ext2_vfsops.c revision 24203
1/* 2 * modified for EXT2FS support in Lites 1.1 3 * 4 * Aug 1995, Godmar Back (gback@cs.utah.edu) 5 * University of Utah, Department of Computer Science 6 */ 7/* 8 * Copyright (c) 1989, 1991, 1993, 1994 9 * The Regents of the University of California. All rights reserved. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the University of 22 * California, Berkeley and its contributors. 23 * 4. Neither the name of the University nor the names of its contributors 24 * may be used to endorse or promote products derived from this software 25 * without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37 * SUCH DAMAGE. 38 * 39 * @(#)ffs_vfsops.c 8.8 (Berkeley) 4/18/94 40 */ 41 42#if !defined(__FreeBSD__) 43#include "quota.h" 44#else 45#include "opt_quota.h" 46#endif 47 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/namei.h> 51#include <sys/proc.h> 52#include <sys/kernel.h> 53#include <sys/vnode.h> 54#include <sys/socket.h> 55#include <sys/mount.h> 56#include <sys/buf.h> 57#include <sys/mbuf.h> 58#include <sys/fcntl.h> 59#include <sys/disklabel.h> 60#include <sys/errno.h> 61#include <sys/malloc.h> 62#include <sys/stat.h> 63 64#include <miscfs/specfs/specdev.h> 65 66#include <ufs/ufs/quota.h> 67#include <ufs/ufs/ufsmount.h> 68#include <ufs/ufs/inode.h> 69#include <ufs/ufs/ufs_extern.h> 70 71#include <gnu/ext2fs/fs.h> 72#include <gnu/ext2fs/ext2_extern.h> 73#include <gnu/ext2fs/ext2_fs.h> 74#include <gnu/ext2fs/ext2_fs_sb.h> 75 76static int ext2_fhtovp __P((struct mount *, struct fid *, struct mbuf *, 77 struct vnode **, int *, struct ucred **)); 78static int ext2_flushfiles __P((struct mount *mp, int flags, struct proc *p)); 79static int ext2_mount __P((struct mount *, 80 char *, caddr_t, struct nameidata *, struct proc *)); 81static int ext2_mountfs __P((struct vnode *, struct mount *, struct proc *)); 82static int ext2_reload __P((struct mount *mountp, struct ucred *cred, 83 struct proc *p)); 84static int ext2_sbupdate __P((struct ufsmount *, int)); 85static int ext2_statfs __P((struct mount *, struct statfs *, struct proc *)); 86static int ext2_sync __P((struct mount *, int, struct ucred *, struct proc *)); 87static int ext2_unmount __P((struct mount *, int, struct proc *)); 88static int ext2_vget __P((struct mount *, ino_t, struct vnode **)); 89static int ext2_vptofh __P((struct vnode *, struct fid *)); 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 /* 148 * Get vnodes for swapdev and rootdev. 149 */ 150 if (bdevvp(swapdev, &swapdev_vp) || bdevvp(rootdev, &rootvp)) 151 panic("ext2_mountroot: can't setup bdevvp's"); 152 153 mp = bsd_malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK); 154 bzero((char *)mp, (u_long)sizeof(struct mount)); 155 mp->mnt_op = &ext2fs_vfsops; 156 mp->mnt_flag = MNT_RDONLY; 157 if (error = ext2_mountfs(rootvp, mp, p)) { 158 bsd_free(mp, M_MOUNT); 159 return (error); 160 } 161 if (error = vfs_lock(mp)) { 162 (void)ext2_unmount(mp, 0, p); 163 bsd_free(mp, M_MOUNT); 164 return (error); 165 } 166#if defined(__FreeBSD__) 167 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 168#else 169 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 170#endif 171 mp->mnt_flag |= MNT_ROOTFS; 172 mp->mnt_vnodecovered = NULLVP; 173 ump = VFSTOUFS(mp); 174 fs = ump->um_e2fs; 175 bzero(fs->fs_fsmnt, sizeof(fs->fs_fsmnt)); 176 fs->fs_fsmnt[0] = '/'; 177 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 178 MNAMELEN); 179 (void) copystr(ROOTNAME, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 180 &size); 181 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 182 (void)ext2_statfs(mp, &mp->mnt_stat, p); 183 vfs_unlock(mp); 184 inittodr(fs->s_es->s_wtime); /* this helps to set the time */ 185 return (0); 186} 187#endif 188 189/* 190 * VFS Operations. 