ffs_vfsops.c revision 128658
1/* 2 * Copyright (c) 1989, 1991, 1993, 1994 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ffs_vfsops.c 8.31 (Berkeley) 5/20/95 30 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: head/sys/ufs/ffs/ffs_vfsops.c 128658 2004-04-26 15:13:46Z bmilekic $"); 34 35#include "opt_mac.h" 36#include "opt_quota.h" 37#include "opt_ufs.h" 38 39#include <sys/param.h> 40#include <sys/systm.h> 41#include <sys/namei.h> 42#include <sys/proc.h> 43#include <sys/kernel.h> 44#include <sys/mac.h> 45#include <sys/vnode.h> 46#include <sys/mount.h> 47#include <sys/bio.h> 48#include <sys/buf.h> 49#include <sys/conf.h> 50#include <sys/fcntl.h> 51#include <sys/disk.h> 52#include <sys/malloc.h> 53#include <sys/mutex.h> 54 55#include <ufs/ufs/extattr.h> 56#include <ufs/ufs/quota.h> 57#include <ufs/ufs/ufsmount.h> 58#include <ufs/ufs/inode.h> 59#include <ufs/ufs/ufs_extern.h> 60 61#include <ufs/ffs/fs.h> 62#include <ufs/ffs/ffs_extern.h> 63 64#include <vm/vm.h> 65#include <vm/uma.h> 66#include <vm/vm_page.h> 67 68uma_zone_t uma_inode, uma_ufs1, uma_ufs2; 69 70static int ffs_sbupdate(struct ufsmount *, int); 71 int ffs_reload(struct mount *,struct ucred *,struct thread *); 72static int ffs_mountfs(struct vnode *, struct mount *, struct thread *); 73static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, 74 ufs2_daddr_t); 75static void ffs_oldfscompat_write(struct fs *, struct ufsmount *); 76static void ffs_ifree(struct ufsmount *ump, struct inode *ip); 77static vfs_init_t ffs_init; 78static vfs_uninit_t ffs_uninit; 79static vfs_extattrctl_t ffs_extattrctl; 80 81static struct vfsops ufs_vfsops = { 82 .vfs_extattrctl = ffs_extattrctl, 83 .vfs_fhtovp = ffs_fhtovp, 84 .vfs_init = ffs_init, 85 .vfs_mount = ffs_mount, 86 .vfs_quotactl = ufs_quotactl, 87 .vfs_root = ufs_root, 88 .vfs_start = ufs_start, 89 .vfs_statfs = ffs_statfs, 90 .vfs_sync = ffs_sync, 91 .vfs_uninit = ffs_uninit, 92 .vfs_unmount = ffs_unmount, 93 .vfs_vget = ffs_vget, 94 .vfs_vptofh = ffs_vptofh, 95}; 96 97VFS_SET(ufs_vfsops, ufs, 0); 98 99/* 100 * ffs_mount 101 * 102 * Called when mounting local physical media 103 * 104 * PARAMETERS: 105 * mountroot 106 * mp mount point structure 107 * path NULL (flag for root mount!!!) 108 * data <unused> 109 * ndp <unused> 110 * p process (user credentials check [statfs]) 111 * 112 * mount 113 * mp mount point structure 114 * path path to mount point 115 * data pointer to argument struct in user space 116 * ndp mount point namei() return (used for 117 * credentials on reload), reused to look 118 * up block device. 119 * p process (user credentials check) 120 * 121 * RETURNS: 0 Success 122 * !0 error number (errno.h) 123 * 124 * LOCK STATE: 125 * 126 * ENTRY 127 * mount point is locked 128 * EXIT 129 * mount point is locked 130 * 131 * NOTES: 132 * A NULL path can be used for a flag since the mount 133 * system call will fail with EFAULT in copyinstr in 134 * namei() if it is a genuine NULL from the user. 135 */ 136int 137ffs_mount(mp, path, data, ndp, td) 138 struct mount *mp; /* mount struct pointer*/ 139 char *path; /* path to mount point*/ 140 caddr_t data; /* arguments to FS specific mount*/ 141 struct nameidata *ndp; /* mount point credentials*/ 142 struct thread *td; /* process requesting mount*/ 143{ 144 size_t size; 145 struct vnode *devvp; 146 struct ufs_args args; 147 struct ufsmount *ump = 0; 148 struct fs *fs; 149 int error, flags; 150 mode_t accessmode; 151 152 if (uma_inode == NULL) { 153 uma_inode = uma_zcreate("FFS inode", 154 sizeof(struct inode), NULL, NULL, NULL, NULL, 155 UMA_ALIGN_PTR, 0); 156 uma_ufs1 = uma_zcreate("FFS1 dinode", 157 sizeof(struct ufs1_dinode), NULL, NULL, NULL, NULL, 158 UMA_ALIGN_PTR, 0); 159 uma_ufs2 = uma_zcreate("FFS2 dinode", 160 sizeof(struct ufs2_dinode), NULL, NULL, NULL, NULL, 161 UMA_ALIGN_PTR, 0); 162 } 163 /* 164 * Use NULL path to indicate we are mounting the root filesystem. 165 */ 166 if (path == NULL) { 167 if ((error = bdevvp(rootdev, &rootvp))) { 168 printf("ffs_mountroot: can't find rootvp\n"); 169 return (error); 170 } 171 172 if ((error = ffs_mountfs(rootvp, mp, td)) != 0) 173 return (error); 174 (void)VFS_STATFS(mp, &mp->mnt_stat, td); 175 return (0); 176 } 177 178 /* 179 * Mounting non-root filesystem or updating a filesystem 180 */ 181 if ((error = copyin(data, (caddr_t)&args, sizeof(struct ufs_args)))!= 0) 182 return (error); 183 184 /* 185 * If updating, check whether changing from read-only to 186 * read/write; if there is no device name, that's all we do. 187 */ 188 if (mp->mnt_flag & MNT_UPDATE) { 189 ump = VFSTOUFS(mp); 190 fs = ump->um_fs; 191 devvp = ump->um_devvp; 192 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 193 if ((error = vn_start_write(NULL, &mp, V_WAIT)) != 0) 194 return (error); 195 /* 196 * Flush any dirty data. 197 */ 198 if ((error = VFS_SYNC(mp, MNT_WAIT, 199 td->td_ucred, td)) != 0) { 200 vn_finished_write(mp); 201 return (error); 202 } 203 /* 204 * Check for and optionally get rid of files open 205 * for writing. 