; 227 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) { 228 flags = WRITECLOSE; 229 if (mp->mnt_flag & MNT_FORCE) 230 flags |= FORCECLOSE; 231 if (vfs_busy(mp, LK_NOWAIT, 0, p)) 232 return (EBUSY); 233 error = ext2_flushfiles(mp, flags, p); 234 vfs_unbusy(mp, p); 235 } 236 if (!error && (mp->mnt_flag & MNT_RELOAD)) 237 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, p); 238 if (error) 239 return (error); 240 if (fs->s_rd_only && (mp->mnt_kern_flag & MNTK_WANTRDWR)) 241 fs->s_rd_only = 0; 242 if (fs->s_rd_only == 0) { 243 /* don't say it's clean */ 244 fs->s_es->s_state &= ~EXT2_VALID_FS; 245 ext2_sbupdate(ump, MNT_WAIT); 246 } 247 if (args.fspec == 0) { 248 /* 249 * Process export requests. 250 */ 251 return (vfs_export(mp, &ump->um_export, &args.export)); 252 } 253 } 254 /* 255 * Not an update, or updating the name: look up the name 256 * and verify that it refers to a sensible block device. 257 */ 258 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p); 259 if (error = namei(ndp)) 260 return (error); 261 devvp = ndp->ni_vp; 262 263 if (devvp->v_type != VBLK) { 264 vrele(devvp); 265 return (ENOTBLK); 266 } 267 if (major(devvp->v_rdev) >= nblkdev) { 268 vrele(devvp); 269 return (ENXIO); 270 } 271 if ((mp->mnt_flag & MNT_UPDATE) == 0) { 272 if (bdevsw[major(devvp->v_rdev)]->d_flags & D_NOCLUSTERR) 273 mp->mnt_flag |= MNT_NOCLUSTERR; 274 if (bdevsw[major(devvp->v_rdev)]->d_flags & D_NOCLUSTERW) 275 mp->mnt_flag |= MNT_NOCLUSTERW; 276 error = ext2_mountfs(devvp, mp, p); 277 } else { 278 if (devvp != ump->um_devvp) 279 error = EINVAL; /* needs translation */ 280 else 281 vrele(devvp); 282 } 283 if (error) { 284 vrele(devvp); 285 return (error); 286 } 287 ump = VFSTOUFS(mp); 288 fs = ump->um_e2fs; 289 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size); 290 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size); 291 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 292 MNAMELEN); 293 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 294 &size); 295 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 296 (void)ext2_statfs(mp, &mp->mnt_stat, p); 297 return (0); 298} 299 300/* 301 * checks that the data in the descriptor blocks make sense 302 * this is taken from ext2/super.c 303 */ 304static int ext2_check_descriptors (struct ext2_sb_info * sb) 305{ 306 int i; 307 int desc_block = 0; 308 unsigned long block = sb->s_es->s_first_data_block; 309 struct ext2_group_desc * gdp = NULL; 310 311 /* ext2_debug ("Checking group descriptors"); */ 312 313 for (i = 0; i < sb->s_groups_count; i++) 314 { 315 /* examine next descriptor block */ 316 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0) 317 gdp = (struct ext2_group_desc *) 318 sb->s_group_desc[desc_block++]->b_data; 319 if (gdp->bg_block_bitmap < block || 320 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 321 { 322 printf ("ext2_check_descriptors: " 323 "Block bitmap for group %d" 324 " not in group (block %lu)!", 325 i, (unsigned long) gdp->bg_block_bitmap); 326 return 0; 327 } 328 if (gdp->bg_inode_bitmap < block || 329 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 330 { 331 printf ("ext2_check_descriptors: " 332 "Inode bitmap for group %d" 333 " not in group (block %lu)!", 334 i, (unsigned long) gdp->bg_inode_bitmap); 335 return 0; 336 } 337 if (gdp->bg_inode_table < block || 338 gdp->bg_inode_table + sb->s_itb_per_group >= 339 block + EXT2_BLOCKS_PER_GROUP(sb)) 340 { 341 printf ("ext2_check_descriptors: " 342 "Inode table for group %d" 343 " not in group (block %lu)!", 344 i, (unsigned long) gdp->bg_inode_table); 345 return 0; 346 } 347 block += EXT2_BLOCKS_PER_GROUP(sb); 348 gdp++; 349 } 350 return 1; 351} 352 353/* 354 * this computes the fields of the ext2_sb_info structure from the 355 * data in the ext2_super_block