191 * 192 * mount system call 193 */ 194static int 195ext2_mount(mp, path, data, ndp, p) 196 register struct mount *mp; 197 char *path; 198 caddr_t data; /* this is actually a (struct ufs_args *) */ 199 struct nameidata *ndp; 200 struct proc *p; 201{ 202 struct vnode *devvp; 203 struct ufs_args args; 204 struct ufsmount *ump = 0; 205 register struct ext2_sb_info *fs; 206 u_int size; 207 int error, flags; 208 209 if (error = copyin(data, (caddr_t)&args, sizeof (struct ufs_args))) 210 return (error); 211 /* 212 * If updating, check whether changing from read-only to 213 * read/write; if there is no device name, that's all we do. 214 */ 215 if (mp->mnt_flag & MNT_UPDATE) { 216 ump = VFSTOUFS(mp); 217 fs = ump->um_e2fs; 218 error = 0; 219 if (fs->s_rd_only == 0 && (mp->mnt_flag & MNT_RDONLY)) { 220 flags = WRITECLOSE; 221 if (mp->mnt_flag & MNT_FORCE) 222 flags |= FORCECLOSE; 223 if (vfs_busy(mp, LK_NOWAIT, 0, p)) 224 return (EBUSY); 225 error = ext2_flushfiles(mp, flags, p); 226 vfs_unbusy(mp, p); 227 } 228 if (!error && (mp->mnt_flag & MNT_RELOAD)) 229 error = ext2_reload(mp, ndp->ni_cnd.cn_cred, p); 230 if (error) 231 return (error); 232 if (fs->s_rd_only && (mp->mnt_flag & MNT_WANTRDWR)) 233 fs->s_rd_only = 0; 234 if (fs->s_rd_only == 0) { 235 /* don't say it's clean */ 236 fs->s_es->s_state &= ~EXT2_VALID_FS; 237 ext2_sbupdate(ump, MNT_WAIT); 238 } 239 if (args.fspec == 0) { 240 /* 241 * Process export requests. 242 */ 243 return (vfs_export(mp, &ump->um_export, &args.export)); 244 } 245 } 246 /* 247 * Not an update, or updating the name: look up the name 248 * and verify that it refers to a sensible block device. 249 */ 250 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p); 251 if (error = namei(ndp)) 252 return (error); 253 devvp = ndp->ni_vp; 254 255 if (devvp->v_type != VBLK) { 256 vrele(devvp); 257 return (ENOTBLK); 258 } 259 if (major(devvp->v_rdev) >= nblkdev) { 260 vrele(devvp); 261 return (ENXIO); 262 } 263 if ((mp->mnt_flag & MNT_UPDATE) == 0) 264 error = ext2_mountfs(devvp, mp, p); 265 else { 266 if (devvp != ump->um_devvp) 267 error = EINVAL; /* needs translation */ 268 else 269 vrele(devvp); 270 } 271 if (error) { 272 vrele(devvp); 273 return (error); 274 } 275 ump = VFSTOUFS(mp); 276 fs = ump->um_e2fs; 277 (void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size); 278 bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size); 279 bcopy((caddr_t)fs->fs_fsmnt, (caddr_t)mp->mnt_stat.f_mntonname, 280 MNAMELEN); 281 (void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 282 &size); 283 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 284 (void)ext2_statfs(mp, &mp->mnt_stat, p); 285 return (0); 286} 287 288/* 289 * checks that the data in the descriptor blocks make sense 290 * this is taken from ext2/super.c 291 */ 292static int ext2_check_descriptors (struct ext2_sb_info * sb) 293{ 294 int i; 295 int desc_block = 0; 296 unsigned long block = sb->s_es->s_first_data_block; 297 struct ext2_group_desc * gdp = NULL; 298 299 /* ext2_debug ("Checking group descriptors"); */ 300 301 for (i = 0; i < sb->s_groups_count; i++) 302 { 303 /* examine next descriptor block */ 304 if ((i % EXT2_DESC_PER_BLOCK(sb)) == 0) 305 gdp = (struct ext2_group_desc *) 306 sb->s_group_desc[desc_block++]->b_data; 307 if (gdp->bg_block_bitmap < block || 308 gdp->bg_block_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 309 { 310 printf ("ext2_check_descriptors: " 311 "Block bitmap for group %d" 312 " not in group (block %lu)!", 313 i, (unsigned long) gdp->bg_block_bitmap); 314 return 0; 315 } 316 if (gdp->bg_inode_bitmap < block || 317 gdp->bg_inode_bitmap >= block + EXT2_BLOCKS_PER_GROUP(sb)) 318 { 319 printf ("ext2_check_descriptors: " 320 "Inode bitmap for group %d" 321 " not in group (block %lu)!", 322 i, (unsigned long) gdp->bg_inode_bitmap); 323 return 0; 324 } 325 if (gdp->bg_inode_table < block || 326 gdp->bg_inode_table + sb->s_itb_per_group >= 327 block + EXT2_BLOCKS_PER_GROUP(sb)) 328 { 329 printf ("ext2_check_descriptors: " 330 "Inode table for