206 */ 207 flags = WRITECLOSE; 208 if (mp->mnt_flag & MNT_FORCE) 209 flags |= FORCECLOSE; 210 if (mp->mnt_flag & MNT_SOFTDEP) { 211 error = softdep_flushfiles(mp, flags, td); 212 } else { 213 error = ffs_flushfiles(mp, flags, td); 214 } 215 if (error) { 216 vn_finished_write(mp); 217 return (error); 218 } 219 if (fs->fs_pendingblocks != 0 || 220 fs->fs_pendinginodes != 0) { 221 printf("%s: %s: blocks %jd files %d\n", 222 fs->fs_fsmnt, "update error", 223 (intmax_t)fs->fs_pendingblocks, 224 fs->fs_pendinginodes); 225 fs->fs_pendingblocks = 0; 226 fs->fs_pendinginodes = 0; 227 } 228 fs->fs_ronly = 1; 229 if ((fs->fs_flags & (FS_UNCLEAN | FS_NEEDSFSCK)) == 0) 230 fs->fs_clean = 1; 231 if ((error = ffs_sbupdate(ump, MNT_WAIT)) != 0) { 232 fs->fs_ronly = 0; 233 fs->fs_clean = 0; 234 vn_finished_write(mp); 235 return (error); 236 } 237 vn_finished_write(mp); 238 } 239 if ((mp->mnt_flag & MNT_RELOAD) && 240 (error = ffs_reload(mp, ndp->ni_cnd.cn_cred, td)) != 0) 241 return (error); 242 if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { 243 /* 244 * If upgrade to read-write by non-root, then verify 245 * that user has necessary permissions on the device. 246 */ 247 if (suser(td)) { 248 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 249 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE, 250 td->td_ucred, td)) != 0) { 251 VOP_UNLOCK(devvp, 0, td); 252 return (error); 253 } 254 VOP_UNLOCK(devvp, 0, td); 255 } 256 fs->fs_flags &= ~FS_UNCLEAN; 257 if (fs->fs_clean == 0) { 258 fs->fs_flags |= FS_UNCLEAN; 259 if ((mp->mnt_flag & MNT_FORCE) || 260 ((fs->fs_flags & FS_NEEDSFSCK) == 0 && 261 (fs->fs_flags & FS_DOSOFTDEP))) { 262 printf("WARNING: %s was not %s\n", 263 fs->fs_fsmnt, "properly dismounted"); 264 } else { 265 printf( 266"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", 267 fs->fs_fsmnt); 268 return (EPERM); 269 } 270 } 271 if ((error = vn_start_write(NULL, &mp, V_WAIT)) != 0) 272 return (error); 273 fs->fs_ronly = 0; 274 fs->fs_clean = 0; 275 if ((error = ffs_sbupdate(ump, MNT_WAIT)) != 0) { 276 vn_finished_write(mp); 277 return (error); 278 } 279 /* check to see if we need to start softdep */ 280 if ((fs->fs_flags & FS_DOSOFTDEP) && 281 (error = softdep_mount(devvp, mp, fs, td->td_ucred))){ 282 vn_finished_write(mp); 283 return (error); 284 } 285 if (fs->fs_snapinum[0] != 0) 286 ffs_snapshot_mount(mp); 287 vn_finished_write(mp); 288 } 289 /* 290 * Soft updates is incompatible with "async", 291 * so if we are doing softupdates stop the user 292 * from setting the async flag in an update. 293 * Softdep_mount() clears it in an initial mount 294 * or ro->rw remount. 295 */ 296 if (mp->mnt_flag & MNT_SOFTDEP) 297 mp->mnt_flag &= ~MNT_ASYNC; 298 /* 299 * If not updating name, process export requests. 300 */ 301 if (args.fspec == 0) 302 return (vfs_export(mp, &args.export)); 303 /* 304 * If this is a snapshot request, take the snapshot. 305 */ 306 if (mp->mnt_flag & MNT_SNAPSHOT) 307 return (ffs_snapshot(mp, args.fspec)); 308 } 309 310 /* 311 * Not an update, or updating the name: look up the name 312 * and verify that it refers to a sensible disk device. 313 */ 314 NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, td); 315 if ((error = namei(ndp)) != 0) 316 return (error); 317 NDFREE(ndp, NDF_ONLY_PNBUF); 318 devvp = ndp->ni_vp; 319 if (!vn_isdisk(devvp, &error)) { 320 vrele(devvp); 321 return (error); 322 } 323 324 /* 325 * If mount by non-root, then verify that user has necessary 326 * permissions on the device. 327 */ 328 if (suser(td)) { 329 accessmode = VREAD; 330 if ((mp->mnt_flag & MNT_RDONLY) == 0) 331 accessmode |= VWRITE; 332 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 333 if ((error = VOP_ACCESS(devvp, accessmode, td->td_ucred, td))!= 0){ 334 vput(devvp); 335 return (error); 336 } 337 VOP_UNLOCK(devvp, 0, td); 338 } 339 340 if (mp->mnt_flag & MNT_UPDATE) { 341 /* 342 * Update only 343 * 344 * If it's not the same vnode, or at least the same device 345 * then it's not correct. 346 */ 347 348 if (devvp != ump->um_devvp && 349 devvp->v_rdev != ump->um_devvp->v_rdev) 350 error = EINVAL; /* needs translation */ 351 vrele(devvp); 352 if (error) 353 return (error); 354 } else { 355 /* 356 * New mount 357 * 358 * We need the name for the mount point (also used for 359 * "last mounted on") copied in. If an error occurs, 360 * the mount point is discarded by the upper level code. 361 * Note that vfs_mount() populates f_mntonname for us. 362 */ 363 if ((error = ffs_mountfs(devvp, mp, td)) != 0) { 364 vrele(devvp); 365 return (error); 366 } 367 } 368 /* 369 * Save "mounted from" device name info for mount point (NULL pad). 