structure read in 356 */ 357static int compute_sb_data(devvp, es, fs) 358 struct vnode * devvp; 359 struct ext2_super_block * es; 360 struct ext2_sb_info * fs; 361{ 362 int db_count, error; 363 int i, j; 364 int logic_sb_block = 1; /* XXX for now */ 365 366#if 1 367#define V(v) 368#else 369#define V(v) printf(#v"= %d\n", fs->v); 370#endif 371 372 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size; 373 V(s_blocksize) 374 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size; 375 V(s_bshift) 376 fs->s_fsbtodb = es->s_log_block_size + 1; 377 V(s_fsbtodb) 378 fs->s_qbmask = fs->s_blocksize - 1; 379 V(s_bmask) 380 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es); 381 V(s_blocksize_bits) 382 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size; 383 V(s_frag_size) 384 if (fs->s_frag_size) 385 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size; 386 V(s_frags_per_block) 387 fs->s_blocks_per_group = es->s_blocks_per_group; 388 V(s_blocks_per_group) 389 fs->s_frags_per_group = es->s_frags_per_group; 390 V(s_frags_per_group) 391 fs->s_inodes_per_group = es->s_inodes_per_group; 392 V(s_inodes_per_group) 393 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE; 394 V(s_inodes_per_block) 395 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block; 396 V(s_itb_per_group) 397 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc); 398 V(s_desc_per_block) 399 /* s_resuid / s_resgid ? */ 400 fs->s_groups_count = (es->s_blocks_count - 401 es->s_first_data_block + 402 EXT2_BLOCKS_PER_GROUP(fs) - 1) / 403 EXT2_BLOCKS_PER_GROUP(fs); 404 V(s_groups_count) 405 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) / 406 EXT2_DESC_PER_BLOCK(fs); 407 fs->s_db_per_group = db_count; 408 V(s_db_per_group) 409 410 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *), 411 M_UFSMNT, M_WAITOK); 412 413 /* adjust logic_sb_block */ 414 if(fs->s_blocksize > SBSIZE) 415 /* Godmar thinks: if the blocksize is greater than 1024, then 416 the superblock is logically part of block zero. 417 */ 418 logic_sb_block = 0; 419 420 for (i = 0; i < db_count; i++) { 421 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1), 422 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]); 423 if(error) { 424 for (j = 0; j < i; j++) 425 brelse(fs->s_group_desc[j]); 426 bsd_free(fs->s_group_desc, M_UFSMNT); 427 printf("EXT2-fs: unable to read group descriptors (%d)\n", error); 428 return EIO; 429 } 430 /* Set the B_LOCKED flag on the buffer, then brelse() it */ 431 LCK_BUF(fs->s_group_desc[i]) 432 } 433 if(!ext2_check_descriptors(fs)) { 434 for (j = 0; j < db_count; j++) 435 ULCK_BUF(fs->s_group_desc[j]) 436 bsd_free(fs->s_group_desc, M_UFSMNT); 437 printf("EXT2-fs: (ext2_check_descriptors failure) " 438 "unable to read group descriptors\n"); 439 return EIO; 440 } 441 442 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) { 443 fs->s_inode_bitmap_number[i] = 0; 444 fs->s_inode_bitmap[i] = NULL; 445 fs->s_block_bitmap_number[i] = 0; 446 fs->s_block_bitmap[i] = NULL; 447 } 448 fs->s_loaded_inode_bitmaps = 0; 449 fs->s_loaded_block_bitmaps = 0; 450 return 0; 451} 452 453/* 454 * Reload all incore data for a filesystem (used after running fsck on 455 * the root filesystem and finding things to fix). The filesystem must 456 * be mounted read-only. 457 * 458 * Things to do to update the mount: 459 * 1) invalidate all cached meta-data. 460 * 2) re-read superblock from disk. 461 * 3) re-read summary information from disk. 462 * 4) invalidate all inactive vnodes. 463 * 5) invalidate all cached file data. 464 * 6) re-read inode data for all active vnodes. 