group %d" 331 " not in group (block %lu)!", 332 i, (unsigned long) gdp->bg_inode_table); 333 return 0; 334 } 335 block += EXT2_BLOCKS_PER_GROUP(sb); 336 gdp++; 337 } 338 return 1; 339} 340 341/* 342 * this computes the fields of the ext2_sb_info structure from the 343 * data in the ext2_super_block structure read in 344 */ 345static int compute_sb_data(devvp, es, fs) 346 struct vnode * devvp; 347 struct ext2_super_block * es; 348 struct ext2_sb_info * fs; 349{ 350 int db_count, error; 351 int i, j; 352 int logic_sb_block = 1; /* XXX for now */ 353 354#if 1 355#define V(v) 356#else 357#define V(v) printf(#v"= %d\n", fs->v); 358#endif 359 360 fs->s_blocksize = EXT2_MIN_BLOCK_SIZE << es->s_log_block_size; 361 V(s_blocksize) 362 fs->s_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->s_log_block_size; 363 V(s_bshift) 364 fs->s_fsbtodb = es->s_log_block_size + 1; 365 V(s_fsbtodb) 366 fs->s_qbmask = fs->s_blocksize - 1; 367 V(s_bmask) 368 fs->s_blocksize_bits = EXT2_BLOCK_SIZE_BITS(es); 369 V(s_blocksize_bits) 370 fs->s_frag_size = EXT2_MIN_FRAG_SIZE << es->s_log_frag_size; 371 V(s_frag_size) 372 if (fs->s_frag_size) 373 fs->s_frags_per_block = fs->s_blocksize / fs->s_frag_size; 374 V(s_frags_per_block) 375 fs->s_blocks_per_group = es->s_blocks_per_group; 376 V(s_blocks_per_group) 377 fs->s_frags_per_group = es->s_frags_per_group; 378 V(s_frags_per_group) 379 fs->s_inodes_per_group = es->s_inodes_per_group; 380 V(s_inodes_per_group) 381 fs->s_inodes_per_block = fs->s_blocksize / EXT2_INODE_SIZE; 382 V(s_inodes_per_block) 383 fs->s_itb_per_group = fs->s_inodes_per_group /fs->s_inodes_per_block; 384 V(s_itb_per_group) 385 fs->s_desc_per_block = fs->s_blocksize / sizeof (struct ext2_group_desc); 386 V(s_desc_per_block) 387 /* s_resuid / s_resgid ? */ 388 fs->s_groups_count = (es->s_blocks_count - 389 es->s_first_data_block + 390 EXT2_BLOCKS_PER_GROUP(fs) - 1) / 391 EXT2_BLOCKS_PER_GROUP(fs); 392 V(s_groups_count) 393 db_count = (fs->s_groups_count + EXT2_DESC_PER_BLOCK(fs) - 1) / 394 EXT2_DESC_PER_BLOCK(fs); 395 fs->s_db_per_group = db_count; 396 V(s_db_per_group) 397 398 fs->s_group_desc = bsd_malloc(db_count * sizeof (struct buf *), 399 M_UFSMNT, M_WAITOK); 400 401 /* adjust logic_sb_block */ 402 if(fs->s_blocksize > SBSIZE) 403 /* Godmar thinks: if the blocksize is greater than 1024, then 404 the superblock is logically part of block zero. 405 */ 406 logic_sb_block = 0; 407 408 for (i = 0; i < db_count; i++) { 409 error = bread(devvp , fsbtodb(fs, logic_sb_block + i + 1), 410 fs->s_blocksize, NOCRED, &fs->s_group_desc[i]); 411 if(error) { 412 for (j = 0; j < i; j++) 413 brelse(fs->s_group_desc[j]); 414 bsd_free(fs->s_group_desc, M_UFSMNT); 415 printf("EXT2-fs: unable to read group descriptors (%d)\n", error); 416 return EIO; 417 } 418 } 419 if(!ext2_check_descriptors(fs)) { 420 for (j = 0; j < db_count; j++) 421 brelse(fs->s_group_desc[j]); 422 bsd_free(fs->s_group_desc, M_UFSMNT); 423 printf("EXT2-fs: (ext2_check_descriptors failure) " 424 "unable to read group descriptors\n"); 425 return EIO; 426 } 427 428 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) { 429 fs->s_inode_bitmap_number[i] = 0; 430 fs->s_inode_bitmap[i] = NULL; 431 fs->s_block_bitmap_number[i] = 0; 432 fs->s_block_bitmap[i] = NULL; 433 } 434 fs->s_loaded_inode_bitmaps = 0; 435 fs->s_loaded_block_bitmaps = 0; 436 return 0; 437} 438 439/* 440 * Reload all incore data for a filesystem (used after running fsck on 441 * the root filesystem and finding things to fix). The filesystem must 442 * be mounted read-only. 443 * 444 * Things to do to update the mount: 445 * 1) invalidate all cached meta-data. 446 * 2) re-read superblock from disk. 447 * 3) re-read summary information from disk. 448 * 4) invalidate all inactive vnodes. 449 * 5) invalidate all cached file data. 450 * 6) re-read inode data for all active vnodes. 