370 */ 371 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size); 372 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 373 /* 374 * Initialize filesystem stat information in mount struct. 375 */ 376 (void)VFS_STATFS(mp, &mp->mnt_stat, td); 377 return (0); 378} 379 380/* 381 * Reload all incore data for a filesystem (used after running fsck on 382 * the root filesystem and finding things to fix). The filesystem must 383 * be mounted read-only. 384 * 385 * Things to do to update the mount: 386 * 1) invalidate all cached meta-data. 387 * 2) re-read superblock from disk. 388 * 3) re-read summary information from disk. 389 * 4) invalidate all inactive vnodes. 390 * 5) invalidate all cached file data. 391 * 6) re-read inode data for all active vnodes. 392 */ 393int 394ffs_reload(mp, cred, td) 395 struct mount *mp; 396 struct ucred *cred; 397 struct thread *td; 398{ 399 struct vnode *vp, *nvp, *devvp; 400 struct inode *ip; 401 void *space; 402 struct buf *bp; 403 struct fs *fs, *newfs; 404 ufs2_daddr_t sblockloc; 405 int i, blks, size, error; 406 int32_t *lp; 407 408 if ((mp->mnt_flag & MNT_RDONLY) == 0) 409 return (EINVAL); 410 /* 411 * Step 1: invalidate all cached meta-data. 412 */ 413 devvp = VFSTOUFS(mp)->um_devvp; 414 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 415 if (vinvalbuf(devvp, 0, cred, td, 0, 0) != 0) 416 panic("ffs_reload: dirty1"); 417 /* 418 * Only VMIO the backing device if the backing device is a real 419 * disk device. See ffs_mountfs() for more details. 420 */ 421 if (vn_isdisk(devvp, NULL)) 422 vfs_object_create(devvp, td, td->td_ucred); 423 VOP_UNLOCK(devvp, 0, td); 424 425 /* 426 * Step 2: re-read superblock from disk. 427 */ 428 fs = VFSTOUFS(mp)->um_fs; 429 if ((error = bread(devvp, btodb(fs->fs_sblockloc), fs->fs_sbsize, 430 NOCRED, &bp)) != 0) 431 return (error); 432 newfs = (struct fs *)bp->b_data; 433 if ((newfs->fs_magic != FS_UFS1_MAGIC && 434 newfs->fs_magic != FS_UFS2_MAGIC) || 435 newfs->fs_bsize > MAXBSIZE || 436 newfs->fs_bsize < sizeof(struct fs)) { 437 brelse(bp); 438 return (EIO); /* XXX needs translation */ 439 } 440 /* 441 * Copy pointer fields back into superblock before copying in XXX 442 * new superblock. These should really be in the ufsmount. XXX 443 * Note that important parameters (eg fs_ncg) are unchanged. 444 */ 445 newfs->fs_csp = fs->fs_csp; 446 newfs->fs_maxcluster = fs->fs_maxcluster; 447 newfs->fs_contigdirs = fs->fs_contigdirs; 448 newfs->fs_active = fs->fs_active; 449 /* The file system is still read-only. */ 450 newfs->fs_ronly = 1; 451 sblockloc = fs->fs_sblockloc; 452 bcopy(newfs, fs, (u_int)fs->fs_sbsize); 453 brelse(bp); 454 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 455 ffs_oldfscompat_read(fs, VFSTOUFS(mp), sblockloc); 456 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 457 printf("%s: reload pending error: blocks %jd files %d\n", 458 fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, 459 fs->fs_pendinginodes); 460 fs->fs_pendingblocks = 0; 461 fs->fs_pendinginodes = 0; 462 } 463 464 /* 465 * Step 3: re-read summary information from disk. 466 */ 467 blks = howmany(fs->fs_cssize, fs->fs_fsize); 468 space = fs->fs_csp; 469 for (i = 0; i < blks; i += fs->fs_frag) { 470 size = fs->fs_bsize; 471 if (i + fs->fs_frag > blks) 472 size = (blks - i) * fs->fs_fsize; 473 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 474 NOCRED, &bp); 475 if (error) 476 return (error); 477 bcopy(bp->b_data, space, (u_int)size); 478 space = (char *)space + size; 479 brelse(bp); 480 } 481 /* 482 * We no longer know anything about clusters per cylinder group. 483 */ 484 if (fs->fs_contigsumsize > 0) { 485 lp = fs->fs_maxcluster; 486 for (i = 0; i < fs->fs_ncg; i++) 487 *lp++ = fs->fs_contigsumsize; 488 } 489 490loop: 491 MNT_ILOCK(mp); 492 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) { 493 if (vp->v_mount != mp) { 494 MNT_IUNLOCK(mp); 495 goto loop; 496 } 497 nvp = TAILQ_NEXT(vp, v_nmntvnodes); 498 VI_LOCK(vp); 499 if (vp->v_iflag & VI_XLOCK) { 500 VI_UNLOCK(vp); 501 continue; 502 } 503 MNT_IUNLOCK(mp); 504 /* 505 * Step 4: invalidate all inactive vnodes. 506 */ 507 if (vp->v_usecount == 0) { 508 vgonel(vp, td); 509 goto loop; 510 } 511 /* 512 * Step 5: invalidate all cached file data. 513 */ 514 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) { 515 goto loop; 516 } 517 if (vinvalbuf(vp, 0, cred, td, 0, 0)) 518 panic("ffs_reload: dirty2"); 519 /* 520 * Step 6: re-read inode data for all active vnodes. 