465 */ 466static int 467ext2_reload(mountp, cred, p) 468 register struct mount *mountp; 469 struct ucred *cred; 470 struct proc *p; 471{ 472 register struct vnode *vp, *nvp, *devvp; 473 struct inode *ip; 474 struct buf *bp; 475 struct ext2_super_block * es; 476 struct ext2_sb_info *fs; 477 int error; 478 479 if ((mountp->mnt_flag & MNT_RDONLY) == 0) 480 return (EINVAL); 481 /* 482 * Step 1: invalidate all cached meta-data. 483 */ 484 devvp = VFSTOUFS(mountp)->um_devvp; 485 if (vinvalbuf(devvp, 0, cred, p, 0, 0)) 486 panic("ext2_reload: dirty1"); 487 /* 488 * Step 2: re-read superblock from disk. 489 * constants have been adjusted for ext2 490 */ 491 if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) 492 return (error); 493 es = (struct ext2_super_block *)bp->b_data; 494 if (es->s_magic != EXT2_SUPER_MAGIC) { 495 if(es->s_magic == EXT2_PRE_02B_MAGIC) 496 printf("This filesystem bears the magic number of a pre " 497 "0.2b version of ext2. This is not supported by " 498 "Lites.\n"); 499 else 500 printf("Wrong magic number: %x (expected %x for ext2 fs\n", 501 es->s_magic, EXT2_SUPER_MAGIC); 502 brelse(bp); 503 return (EIO); /* XXX needs translation */ 504 } 505 fs = VFSTOUFS(mountp)->um_e2fs; 506 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block)); 507 508 if(error = compute_sb_data(devvp, es, fs)) { 509 brelse(bp); 510 return error; 511 } 512#ifdef UNKLAR 513 if (fs->fs_sbsize < SBSIZE) 514 bp->b_flags |= B_INVAL; 515#endif 516 brelse(bp); 517 518loop: 519 for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 520 nvp = vp->v_mntvnodes.le_next; 521 /* 522 * Step 4: invalidate all inactive vnodes. 523 */ 524 if (vp->v_usecount == 0) { 525 vgone(vp); 526 continue; 527 } 528 /* 529 * Step 5: invalidate all cached file data. 530 */ 531 if (vget(vp, LK_EXCLUSIVE, p)) 532 goto loop; 533 if (vinvalbuf(vp, 0, cred, p, 0, 0)) 534 panic("ext2_reload: dirty2"); 535 /* 536 * Step 6: re-read inode data for all active vnodes. 537 */ 538 ip = VTOI(vp); 539 if (error = 540 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 541 (int)fs->s_blocksize, NOCRED, &bp)) { 542 vput(vp); 543 return (error); 544 } 545 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + 546 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)), 547 &ip->i_din); 548 brelse(bp); 549 vput(vp); 550 if (vp->v_mount != mountp) 551 goto loop; 552 } 553 return (0); 554} 555 556/* 557 * Common code for mount and mountroot 558 */ 559static int 560ext2_mountfs(devvp, mp, p) 561 register struct vnode *devvp; 562 struct mount *mp; 563 struct proc *p; 564{ 565 register struct ufsmount *ump; 566 struct buf *bp; 567 register struct ext2_sb_info *fs; 568 struct ext2_super_block * es; 569 dev_t dev = devvp->v_rdev; 570 struct partinfo dpart; 571 int havepart = 0; 572 int error, i, size; 573 int ronly; 574#if !defined(__FreeBSD__) 575 extern struct vnode *rootvp; 576#endif 577 578 /* 579 * Disallow multiple mounts of the same device. 580 * Disallow mounting of a device that is currently in use 581 * (except for root, which might share swap device for miniroot). 582 * Flush out any old buffers remaining from a previous use. 583 */ 584 if (error = vfs_mountedon(devvp)) 585 return (error); 586 if (vcount(devvp) > 1 && devvp != rootvp) 587 return (EBUSY); 588 if (error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0)) 589 return (error); 590#ifdef READONLY 591/* turn on this to force it to be read-only */ 592 mp->mnt_flag |= MNT_RDONLY; 593#endif 594 595 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 596 if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p)) 597 return (error); 598 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0) 599 size = DEV_BSIZE; 600 else { 601 havepart = 1; 602 size = dpart.disklab->d_secsize; 603 } 604 605 bp = NULL; 606 ump = NULL; 607 if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) 608 goto out; 609 es = (struct ext2_super_block *)bp->b_data; 610 if (es->s_magic != EXT2_SUPER_MAGIC) { 611 if(es->s_magic == EXT2_PRE_02B_MAGIC) 612 printf("This filesystem bears the magic number of a pre " 613 "0.2b version of ext2. This is not supported by " 614 "Lites.