451 */ 452static int 453ext2_reload(mountp, cred, p) 454 register struct mount *mountp; 455 struct ucred *cred; 456 struct proc *p; 457{ 458 register struct vnode *vp, *nvp, *devvp; 459 struct inode *ip; 460 struct buf *bp; 461 struct ext2_super_block * es; 462 struct ext2_sb_info *fs; 463 int error; 464 465 if ((mountp->mnt_flag & MNT_RDONLY) == 0) 466 return (EINVAL); 467 /* 468 * Step 1: invalidate all cached meta-data. 469 */ 470 devvp = VFSTOUFS(mountp)->um_devvp; 471 if (vinvalbuf(devvp, 0, cred, p, 0, 0)) 472 panic("ext2_reload: dirty1"); 473 /* 474 * Step 2: re-read superblock from disk. 475 * constants have been adjusted for ext2 476 */ 477 if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) 478 return (error); 479 es = (struct ext2_super_block *)bp->b_data; 480 if (es->s_magic != EXT2_SUPER_MAGIC) { 481 if(es->s_magic == EXT2_PRE_02B_MAGIC) 482 printf("This filesystem bears the magic number of a pre " 483 "0.2b version of ext2. This is not supported by " 484 "Lites.\n"); 485 else 486 printf("Wrong magic number: %x (expected %x for ext2 fs\n", 487 es->s_magic, EXT2_SUPER_MAGIC); 488 brelse(bp); 489 return (EIO); /* XXX needs translation */ 490 } 491 fs = VFSTOUFS(mountp)->um_e2fs; 492 bcopy(bp->b_data, fs->s_es, sizeof(struct ext2_super_block)); 493 494 if(error = compute_sb_data(devvp, es, fs)) { 495 brelse(bp); 496 return error; 497 } 498#ifdef UNKLAR 499 if (fs->fs_sbsize < SBSIZE) 500 bp->b_flags |= B_INVAL; 501#endif 502 brelse(bp); 503 504loop: 505 for (vp = mountp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) { 506 nvp = vp->v_mntvnodes.le_next; 507 /* 508 * Step 4: invalidate all inactive vnodes. 509 */ 510 if (vp->v_usecount == 0) { 511 vgone(vp); 512 continue; 513 } 514 /* 515 * Step 5: invalidate all cached file data. 516 */ 517 if (vget(vp, LK_EXCLUSIVE, p)) 518 goto loop; 519 if (vinvalbuf(vp, 0, cred, p, 0, 0)) 520 panic("ext2_reload: dirty2"); 521 /* 522 * Step 6: re-read inode data for all active vnodes. 523 */ 524 ip = VTOI(vp); 525 if (error = 526 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 527 (int)fs->s_blocksize, NOCRED, &bp)) { 528 vput(vp); 529 return (error); 530 } 531 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + 532 EXT2_INODE_SIZE * ino_to_fsbo(fs, ip->i_number)), 533 &ip->i_din); 534 brelse(bp); 535 vput(vp); 536 if (vp->v_mount != mountp) 537 goto loop; 538 } 539 return (0); 540} 541 542/* 543 * Common code for mount and mountroot 544 */ 545static int 546ext2_mountfs(devvp, mp, p) 547 register struct vnode *devvp; 548 struct mount *mp; 549 struct proc *p; 550{ 551 register struct ufsmount *ump; 552 struct buf *bp; 553 register struct ext2_sb_info *fs; 554 struct ext2_super_block * es; 555 dev_t dev = devvp->v_rdev; 556 struct partinfo dpart; 557 int havepart = 0; 558 int error, i, size; 559 int ronly; 560#if !defined(__FreeBSD__) 561 extern struct vnode *rootvp; 562#endif 563 564 /* 565 * Disallow multiple mounts of the same device. 566 * Disallow mounting of a device that is currently in use 567 * (except for root, which might share swap device for miniroot). 568 * Flush out any old buffers remaining from a previous use. 569 */ 570 if (error = vfs_mountedon(devvp)) 571 return (error); 572 if (vcount(devvp) > 1 && devvp != rootvp) 573 return (EBUSY); 574 if (error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0)) 575 return (error); 576#ifdef READONLY 577/* turn on this to force it to be read-only */ 578 mp->mnt_flag |= MNT_RDONLY; 579#endif 580 581 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 582 if (error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p)) 583 return (error); 584 if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0) 585 size = DEV_BSIZE; 586 else { 587 havepart = 1; 588 size = dpart.disklab->d_secsize; 589 } 590 591 bp = NULL; 592 ump = NULL; 593 if (error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) 594 goto out; 595 es = (struct ext2_super_block *)bp->b_data; 596 if (es->s_magic != EXT2_SUPER_MAGIC) { 597 if(es->s_magic == EXT2_PRE_02B_MAGIC) 598 printf("This filesystem bears the magic number of a pre " 599 "0.2b version of ext2. This is not supported by " 600 "Lites.