521 */ 522 ip = VTOI(vp); 523 error = 524 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 525 (int)fs->fs_bsize, NOCRED, &bp); 526 if (error) { 527 VOP_UNLOCK(vp, 0, td); 528 vrele(vp); 529 return (error); 530 } 531 ffs_load_inode(bp, ip, fs, ip->i_number); 532 ip->i_effnlink = ip->i_nlink; 533 brelse(bp); 534 VOP_UNLOCK(vp, 0, td); 535 vrele(vp); 536 MNT_ILOCK(mp); 537 } 538 MNT_IUNLOCK(mp); 539 return (0); 540} 541 542/* 543 * Possible superblock locations ordered from most to least likely. 544 */ 545static int sblock_try[] = SBLOCKSEARCH; 546 547/* 548 * Common code for mount and mountroot 549 */ 550static int 551ffs_mountfs(devvp, mp, td) 552 struct vnode *devvp; 553 struct mount *mp; 554 struct thread *td; 555{ 556 struct ufsmount *ump; 557 struct buf *bp; 558 struct fs *fs; 559 dev_t dev; 560 void *space; 561 ufs2_daddr_t sblockloc; 562 int error, i, blks, size, ronly; 563 int32_t *lp; 564 struct ucred *cred; 565 size_t strsize; 566 567 dev = devvp->v_rdev; 568 cred = td ? td->td_ucred : NOCRED; 569 /* 570 * Disallow multiple mounts of the same device. 571 * Disallow mounting of a device that is currently in use 572 * (except for root, which might share swap device for miniroot). 573 * Flush out any old buffers remaining from a previous use. 574 */ 575 error = vfs_mountedon(devvp); 576 if (error) 577 return (error); 578 if (vcount(devvp) > 1 && devvp != rootvp) 579 return (EBUSY); 580 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 581 error = vinvalbuf(devvp, V_SAVE, cred, td, 0, 0); 582 if (error) { 583 VOP_UNLOCK(devvp, 0, td); 584 return (error); 585 } 586 587 /* 588 * Only VMIO the backing device if the backing device is a real 589 * disk device. 590 * Note that it is optional that the backing device be VMIOed. This 591 * increases the opportunity for metadata caching. 592 */ 593 if (vn_isdisk(devvp, NULL)) 594 vfs_object_create(devvp, td, cred); 595 596 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 597 /* 598 * XXX: open the device with read and write access even if only 599 * read access is needed now. Write access is needed if the 600 * filesystem is ever mounted read/write, and we don't change the 601 * access mode for remounts. 602 */ 603#ifdef notyet 604 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD | FWRITE, FSCRED, td, -1); 605#else 606 error = VOP_OPEN(devvp, FREAD | FWRITE, FSCRED, td, -1); 607#endif 608 VOP_UNLOCK(devvp, 0, td); 609 if (error) 610 return (error); 611 if (devvp->v_rdev->si_iosize_max != 0) 612 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max; 613 if (mp->mnt_iosize_max > MAXPHYS) 614 mp->mnt_iosize_max = MAXPHYS; 615 616 bp = NULL; 617 ump = NULL; 618 fs = NULL; 619 sblockloc = 0; 620 /* 621 * Try reading the superblock in each of its possible locations. 622 */ 623 for (i = 0; sblock_try[i] != -1; i++) { 624 if ((error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, 625 cred, &bp)) != 0) 626 goto out; 627 fs = (struct fs *)bp->b_data; 628 sblockloc = sblock_try[i]; 629 if ((fs->fs_magic == FS_UFS1_MAGIC || 630 (fs->fs_magic == FS_UFS2_MAGIC && 631 (fs->fs_sblockloc == sblockloc || 632 (fs->fs_old_flags & FS_FLAGS_UPDATED) == 0))) && 633 fs->fs_bsize <= MAXBSIZE && 634 fs->fs_bsize >= sizeof(struct fs)) 635 break; 636 brelse(bp); 637 bp = NULL; 638 } 639 if (sblock_try[i] == -1) { 640 error = EINVAL; /* XXX needs translation */ 641 goto out; 642 } 643 fs->fs_fmod = 0; 644 fs->fs_flags &= ~FS_INDEXDIRS; /* no support for directory indicies */ 645 fs->fs_flags &= ~FS_UNCLEAN; 646 if (fs->fs_clean == 0) { 647 fs->fs_flags |= FS_UNCLEAN; 648 if (ronly || (mp->mnt_flag & MNT_FORCE) || 649 ((fs->fs_flags & FS_NEEDSFSCK) == 0 && 650 (fs->fs_flags & FS_DOSOFTDEP))) { 651 printf( 652"WARNING: %s was not properly dismounted\n", 653 fs->fs_fsmnt); 654 } else { 655 printf( 656"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", 657 fs->fs_fsmnt); 658 error = EPERM; 659 goto out; 660 } 661 if ((fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) && 662 (mp->mnt_flag & MNT_FORCE)) { 663 printf("%s: lost blocks %jd files %d\n", fs->fs_fsmnt, 664 (intmax_t)fs->fs_pendingblocks, 665 fs->fs_pendinginodes); 666 fs->fs_pendingblocks = 0; 667 fs->fs_pendinginodes = 0; 668 } 669 } 670 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 671 printf("%s: mount pending error: blocks %jd files %d\n", 672 fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, 673 fs->fs_pendinginodes); 674 fs->fs_pendingblocks = 0; 675 fs->fs_pendinginodes = 0; 676 } 677 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK | M_ZERO); 678 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT, 679 M_WAITOK); 680 if (fs->fs_magic == FS_UFS1_MAGIC) { 681 ump->um_fstype = UFS1; 682 ump->um_balloc = ffs_balloc_ufs1; 683 } else { 684 ump->um_fstype = UFS2; 685 ump->um_balloc = ffs_balloc_ufs2; 686 } 