\n"); 615 else 616 printf("Wrong magic number: %x (expected %x for EXT2FS)\n", 617 es->s_magic, EXT2_SUPER_MAGIC); 618 error = EINVAL; /* XXX needs translation */ 619 goto out; 620 } 621 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK); 622 bzero((caddr_t)ump, sizeof *ump); 623 ump->um_malloctype = M_EXT2NODE; 624 ump->um_blkatoff = ext2_blkatoff; 625 ump->um_truncate = ext2_truncate; 626 ump->um_update = ext2_update; 627 ump->um_valloc = ext2_valloc; 628 ump->um_vfree = ext2_vfree; 629 /* I don't know whether this is the right strategy. Note that 630 we dynamically allocate both a ext2_sb_info and a ext2_super_block 631 while Linux keeps the super block in a locked buffer 632 */ 633 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info), 634 M_UFSMNT, M_WAITOK); 635 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block), 636 M_UFSMNT, M_WAITOK); 637 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block)); 638 if(error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)) { 639 brelse(bp); 640 return error; 641 } 642 brelse(bp); 643 bp = NULL; 644 fs = ump->um_e2fs; 645 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */ 646 if (!(fs->s_es->s_state & EXT2_VALID_FS)) { 647 printf("WARNING: %s was not properly dismounted\n", 648 fs->fs_fsmnt); 649 } 650 /* if the fs is not mounted read-only, make sure the super block is 651 always written back on a sync() 652 */ 653 if (ronly == 0) { 654 fs->s_dirt = 1; /* mark it modified */ 655 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */ 656 } 657 mp->mnt_data = (qaddr_t)ump; 658 mp->mnt_stat.f_fsid.val[0] = (long)dev; 659 mp->mnt_stat.f_fsid.val[1] = MOUNT_EXT2FS; 660 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; 661 mp->mnt_flag |= MNT_LOCAL; 662 ump->um_mountp = mp; 663 ump->um_dev = dev; 664 ump->um_devvp = devvp; 665 /* setting those two parameters allows us to use 666 ufs_bmap w/o changse ! 667 */ 668 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); 669 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1; 670 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); 671 for (i = 0; i < MAXQUOTAS; i++) 672 ump->um_quotas[i] = NULLVP; 673 devvp->v_specflags |= SI_MOUNTEDON; 674 if (ronly == 0) 675 ext2_sbupdate(ump, MNT_WAIT); 676 return (0); 677out: 678 if (bp) 679 brelse(bp); 680 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p); 681 if (ump) { 682 bsd_free(ump->um_fs, M_UFSMNT); 683 bsd_free(ump, M_UFSMNT); 684 mp->mnt_data = (qaddr_t)0; 685 } 686 return (error); 687} 688 689/* 690 * unmount system call 691 */ 692static int 693ext2_unmount(mp, mntflags, p) 694 struct mount *mp; 695 int mntflags; 696 struct proc *p; 697{ 698 register struct ufsmount *ump; 699 register struct ext2_sb_info *fs; 700 int error, flags, ronly, i; 701 702 flags = 0; 703 if (mntflags & MNT_FORCE) { 704 if (mp->mnt_flag & MNT_ROOTFS) 705 return (EINVAL); 706 flags |= FORCECLOSE; 707 } 708 if (error = ext2_flushfiles(mp, flags, p)) 709 return (error); 710 ump = VFSTOUFS(mp); 711 fs = ump->um_e2fs; 712 ronly = fs->s_rd_only; 713 if (!ronly) { 714 fs->s_es->s_state |= EXT2_VALID_FS; /* was fs_clean = 1 */ 715 ext2_sbupdate(ump, MNT_WAIT); 716 } 717 718 /* release buffers containing group descriptors */ 719 for(i = 0; i < fs->s_db_per_group; i++) 720 ULCK_BUF(fs->s_group_desc[i]) 721 722 /* release cached inode/block bitmaps */ 723 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 724 if (fs->s_inode_bitmap[i]) 725 ULCK_BUF(fs->s_inode_bitmap[i]) 726 727 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 728 if (fs->s_block_bitmap[i]) 729 ULCK_BUF(fs->s_block_bitmap[i]) 730 731 ump->um_devvp->v_specflags &= ~SI_MOUNTEDON; 732 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, 733 NOCRED, p); 734 vrele(ump->um_devvp); 735 bsd_free(fs->s_es, M_UFSMNT); 736 bsd_free(fs, M_UFSMNT); 737 bsd_free(ump, M_UFSMNT); 738 mp->mnt_data = (qaddr_t)0; 739 mp->mnt_flag &= ~MNT_LOCAL; 740 return (error); 741} 742 743/* 744 * Flush out all the files in a filesystem. 