\n"); 601 else 602 printf("Wrong magic number: %x (expected %x for EXT2FS)\n", 603 es->s_magic, EXT2_SUPER_MAGIC); 604 error = EINVAL; /* XXX needs translation */ 605 goto out; 606 } 607 ump = bsd_malloc(sizeof *ump, M_UFSMNT, M_WAITOK); 608 bzero((caddr_t)ump, sizeof *ump); 609 /* I don't know whether this is the right strategy. Note that 610 we dynamically allocate both a ext2_sb_info and a ext2_super_block 611 while Linux keeps the super block in a locked buffer 612 */ 613 ump->um_e2fs = bsd_malloc(sizeof(struct ext2_sb_info), 614 M_UFSMNT, M_WAITOK); 615 ump->um_e2fs->s_es = bsd_malloc(sizeof(struct ext2_super_block), 616 M_UFSMNT, M_WAITOK); 617 bcopy(es, ump->um_e2fs->s_es, (u_int)sizeof(struct ext2_super_block)); 618 if(error = compute_sb_data(devvp, ump->um_e2fs->s_es, ump->um_e2fs)) { 619 brelse(bp); 620 return error; 621 } 622 brelse(bp); 623 bp = NULL; 624 fs = ump->um_e2fs; 625 fs->s_rd_only = ronly; /* ronly is set according to mnt_flags */ 626 if (!(fs->s_es->s_state & EXT2_VALID_FS)) { 627 printf("WARNING: %s was not properly dismounted\n", 628 fs->fs_fsmnt); 629 } 630 /* if the fs is not mounted read-only, make sure the super block is 631 always written back on a sync() 632 */ 633 if (ronly == 0) { 634 fs->s_dirt = 1; /* mark it modified */ 635 fs->s_es->s_state &= ~EXT2_VALID_FS; /* set fs invalid */ 636 } 637 mp->mnt_data = (qaddr_t)ump; 638 mp->mnt_stat.f_fsid.val[0] = (long)dev; 639 mp->mnt_stat.f_fsid.val[1] = MOUNT_EXT2FS; 640 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN; 641 mp->mnt_flag |= MNT_LOCAL; 642 ump->um_mountp = mp; 643 ump->um_dev = dev; 644 ump->um_devvp = devvp; 645 /* setting those two parameters allows us to use 646 ufs_bmap w/o changse ! 647 */ 648 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs); 649 ump->um_bptrtodb = fs->s_es->s_log_block_size + 1; 650 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs); 651 for (i = 0; i < MAXQUOTAS; i++) 652 ump->um_quotas[i] = NULLVP; 653 devvp->v_specflags |= SI_MOUNTEDON; 654 if (ronly == 0) 655 ext2_sbupdate(ump, MNT_WAIT); 656 return (0); 657out: 658 if (bp) 659 brelse(bp); 660 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p); 661 if (ump) { 662 bsd_free(ump->um_fs, M_UFSMNT); 663 bsd_free(ump, M_UFSMNT); 664 mp->mnt_data = (qaddr_t)0; 665 } 666 return (error); 667} 668 669/* 670 * unmount system call 671 */ 672static int 673ext2_unmount(mp, mntflags, p) 674 struct mount *mp; 675 int mntflags; 676 struct proc *p; 677{ 678 register struct ufsmount *ump; 679 register struct ext2_sb_info *fs; 680 int error, flags, ronly, i; 681 682 flags = 0; 683 if (mntflags & MNT_FORCE) { 684 if (mp->mnt_flag & MNT_ROOTFS) 685 return (EINVAL); 686 flags |= FORCECLOSE; 687 } 688 if (error = ext2_flushfiles(mp, flags, p)) 689 return (error); 690 ump = VFSTOUFS(mp); 691 fs = ump->um_e2fs; 692 ronly = fs->s_rd_only; 693 if (!ronly) { 694 fs->s_es->s_state |= EXT2_VALID_FS; /* was fs_clean = 1 */ 695 ext2_sbupdate(ump, MNT_WAIT); 696 } 697 /* release buffers containing group descriptors */ 698 for(i = 0; i < fs->s_db_per_group; i++) 699 brelse(fs->s_group_desc[i]); 700 /* release cached inode/block bitmaps */ 701 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 702 if (fs->s_inode_bitmap[i]) 703 brelse (fs->s_inode_bitmap[i]); 704 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 705 if (fs->s_block_bitmap[i]) 706 brelse (fs->s_block_bitmap[i]); 707 708 ump->um_devvp->v_specflags &= ~SI_MOUNTEDON; 709 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD|FWRITE, 710 NOCRED, p); 711 vrele(ump->um_devvp); 712 bsd_free(fs->s_es, M_UFSMNT); 713 bsd_free(fs, M_UFSMNT); 714 bsd_free(ump, M_UFSMNT); 715 mp->mnt_data = (qaddr_t)0; 716 mp->mnt_flag &= ~MNT_LOCAL; 717 return (error); 718} 719 720/* 721 * Flush out all the files in a filesystem. 