687 ump->um_blkatoff = ffs_blkatoff; 688 ump->um_truncate = ffs_truncate; 689 ump->um_update = ffs_update; 690 ump->um_valloc = ffs_valloc; 691 ump->um_vfree = ffs_vfree; 692 ump->um_ifree = ffs_ifree; 693 bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize); 694 if (fs->fs_sbsize < SBLOCKSIZE) 695 bp->b_flags |= B_INVAL | B_NOCACHE; 696 brelse(bp); 697 bp = NULL; 698 fs = ump->um_fs; 699 ffs_oldfscompat_read(fs, ump, sblockloc); 700 fs->fs_ronly = ronly; 701 size = fs->fs_cssize; 702 blks = howmany(size, fs->fs_fsize); 703 if (fs->fs_contigsumsize > 0) 704 size += fs->fs_ncg * sizeof(int32_t); 705 size += fs->fs_ncg * sizeof(u_int8_t); 706 space = malloc((u_long)size, M_UFSMNT, M_WAITOK); 707 fs->fs_csp = space; 708 for (i = 0; i < blks; i += fs->fs_frag) { 709 size = fs->fs_bsize; 710 if (i + fs->fs_frag > blks) 711 size = (blks - i) * fs->fs_fsize; 712 if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 713 cred, &bp)) != 0) { 714 free(fs->fs_csp, M_UFSMNT); 715 goto out; 716 } 717 bcopy(bp->b_data, space, (u_int)size); 718 space = (char *)space + size; 719 brelse(bp); 720 bp = NULL; 721 } 722 if (fs->fs_contigsumsize > 0) { 723 fs->fs_maxcluster = lp = space; 724 for (i = 0; i < fs->fs_ncg; i++) 725 *lp++ = fs->fs_contigsumsize; 726 space = lp; 727 } 728 size = fs->fs_ncg * sizeof(u_int8_t); 729 fs->fs_contigdirs = (u_int8_t *)space; 730 bzero(fs->fs_contigdirs, size); 731 fs->fs_active = NULL; 732 mp->mnt_data = (qaddr_t)ump; 733 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0]; 734 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1]; 735 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 || 736 vfs_getvfs(&mp->mnt_stat.f_fsid)) 737 vfs_getnewfsid(mp); 738 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 739 mp->mnt_flag |= MNT_LOCAL; 740 if ((fs->fs_flags & FS_MULTILABEL) != 0) 741#ifdef MAC 742 mp->mnt_flag |= MNT_MULTILABEL; 743#else 744 printf( 745"WARNING: %s: multilabel flag on fs but no MAC support\n", 746 fs->fs_fsmnt); 747#endif 748 if ((fs->fs_flags & FS_ACLS) != 0) 749#ifdef UFS_ACL 750 mp->mnt_flag |= MNT_ACLS; 751#else 752 printf( 753"WARNING: %s: ACLs flag on fs but no ACLs support\n", 754 fs->fs_fsmnt); 755#endif 756 ump->um_mountp = mp; 757 ump->um_dev = dev; 758 ump->um_devvp = devvp; 759 ump->um_nindir = fs->fs_nindir; 760 ump->um_bptrtodb = fs->fs_fsbtodb; 761 ump->um_seqinc = fs->fs_frag; 762 for (i = 0; i < MAXQUOTAS; i++) 763 ump->um_quotas[i] = NULLVP; 764#ifdef UFS_EXTATTR 765 ufs_extattr_uepm_init(&ump->um_extattr); 766#endif 767 devvp->v_rdev->si_mountpoint = mp; 768 769 /* 770 * Set FS local "last mounted on" information (NULL pad) 771 */ 772 copystr( mp->mnt_stat.f_mntonname, /* mount point*/ 773 fs->fs_fsmnt, /* copy area*/ 774 sizeof(fs->fs_fsmnt) - 1, /* max size*/ 775 &strsize); /* real size*/ 776 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize); 777 778 if( mp->mnt_flag & MNT_ROOTFS) { 779 /* 780 * Root mount; update timestamp in mount structure. 781 * this will be used by the common root mount code 782 * to update the system clock. 783 */ 784 mp->mnt_time = fs->fs_time; 785 } 786 787 if (ronly == 0) { 788 if ((fs->fs_flags & FS_DOSOFTDEP) && 789 (error = softdep_mount(devvp, mp, fs, cred)) != 0) { 790 free(fs->fs_csp, M_UFSMNT); 791 goto out; 792 } 793 if (fs->fs_snapinum[0] != 0) 794 ffs_snapshot_mount(mp); 795 fs->fs_fmod = 1; 796 fs->fs_clean = 0; 797 (void) ffs_sbupdate(ump, MNT_WAIT); 798 } 799#ifdef UFS_EXTATTR 800#ifdef UFS_EXTATTR_AUTOSTART 801 /* 802 * 803 * Auto-starting does the following: 804 * - check for /.attribute in the fs, and extattr_start if so 805 * - for each file in .attribute, enable that file with 806 * an attribute of the same name. 807 * Not clear how to report errors -- probably eat them. 808 * This would all happen while the filesystem was busy/not 809 * available, so would effectively be "atomic". 810 */ 811 (void) ufs_extattr_autostart(mp, td); 812#endif /* !UFS_EXTATTR_AUTOSTART */ 813#endif /* !UFS_EXTATTR */ 814 return (0); 815out: 816 devvp->v_rdev->si_mountpoint = NULL; 817 if (bp) 818 brelse(bp); 819 /* XXX: see comment above VOP_OPEN. */ 820#ifdef notyet 821 (void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD | FWRITE, cred, td); 822#else 823 (void)VOP_CLOSE(devvp, FREAD | FWRITE, cred, td); 824#endif 825 if (ump) { 826 free(ump->um_fs, M_UFSMNT); 827 free(ump, M_UFSMNT); 828 mp->mnt_data = (qaddr_t)0; 829 } 830 return (error); 831} 832 833#include <sys/sysctl.h> 834int bigcgs = 0; 835SYSCTL_INT(_debug, OID_AUTO, bigcgs, CTLFLAG_RW, &bigcgs, 0, ""); 836 837/* 838 * Sanity checks for loading old filesystem superblocks. 839 * See ffs_oldfscompat_write below for unwound actions. 840 * 841 * XXX - Parts get retired eventually. 