745 */ 746static int 747ext2_flushfiles(mp, flags, p) 748 register struct mount *mp; 749 int flags; 750 struct proc *p; 751{ 752 register struct ufsmount *ump; 753 int error; 754#if QUOTA 755 int i; 756#endif 757 758 ump = VFSTOUFS(mp); 759#if QUOTA 760 if (mp->mnt_flag & MNT_QUOTA) { 761 if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) 762 return (error); 763 for (i = 0; i < MAXQUOTAS; i++) { 764 if (ump->um_quotas[i] == NULLVP) 765 continue; 766 quotaoff(p, mp, i); 767 } 768 /* 769 * Here we fall through to vflush again to ensure 770 * that we have gotten rid of all the system vnodes. 771 */ 772 } 773#endif 774 error = vflush(mp, NULLVP, flags); 775 return (error); 776} 777 778/* 779 * Get file system statistics. 780 * taken from ext2/super.c ext2_statfs 781 */ 782static int 783ext2_statfs(mp, sbp, p) 784 struct mount *mp; 785 register struct statfs *sbp; 786 struct proc *p; 787{ 788 unsigned long overhead; 789 unsigned long overhead_per_group; 790 791 register struct ufsmount *ump; 792 register struct ext2_sb_info *fs; 793 register struct ext2_super_block *es; 794 795 ump = VFSTOUFS(mp); 796 fs = ump->um_e2fs; 797 es = fs->s_es; 798 799 if (es->s_magic != EXT2_SUPER_MAGIC) 800 panic("ext2_statfs - magic number spoiled"); 801 802 /* 803 * Compute the overhead (FS structures) 804 */ 805 overhead_per_group = 1 /* super block */ + 806 fs->s_db_per_group + 807 1 /* block bitmap */ + 808 1 /* inode bitmap */ + 809 fs->s_itb_per_group; 810 overhead = es->s_first_data_block + 811 fs->s_groups_count * overhead_per_group; 812 813 sbp->f_type = MOUNT_EXT2FS; 814 sbp->f_bsize = EXT2_FRAG_SIZE(fs); 815 sbp->f_iosize = EXT2_BLOCK_SIZE(fs); 816 sbp->f_blocks = es->s_blocks_count - overhead; 817 sbp->f_bfree = es->s_free_blocks_count; 818 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count; 819 sbp->f_files = es->s_inodes_count; 820 sbp->f_ffree = es->s_free_inodes_count; 821 if (sbp != &mp->mnt_stat) { 822 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 823 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 824 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 825 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 826 } 827 return (0); 828} 829 830/* 831 * Go through the disk queues to initiate sandbagged IO; 832 * go through the inodes to write those that have been modified; 833 * initiate the writing of the super block if it has been modified. 834 * 835 * Note: we are always called with the filesystem marked `MPBUSY'. 836 */ 837static int 838ext2_sync(mp, waitfor, cred, p) 839 struct mount *mp; 840 int waitfor; 841 struct ucred *cred; 842 struct proc *p; 843{ 844 register struct vnode *vp; 845 register struct inode *ip; 846 register struct ufsmount *ump = VFSTOUFS(mp); 847 register struct ext2_sb_info *fs; 848 int error, allerror = 0; 849 850 fs = ump->um_e2fs; 851 /* 852 * Write back modified superblock. 853 * Consistency check that the superblock 854 * is still in the buffer cache. 855 */ 856 if (fs->s_dirt) { 857#if !defined(__FreeBSD__) 858 struct timeval time; 859#endif 860 861 if (fs->s_rd_only != 0) { /* XXX */ 862 printf("fs = %s\n", fs->fs_fsmnt); 863 panic("update: rofs mod"); 864 } 865 fs->s_dirt = 0; 866#if !defined(__FreeBSD__) 867 get_time(&time); 868#endif 869 fs->s_es->s_wtime = time.tv_sec; 870 allerror = ext2_sbupdate(ump, waitfor); 871 } 872 /* 873 * Write back each (modified) inode. 