722 */ 723static int 724ext2_flushfiles(mp, flags, p) 725 register struct mount *mp; 726 int flags; 727 struct proc *p; 728{ 729 register struct ufsmount *ump; 730 int error; 731#if QUOTA 732 int i; 733#endif 734 735 ump = VFSTOUFS(mp); 736#if QUOTA 737 if (mp->mnt_flag & MNT_QUOTA) { 738 if (error = vflush(mp, NULLVP, SKIPSYSTEM|flags)) 739 return (error); 740 for (i = 0; i < MAXQUOTAS; i++) { 741 if (ump->um_quotas[i] == NULLVP) 742 continue; 743 quotaoff(p, mp, i); 744 } 745 /* 746 * Here we fall through to vflush again to ensure 747 * that we have gotten rid of all the system vnodes. 748 */ 749 } 750#endif 751 error = vflush(mp, NULLVP, flags); 752 return (error); 753} 754 755/* 756 * Get file system statistics. 757 * taken from ext2/super.c ext2_statfs 758 */ 759static int 760ext2_statfs(mp, sbp, p) 761 struct mount *mp; 762 register struct statfs *sbp; 763 struct proc *p; 764{ 765 unsigned long overhead; 766 unsigned long overhead_per_group; 767 768 register struct ufsmount *ump; 769 register struct ext2_sb_info *fs; 770 register struct ext2_super_block *es; 771 772 ump = VFSTOUFS(mp); 773 fs = ump->um_e2fs; 774 es = fs->s_es; 775 776 if (es->s_magic != EXT2_SUPER_MAGIC) 777 panic("ext2_statfs - magic number spoiled"); 778 779 /* 780 * Compute the overhead (FS structures) 781 */ 782 overhead_per_group = 1 /* super block */ + 783 fs->s_db_per_group + 784 1 /* block bitmap */ + 785 1 /* inode bitmap */ + 786 fs->s_itb_per_group; 787 overhead = es->s_first_data_block + 788 fs->s_groups_count * overhead_per_group; 789 790 sbp->f_type = MOUNT_EXT2FS; 791 sbp->f_bsize = EXT2_FRAG_SIZE(fs); 792 sbp->f_iosize = EXT2_BLOCK_SIZE(fs); 793 sbp->f_blocks = es->s_blocks_count - overhead; 794 sbp->f_bfree = es->s_free_blocks_count; 795 sbp->f_bavail = sbp->f_bfree - es->s_r_blocks_count; 796 sbp->f_files = es->s_inodes_count; 797 sbp->f_ffree = es->s_free_inodes_count; 798 if (sbp != &mp->mnt_stat) { 799 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 800 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 801 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 802 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 803 } 804 return (0); 805} 806 807/* 808 * Go through the disk queues to initiate sandbagged IO; 809 * go through the inodes to write those that have been modified; 810 * initiate the writing of the super block if it has been modified. 811 * 812 * Note: we are always called with the filesystem marked `MPBUSY'. 813 */ 814static int 815ext2_sync(mp, waitfor, cred, p) 816 struct mount *mp; 817 int waitfor; 818 struct ucred *cred; 819 struct proc *p; 820{ 821 register struct vnode *vp; 822 register struct inode *ip; 823 register struct ufsmount *ump = VFSTOUFS(mp); 824 register struct ext2_sb_info *fs; 825 int error, allerror = 0; 826 827 fs = ump->um_e2fs; 828 /* 829 * Write back modified superblock. 830 * Consistency check that the superblock 831 * is still in the buffer cache. 832 */ 833 if (fs->s_dirt) { 834#if !defined(__FreeBSD__) 835 struct timeval time; 836#endif 837 838 if (fs->s_rd_only != 0) { /* XXX */ 839 printf("fs = %s\n", fs->fs_fsmnt); 840 panic("update: rofs mod"); 841 } 842 fs->s_dirt = 0; 843#if !defined(__FreeBSD__) 844 get_time(&time); 845#endif 846 fs->s_es->s_wtime = time.tv_sec; 847 allerror = ext2_sbupdate(ump, waitfor); 848 } 849 /* 850 * Write back each (modified) inode. 851 */ 852loop: 853 for (vp = mp->mnt_vnodelist.lh_first; 854 vp != NULL; 855 vp = vp->v_mntvnodes.le_next) { 856 /* 857 * If the vnode that we are about to sync is no longer 858 * associated with this mount point, start over. 859 */ 860 if (vp->v_mount != mp) 861 goto loop; 862 if (VOP_ISLOCKED(vp)) 863 continue; 864 ip = VTOI(vp); 865 if ((ip->i_flag & 866 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 867 vp->v_dirtyblkhd.lh_first == NULL) 868 continue; 869 if (vget(vp, LK_EXCLUSIVE, p)) 870 goto loop; 871 if (error = VOP_FSYNC(vp, cred, waitfor, p)) 872 allerror = error; 873 vput(vp); 874 } 875 /* 876 * Force stale file system control information to be flushed. 