842 * Unfortunately new bits get added. 843 */ 844static void 845ffs_oldfscompat_read(fs, ump, sblockloc) 846 struct fs *fs; 847 struct ufsmount *ump; 848 ufs2_daddr_t sblockloc; 849{ 850 off_t maxfilesize; 851 852 /* 853 * If not yet done, update fs_flags location and value of fs_sblockloc. 854 */ 855 if ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0) { 856 fs->fs_flags = fs->fs_old_flags; 857 fs->fs_old_flags |= FS_FLAGS_UPDATED; 858 fs->fs_sblockloc = sblockloc; 859 } 860 /* 861 * If not yet done, update UFS1 superblock with new wider fields. 862 */ 863 if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_maxbsize != fs->fs_bsize) { 864 fs->fs_maxbsize = fs->fs_bsize; 865 fs->fs_time = fs->fs_old_time; 866 fs->fs_size = fs->fs_old_size; 867 fs->fs_dsize = fs->fs_old_dsize; 868 fs->fs_csaddr = fs->fs_old_csaddr; 869 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir; 870 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree; 871 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree; 872 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree; 873 } 874 if (fs->fs_magic == FS_UFS1_MAGIC && 875 fs->fs_old_inodefmt < FS_44INODEFMT) { 876 fs->fs_maxfilesize = (u_quad_t) 1LL << 39; 877 fs->fs_qbmask = ~fs->fs_bmask; 878 fs->fs_qfmask = ~fs->fs_fmask; 879 } 880 if (fs->fs_magic == FS_UFS1_MAGIC) { 881 ump->um_savedmaxfilesize = fs->fs_maxfilesize; 882 maxfilesize = (u_int64_t)0x40000000 * fs->fs_bsize - 1; 883 if (fs->fs_maxfilesize > maxfilesize) 884 fs->fs_maxfilesize = maxfilesize; 885 } 886 /* Compatibility for old filesystems */ 887 if (fs->fs_avgfilesize <= 0) 888 fs->fs_avgfilesize = AVFILESIZ; 889 if (fs->fs_avgfpdir <= 0) 890 fs->fs_avgfpdir = AFPDIR; 891 if (bigcgs) { 892 fs->fs_save_cgsize = fs->fs_cgsize; 893 fs->fs_cgsize = fs->fs_bsize; 894 } 895} 896 897/* 898 * Unwinding superblock updates for old filesystems. 899 * See ffs_oldfscompat_read above for details. 900 * 901 * XXX - Parts get retired eventually. 902 * Unfortunately new bits get added. 903 */ 904static void 905ffs_oldfscompat_write(fs, ump) 906 struct fs *fs; 907 struct ufsmount *ump; 908{ 909 910 /* 911 * Copy back UFS2 updated fields that UFS1 inspects. 912 */ 913 if (fs->fs_magic == FS_UFS1_MAGIC) { 914 fs->fs_old_time = fs->fs_time; 915 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir; 916 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree; 917 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree; 918 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree; 919 fs->fs_maxfilesize = ump->um_savedmaxfilesize; 920 } 921 if (bigcgs) { 922 fs->fs_cgsize = fs->fs_save_cgsize; 923 fs->fs_save_cgsize = 0; 924 } 925} 926 927/* 928 * unmount system call 929 */ 930int 931ffs_unmount(mp, mntflags, td) 932 struct mount *mp; 933 int mntflags; 934 struct thread *td; 935{ 936 struct ufsmount *ump = VFSTOUFS(mp); 937 struct fs *fs; 938 int error, flags; 939 940 flags = 0; 941 if (mntflags & MNT_FORCE) { 942 flags |= FORCECLOSE; 943 } 944#ifdef UFS_EXTATTR 945 if ((error = ufs_extattr_stop(mp, td))) { 946 if (error != EOPNOTSUPP) 947 printf("ffs_unmount: ufs_extattr_stop returned %d\n", 948 error); 949 } else { 950 ufs_extattr_uepm_destroy(&ump->um_extattr); 951 } 952#endif 953 if (mp->mnt_flag & MNT_SOFTDEP) { 954 if ((error = softdep_flushfiles(mp, flags, td)) != 0) 955 return (error); 956 } else { 957 if ((error = ffs_flushfiles(mp, flags, td)) != 0) 958 return (error); 959 } 960 fs = ump->um_fs; 961 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 962 printf("%s: unmount pending error: blocks %jd files %d\n", 963 fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, 964 fs->fs_pendinginodes); 965 fs->fs_pendingblocks = 0; 966 fs->fs_pendinginodes = 0; 967 } 968 if (fs->fs_ronly == 0) { 969 fs->fs_clean = fs->fs_flags & (FS_UNCLEAN|FS_NEEDSFSCK) ? 0 : 1; 970 error = ffs_sbupdate(ump, MNT_WAIT); 971 if (error) { 972 fs->fs_clean = 0; 973 return (error); 974 } 975 } 976 ump->um_devvp->v_rdev->si_mountpoint = NULL; 977 978 vinvalbuf(ump->um_devvp, V_SAVE, NOCRED, td, 0, 0); 979 /* XXX: see comment above VOP_OPEN. */ 980#ifdef notyet 981 error = VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE, 982 NOCRED, td); 983#else 984 error = VOP_CLOSE(ump->um_devvp, FREAD | FWRITE, NOCRED, td); 985#endif 986 vrele(ump->um_devvp); 987 free(fs->fs_csp, M_UFSMNT); 988 free(fs, M_UFSMNT); 989 free(ump, M_UFSMNT); 990 mp->mnt_data = (qaddr_t)0; 991 mp->mnt_flag &= ~MNT_LOCAL; 992 return (error); 993} 994 995/* 996 * Flush out all the files in a filesystem. 