874 */ 875loop: 876 for (vp = mp->mnt_vnodelist.lh_first; 877 vp != NULL; 878 vp = vp->v_mntvnodes.le_next) { 879 /* 880 * If the vnode that we are about to sync is no longer 881 * associated with this mount point, start over. 882 */ 883 if (vp->v_mount != mp) 884 goto loop; 885 if (VOP_ISLOCKED(vp)) 886 continue; 887 ip = VTOI(vp); 888 if ((ip->i_flag & 889 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 890 vp->v_dirtyblkhd.lh_first == NULL) 891 continue; 892 if (vget(vp, LK_EXCLUSIVE, p)) 893 goto loop; 894 if (error = VOP_FSYNC(vp, cred, waitfor, p)) 895 allerror = error; 896 vput(vp); 897 } 898 /* 899 * Force stale file system control information to be flushed. 900 */ 901 if (error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) 902 allerror = error; 903#if QUOTA 904 qsync(mp); 905#endif 906 return (allerror); 907} 908 909/* 910 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it 911 * in from disk. If it is in core, wait for the lock bit to clear, then 912 * return the inode locked. Detection and handling of mount points must be 913 * done by the calling routine. 914 */ 915static int 916ext2_vget(mp, ino, vpp) 917 struct mount *mp; 918 ino_t ino; 919 struct vnode **vpp; 920{ 921 register struct ext2_sb_info *fs; 922 register struct inode *ip; 923 struct ufsmount *ump; 924 struct buf *bp; 925 struct vnode *vp; 926 dev_t dev; 927 int i, error; 928 int used_blocks; 929 930 ump = VFSTOUFS(mp); 931 dev = ump->um_dev; 932restart: 933 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) 934 return (0); 935 936#ifdef __FreeBSD__ 937 /* 938 * Lock out the creation of new entries in the FFS hash table in 939 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 940 * may occur! 941 */ 942 if (ext2fs_inode_hash_lock) { 943 while (ext2fs_inode_hash_lock) { 944 ext2fs_inode_hash_lock = -1; 945 tsleep(&ext2fs_inode_hash_lock, PVM, "ffsvgt", 0); 946 } 947 goto restart; 948 } 949 ext2fs_inode_hash_lock = 1; 950#endif 951 952 /* Allocate a new vnode/inode. */ 953 if (error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) { 954 *vpp = NULL; 955 return (error); 956 } 957 MALLOC(ip, struct inode *, sizeof(struct inode), M_EXT2NODE, M_WAITOK); 958#ifndef __FreeBSD__ 959 insmntque(vp, mp); 960#endif 961 bzero((caddr_t)ip, sizeof(struct inode)); 962 vp->v_data = ip; 963 ip->i_vnode = vp; 964 ip->i_e2fs = fs = ump->um_e2fs; 965 ip->i_dev = dev; 966 ip->i_number = ino; 967#if QUOTA 968 for (i = 0; i < MAXQUOTAS; i++) 969 ip->i_dquot[i] = NODQUOT; 970#endif 971 /* 972 * Put it onto its hash chain and lock it so that other requests for 973 * this inode will block if they arrive while we are sleeping waiting 974 * for old data structures to be purged or for the contents of the 975 * disk portion of this inode to be read. 976 */ 977 ufs_ihashins(ip); 978 979#ifdef __FreeBSD__ 980 if (ext2fs_inode_hash_lock < 0) 981 wakeup(&ext2fs_inode_hash_lock); 982 ext2fs_inode_hash_lock = 0; 983#endif 984 985 /* Read in the disk contents for the inode, copy into the inode. */ 986 /* Read in the disk contents for the inode, copy into the inode. */ 987#if 0 988printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino))); 989#endif 990 if (error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 991 (int)fs->s_blocksize, NOCRED, &bp)) { 992 /* 993 * The inode does not contain anything useful, so it would 994 * be misleading to leave it on its hash chain. With mode 995 * still zero, it will be unlinked and returned to the free 996 * list by vput(). 