877 */ 878 if (error = VOP_FSYNC(ump->um_devvp, cred, waitfor, p)) 879 allerror = error; 880#if QUOTA 881 qsync(mp); 882#endif 883 return (allerror); 884} 885 886/* 887 * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it 888 * in from disk. If it is in core, wait for the lock bit to clear, then 889 * return the inode locked. Detection and handling of mount points must be 890 * done by the calling routine. 891 */ 892static int 893ext2_vget(mp, ino, vpp) 894 struct mount *mp; 895 ino_t ino; 896 struct vnode **vpp; 897{ 898 register struct ext2_sb_info *fs; 899 register struct inode *ip; 900 struct ufsmount *ump; 901 struct buf *bp; 902 struct vnode *vp; 903 dev_t dev; 904 int i, type, error; 905 int used_blocks; 906 907 ump = VFSTOUFS(mp); 908 dev = ump->um_dev; 909restart: 910 if ((*vpp = ufs_ihashget(dev, ino)) != NULL) 911 return (0); 912 913#ifdef __FreeBSD__ 914 /* 915 * Lock out the creation of new entries in the FFS hash table in 916 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate 917 * may occur! 918 */ 919 if (ext2fs_inode_hash_lock) { 920 while (ext2fs_inode_hash_lock) { 921 ext2fs_inode_hash_lock = -1; 922 tsleep(&ext2fs_inode_hash_lock, PVM, "ffsvgt", 0); 923 } 924 goto restart; 925 } 926 ext2fs_inode_hash_lock = 1; 927#endif 928 929 /* Allocate a new vnode/inode. */ 930 if (error = getnewvnode(VT_UFS, mp, ext2_vnodeop_p, &vp)) { 931 *vpp = NULL; 932 return (error); 933 } 934 /* I don't really know what this 'type' does. I suppose it's some kind 935 * of memory accounting. Let's just book this memory on FFS's account 936 * If I'm not mistaken, this stuff isn't implemented anyway in Lites 937 */ 938 type = ump->um_devvp->v_tag == VT_MFS ? M_MFSNODE : M_FFSNODE; /* XXX */ 939 MALLOC(ip, struct inode *, sizeof(struct inode), type, M_WAITOK); 940#ifndef __FreeBSD__ 941 insmntque(vp, mp); 942#endif 943 bzero((caddr_t)ip, sizeof(struct inode)); 944 vp->v_data = ip; 945 ip->i_vnode = vp; 946 ip->i_e2fs = fs = ump->um_e2fs; 947 ip->i_dev = dev; 948 ip->i_number = ino; 949#if QUOTA 950 for (i = 0; i < MAXQUOTAS; i++) 951 ip->i_dquot[i] = NODQUOT; 952#endif 953 /* 954 * Put it onto its hash chain and lock it so that other requests for 955 * this inode will block if they arrive while we are sleeping waiting 956 * for old data structures to be purged or for the contents of the 957 * disk portion of this inode to be read. 958 */ 959 ufs_ihashins(ip); 960 961#ifdef __FreeBSD__ 962 if (ext2fs_inode_hash_lock < 0) 963 wakeup(&ext2fs_inode_hash_lock); 964 ext2fs_inode_hash_lock = 0; 965#endif 966 967 /* Read in the disk contents for the inode, copy into the inode. */ 968 /* Read in the disk contents for the inode, copy into the inode. */ 969#if 0 970printf("ext2_vget(%d) dbn= %d ", ino, fsbtodb(fs, ino_to_fsba(fs, ino))); 971#endif 972 if (error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 973 (int)fs->s_blocksize, NOCRED, &bp)) { 974 /* 975 * The inode does not contain anything useful, so it would 976 * be misleading to leave it on its hash chain. With mode 977 * still zero, it will be unlinked and returned to the free 978 * list by vput(). 979 */ 980 vput(vp); 981 brelse(bp); 982 *vpp = NULL; 983 return (error); 984 } 985 /* convert ext2 inode to dinode */ 986 ext2_ei2di((struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE * 987 ino_to_fsbo(fs, ino)), &ip->i_din); 988 ip->i_block_group = ino_to_cg(fs, ino); 989 ip->i_next_alloc_block = 0; 990 ip->i_next_alloc_goal = 0; 991 ip->i_prealloc_count = 0; 992 ip->i_prealloc_block = 0; 993 /* now we want to make sure that block pointers for unused 994 blocks are zeroed out - ext2_balloc depends on this 995 although for regular files and directories only 996 */ 997 if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) { 998 used_blocks = (ip->i_size+fs->s_blocksize-1) / fs->s_blocksize; 999 for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++) 1000 ip->i_db[i] = 0; 1001 } 1002/* 1003 ext2_print_inode(ip); 1004*/ 1005 brelse(bp); 1006 1007 /* 1008 * Initialize the vnode from the inode, check for aliases. 