997 */ 998int 999ffs_flushfiles(mp, flags, td) 1000 struct mount *mp; 1001 int flags; 1002 struct thread *td; 1003{ 1004 struct ufsmount *ump; 1005 int error; 1006 1007 ump = VFSTOUFS(mp); 1008#ifdef QUOTA 1009 if (mp->mnt_flag & MNT_QUOTA) { 1010 int i; 1011 error = vflush(mp, 0, SKIPSYSTEM|flags); 1012 if (error) 1013 return (error); 1014 for (i = 0; i < MAXQUOTAS; i++) { 1015 if (ump->um_quotas[i] == NULLVP) 1016 continue; 1017 quotaoff(td, mp, i); 1018 } 1019 /* 1020 * Here we fall through to vflush again to ensure 1021 * that we have gotten rid of all the system vnodes. 1022 */ 1023 } 1024#endif 1025 ASSERT_VOP_LOCKED(ump->um_devvp, "ffs_flushfiles"); 1026 if (ump->um_devvp->v_vflag & VV_COPYONWRITE) { 1027 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0) 1028 return (error); 1029 ffs_snapshot_unmount(mp); 1030 /* 1031 * Here we fall through to vflush again to ensure 1032 * that we have gotten rid of all the system vnodes. 1033 */ 1034 } 1035 /* 1036 * Flush all the files. 1037 */ 1038 if ((error = vflush(mp, 0, flags)) != 0) 1039 return (error); 1040 /* 1041 * Flush filesystem metadata. 1042 */ 1043 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td); 1044 error = VOP_FSYNC(ump->um_devvp, td->td_ucred, MNT_WAIT, td); 1045 VOP_UNLOCK(ump->um_devvp, 0, td); 1046 return (error); 1047} 1048 1049/* 1050 * Get filesystem statistics. 1051 */ 1052int 1053ffs_statfs(mp, sbp, td) 1054 struct mount *mp; 1055 struct statfs *sbp; 1056 struct thread *td; 1057{ 1058 struct ufsmount *ump; 1059 struct fs *fs; 1060 1061 ump = VFSTOUFS(mp); 1062 fs = ump->um_fs; 1063 if (fs->fs_magic != FS_UFS1_MAGIC && fs->fs_magic != FS_UFS2_MAGIC) 1064 panic("ffs_statfs"); 1065 if (fs->fs_magic == FS_UFS2_MAGIC) 1066 bcopy((caddr_t)"ufs2", 1067 (caddr_t)&sbp->f_fstypename[0], MFSNAMELEN); 1068 sbp->f_version = STATFS_VERSION; 1069 sbp->f_bsize = fs->fs_fsize; 1070 sbp->f_iosize = fs->fs_bsize; 1071 sbp->f_blocks = fs->fs_dsize; 1072 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag + 1073 fs->fs_cstotal.cs_nffree + dbtofsb(fs, fs->fs_pendingblocks); 1074 sbp->f_bavail = freespace(fs, fs->fs_minfree) + 1075 dbtofsb(fs, fs->fs_pendingblocks); 1076 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO; 1077 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes; 1078 sbp->f_namemax = NAME_MAX; 1079 if (sbp != &mp->mnt_stat) { 1080 sbp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 1081 sbp->f_type = mp->mnt_vfc->vfc_typenum; 1082 sbp->f_syncwrites = mp->mnt_stat.f_syncwrites; 1083 sbp->f_asyncwrites = mp->mnt_stat.f_asyncwrites; 1084 sbp->f_syncreads = mp->mnt_stat.f_syncreads; 1085 sbp->f_asyncreads = mp->mnt_stat.f_asyncreads; 1086 sbp->f_owner = mp->mnt_stat.f_owner; 1087 sbp->f_fsid = mp->mnt_stat.f_fsid; 1088 if (fs->fs_magic == FS_UFS2_MAGIC) 1089 bcopy((caddr_t)"ufs2", 1090 (caddr_t)&sbp->f_fstypename[0], MFSNAMELEN); 1091 else 1092 bcopy((caddr_t)mp->mnt_stat.f_fstypename, 1093 (caddr_t)&sbp->f_fstypename[0], MFSNAMELEN); 1094 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 1095 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 1096 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 1097 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 1098 } 1099 return (0); 1100} 1101 1102/* 1103 * Go through the disk queues to initiate sandbagged IO; 1104 * go through the inodes to write those that have been modified; 1105 * initiate the writing of the super block if it has been modified. 1106 * 1107 * Note: we are always called with the filesystem marked `MPBUSY'. 1108 */ 1109int 1110ffs_sync(mp, waitfor, cred, td) 1111 struct mount *mp; 1112 int waitfor; 1113 struct ucred *cred; 1114 struct thread *td; 1115{ 1116 struct vnode *nvp, *vp, *devvp; 1117 struct inode *ip; 1118 struct ufsmount *ump = VFSTOUFS(mp); 1119 struct fs *fs; 1120 int error, count, wait, lockreq, allerror = 0; 1121 1122 fs = ump->um_fs; 1123 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */ 1124 printf("fs = %s\n", fs->fs_fsmnt); 1125 panic("ffs_sync: rofs mod"); 1126 } 1127 /* 1128 * Write back each (modified) inode. 1129 */ 1130 wait = 0; 1131 lockreq = LK_EXCLUSIVE | LK_NOWAIT; 1132 if (waitfor == MNT_WAIT) { 1133 wait = 1; 1134 lockreq = LK_EXCLUSIVE; 1135 } 1136 lockreq |= LK_INTERLOCK; 1137 MNT_ILOCK(mp); 1138loop: 1139 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp != NULL; vp = nvp) { 1140 /* 1141 * If the vnode that we are about to sync is no longer 1142 * associated with this mount point, start over. 1143 */ 1144 if (vp->v_mount != mp) 1145 goto loop; 1146 1147 /* 1148 * Depend on the mntvnode_slock to keep things stable enough 1149 * for a quick test. Since there might be hundreds of 1150 * thousands of vnodes, we cannot afford even a subroutine 1151 * call unless there's a good chance that we have work to do. 1152 */ 1153 nvp = TAILQ_NEXT(vp, v_nmntvnodes); 1154 VI_LOCK(vp); 1155 if (vp->v_iflag & VI_XLOCK) { 1156 VI_UNLOCK(vp); 1157 continue; 1158 } 1159 ip = VTOI(vp); 1160 if (vp->v_type == VNON || ((ip->i_flag & 1161 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 1162 TAILQ_EMPTY(&vp->v_dirtyblkhd))) { 1163 VI_UNLOCK(vp); 1164 continue; 1165 } 1166 MNT_IUNLOCK(mp); 1167 if ((error = vget(vp, lockreq, td)) != 0) { 1168 MNT_ILOCK(mp); 1169 if (error == ENOENT) 1170 goto loop; 1171 continue; 1172 } 1173 if ((error = VOP_FSYNC(vp, cred, waitfor, td)) != 0) 1174 allerror = error; 1175 VOP_UNLOCK(vp, 0, td); 1176 vrele(vp); 1177 MNT_ILOCK(mp); 1178 if (TAILQ_NEXT(vp, v_nmntvnodes) != nvp) 1179 goto loop; 1180 } 1181 MNT_IUNLOCK(mp); 1182 /* 1183 * Force stale filesystem control information to be flushed. 