997 */ 998 vput(vp); 999 brelse(bp); 1000 *vpp = NULL; 1001 return (error); 1002 } 1003 /* convert ext2 inode to dinode */ 1004 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE * 1005 ino_to_fsbo(fs, ino)), &ip->i_din); 1006 ip->i_block_group = ino_to_cg(fs, ino); 1007 ip->i_next_alloc_block = 0; 1008 ip->i_next_alloc_goal = 0; 1009 ip->i_prealloc_count = 0; 1010 ip->i_prealloc_block = 0; 1011 /* now we want to make sure that block pointers for unused 1012 blocks are zeroed out - ext2_balloc depends on this 1013 although for regular files and directories only 1014 */ 1015 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) { 1016 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize; 1017 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) 1018 ip->i_db[i] = 0; 1019 } 1020/* 1021 ext2_print_inode(ip); 1022*/ 1023 brelse(bp); 1024 1025 /* 1026 * Initialize the vnode from the inode, check for aliases. 1027 * Note that the underlying vnode may have changed. 1028 */ 1029 if (error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) { 1030 vput(vp); 1031 *vpp = NULL; 1032 return (error); 1033 } 1034 /* 1035 * Finish inode initialization now that aliasing has been resolved. 1036 */ 1037 ip->i_devvp = ump->um_devvp; 1038 VREF(ip->i_devvp); 1039 /* 1040 * Set up a generation number for this inode if it does not 1041 * already have one. This should only happen on old filesystems. 1042 */ 1043 if (ip->i_gen == 0) { 1044#if !defined(__FreeBSD__) 1045 struct timeval time; 1046 get_time(&time); 1047#endif 1048 if (++nextgennumber < (u_long)time.tv_sec) 1049 nextgennumber = time.tv_sec; 1050 ip->i_gen = nextgennumber; 1051 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1052 ip->i_flag |= IN_MODIFIED; 1053 } 1054 *vpp = vp; 1055 return (0); 1056} 1057 1058/* 1059 * File handle to vnode 1060 * 1061 * Have to be really careful about stale file handles: 1062 * - check that the inode number is valid 1063 * - call ext2_vget() to get the locked inode 1064 * - check for an unallocated inode (i_mode == 0) 1065 * - check that the given client host has export rights and return 1066 * those rights via. exflagsp and credanonp 1067 */ 1068static int 1069ext2_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) 1070 register struct mount *mp; 1071 struct fid *fhp; 1072 struct sockaddr *nam; 1073 struct vnode **vpp; 1074 int *exflagsp; 1075 struct ucred **credanonp; 1076{ 1077 register struct ufid *ufhp; 1078 struct ext2_sb_info *fs; 1079 1080 ufhp = (struct ufid *)fhp; 1081 fs = VFSTOUFS(mp)->um_e2fs; 1082 if (ufhp->ufid_ino < ROOTINO || 1083 ufhp->ufid_ino >= fs->s_groups_count * fs->s_es->s_inodes_per_group) 1084 return (ESTALE); 1085 return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp)); 1086} 1087 1088/* 1089 * Vnode pointer to File handle 1090 */ 1091/* ARGSUSED */ 1092static int 1093ext2_vptofh(vp, fhp) 1094 struct vnode *vp; 1095 struct fid *fhp; 1096{ 1097 register struct inode *ip; 1098 register struct ufid *ufhp; 1099 1100 ip = VTOI(vp); 1101 ufhp = (struct ufid *)fhp; 1102 ufhp->ufid_len = sizeof(struct ufid); 1103 ufhp->ufid_ino = ip->i_number; 1104 ufhp->ufid_gen = ip->i_gen; 1105 return (0); 1106} 1107 1108/* 1109 * Write a superblock and associated information back to disk. 1110 */ 1111static int 1112ext2_sbupdate(mp, waitfor) 1113 struct ufsmount *mp; 1114 int waitfor; 1115{ 1116 register struct ext2_sb_info *fs = mp->um_e2fs; 1117 register struct ext2_super_block *es = fs->s_es; 1118 register struct buf *bp; 1119 int i, error = 0; 1120/* 1121printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no"); 1122*/ 1123 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0); 1124 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block)); 1125 if (waitfor == MNT_WAIT) 1126 error = bwrite(bp); 1127 else 1128 bawrite(bp); 1129 1130 /* 1131 * The buffers for group descriptors, inode bitmaps and block bitmaps 1132 * are not busy at this point and are (hopefully) written by the 1133 * usual sync mechanism. No need to write them here 1134 */ 1135 1136 return (error); 1137} 1138