1009 * Note that the underlying vnode may have changed. 1010 */ 1011 if (error = ufs_vinit(mp, ext2_specop_p, ext2_fifoop_p, &vp)) { 1012 vput(vp); 1013 *vpp = NULL; 1014 return (error); 1015 } 1016 /* 1017 * Finish inode initialization now that aliasing has been resolved. 1018 */ 1019 ip->i_devvp = ump->um_devvp; 1020 VREF(ip->i_devvp); 1021 /* 1022 * Set up a generation number for this inode if it does not 1023 * already have one. This should only happen on old filesystems. 1024 */ 1025 if (ip->i_gen == 0) { 1026#if !defined(__FreeBSD__) 1027 struct timeval time; 1028 get_time(&time); 1029#endif 1030 if (++nextgennumber < (u_long)time.tv_sec) 1031 nextgennumber = time.tv_sec; 1032 ip->i_gen = nextgennumber; 1033 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) 1034 ip->i_flag |= IN_MODIFIED; 1035 } 1036 *vpp = vp; 1037 return (0); 1038} 1039 1040/* 1041 * File handle to vnode 1042 * 1043 * Have to be really careful about stale file handles: 1044 * - check that the inode number is valid 1045 * - call ext2_vget() to get the locked inode 1046 * - check for an unallocated inode (i_mode == 0) 1047 * - check that the given client host has export rights and return 1048 * those rights via. exflagsp and credanonp 1049 */ 1050static int 1051ext2_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) 1052 register struct mount *mp; 1053 struct fid *fhp; 1054 struct mbuf *nam; 1055 struct vnode **vpp; 1056 int *exflagsp; 1057 struct ucred **credanonp; 1058{ 1059 register struct ufid *ufhp; 1060 struct ext2_sb_info *fs; 1061 1062 ufhp = (struct ufid *)fhp; 1063 fs = VFSTOUFS(mp)->um_e2fs; 1064 if (ufhp->ufid_ino < ROOTINO || 1065 ufhp->ufid_ino >= fs->s_groups_count * fs->s_es->s_inodes_per_group) 1066 return (ESTALE); 1067 return (ufs_check_export(mp, ufhp, nam, vpp, exflagsp, credanonp)); 1068} 1069 1070/* 1071 * Vnode pointer to File handle 1072 */ 1073/* ARGSUSED */ 1074static int 1075ext2_vptofh(vp, fhp) 1076 struct vnode *vp; 1077 struct fid *fhp; 1078{ 1079 register struct inode *ip; 1080 register struct ufid *ufhp; 1081 1082 ip = VTOI(vp); 1083 ufhp = (struct ufid *)fhp; 1084 ufhp->ufid_len = sizeof(struct ufid); 1085 ufhp->ufid_ino = ip->i_number; 1086 ufhp->ufid_gen = ip->i_gen; 1087 return (0); 1088} 1089 1090/* 1091 * Write a superblock and associated information back to disk. 1092 */ 1093static int 1094ext2_sbupdate(mp, waitfor) 1095 struct ufsmount *mp; 1096 int waitfor; 1097{ 1098 register struct ext2_sb_info *fs = mp->um_e2fs; 1099 register struct ext2_super_block *es = fs->s_es; 1100 register struct buf *bp; 1101 int i, error = 0; 1102/* 1103printf("\nupdating superblock, waitfor=%s\n", waitfor == MNT_WAIT ? "yes":"no"); 1104*/ 1105 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0); 1106 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2_super_block)); 1107 if (waitfor == MNT_WAIT) 1108 error = bwrite(bp); 1109 else 1110 bawrite(bp); 1111 1112 /* write group descriptors back on disk */ 1113 for(i = 0; i < fs->s_db_per_group; i++) 1114 /* Godmar thinks: we must avoid using any of the b*write 1115 * functions here: we want to keep the buffer locked 1116 * so we use my 'housemade' write routine: 1117 */ 1118 error |= ll_w_block(fs->s_group_desc[i], waitfor == MNT_WAIT); 1119 1120 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 1121 if (fs->s_inode_bitmap[i]) 1122 ll_w_block (fs->s_inode_bitmap[i], 1); 1123 for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) 1124 if (fs->s_block_bitmap[i]) 1125 ll_w_block (fs->s_block_bitmap[i], 1); 1126 1127 return (error); 1128} 1129