1184 */ 1185 if (waitfor == MNT_WAIT) { 1186 if ((error = softdep_flushworklist(ump->um_mountp, &count, td))) 1187 allerror = error; 1188 /* Flushed work items may create new vnodes to clean */ 1189 if (allerror == 0 && count) { 1190 MNT_ILOCK(mp); 1191 goto loop; 1192 } 1193 } 1194#ifdef QUOTA 1195 qsync(mp); 1196#endif 1197 devvp = ump->um_devvp; 1198 VI_LOCK(devvp); 1199 if (waitfor != MNT_LAZY && 1200 (devvp->v_numoutput > 0 || TAILQ_FIRST(&devvp->v_dirtyblkhd))) { 1201 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY | LK_INTERLOCK, td); 1202 if ((error = VOP_FSYNC(devvp, cred, waitfor, td)) != 0) 1203 allerror = error; 1204 VOP_UNLOCK(devvp, 0, td); 1205 if (allerror == 0 && waitfor == MNT_WAIT) { 1206 MNT_ILOCK(mp); 1207 goto loop; 1208 } 1209 } else 1210 VI_UNLOCK(devvp); 1211 /* 1212 * Write back modified superblock. 1213 */ 1214 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0) 1215 allerror = error; 1216 return (allerror); 1217} 1218 1219int 1220ffs_vget(mp, ino, flags, vpp) 1221 struct mount *mp; 1222 ino_t ino; 1223 int flags; 1224 struct vnode **vpp; 1225{ 1226 struct thread *td = curthread; /* XXX */ 1227 struct fs *fs; 1228 struct inode *ip; 1229 struct ufsmount *ump; 1230 struct buf *bp; 1231 struct vnode *vp; 1232 dev_t dev; 1233 int error; 1234 1235 ump = VFSTOUFS(mp); 1236 dev = ump->um_dev; 1237 1238 /* 1239 * We do not lock vnode creation as it is believed to be too 1240 * expensive for such rare case as simultaneous creation of vnode 1241 * for same ino by different processes. We just allow them to race 1242 * and check later to decide who wins. Let the race begin! 1243 */ 1244 if ((error = ufs_ihashget(dev, ino, flags, vpp)) != 0) 1245 return (error); 1246 if (*vpp != NULL) 1247 return (0); 1248 1249 /* 1250 * If this MALLOC() is performed after the getnewvnode() 1251 * it might block, leaving a vnode with a NULL v_data to be 1252 * found by ffs_sync() if a sync happens to fire right then, 1253 * which will cause a panic because ffs_sync() blindly 1254 * dereferences vp->v_data (as well it should). 1255 */ 1256 ip = uma_zalloc(uma_inode, M_WAITOK); 1257 1258 /* Allocate a new vnode/inode. */ 1259 error = getnewvnode("ufs", mp, ffs_vnodeop_p, &vp); 1260 if (error) { 1261 *vpp = NULL; 1262 uma_zfree(uma_inode, ip); 1263 return (error); 1264 } 1265 bzero((caddr_t)ip, sizeof(struct inode)); 1266 /* 1267 * FFS supports recursive locking. 1268 */ 1269 vp->v_vnlock->lk_flags |= LK_CANRECURSE; 1270 vp->v_data = ip; 1271 ip->i_vnode = vp; 1272 ip->i_ump = ump; 1273 ip->i_fs = fs = ump->um_fs; 1274 ip->i_dev = dev; 1275 ip->i_number = ino; 1276#ifdef QUOTA 1277 { 1278 int i; 1279 for (i = 0; i < MAXQUOTAS; i++) 1280 ip->i_dquot[i] = NODQUOT; 1281 } 1282#endif 1283 /* 1284 * Exclusively lock the vnode before adding to hash. Note, that we 1285 * must not release nor downgrade the lock (despite flags argument 1286 * says) till it is fully initialized. 1287 */ 1288 lockmgr(vp->v_vnlock, LK_EXCLUSIVE, (struct mtx *)0, td); 1289 1290 /* 1291 * Atomicaly (in terms of ufs_hash operations) check the hash for 1292 * duplicate of vnode being created and add it to the hash. If a 1293 * duplicate vnode was found, it will be vget()ed from hash for us. 1294 */ 1295 if ((error = ufs_ihashins(ip, flags, vpp)) != 0) { 1296 vput(vp); 1297 *vpp = NULL; 1298 return (error); 1299 } 1300 1301 /* We lost the race, then throw away our vnode and return existing */ 1302 if (*vpp != NULL) { 1303 vput(vp); 1304 return (0); 1305 } 1306 1307 /* Read in the disk contents for the inode, copy into the inode. */ 1308 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 1309 (int)fs->fs_bsize, NOCRED, &bp); 1310 if (error) { 1311 /* 1312 * The inode does not contain anything useful, so it would 1313 * be misleading to leave it on its hash chain. With mode 1314 * still zero, it will be unlinked and returned to the free 1315 * list by vput(). 1316 */ 1317 brelse(bp); 1318 vput(vp); 1319 *vpp = NULL; 1320 return (error); 1321 } 1322 if (ip->i_ump->um_fstype == UFS1) 1323 ip->i_din1 = uma_zalloc(uma_ufs1, M_WAITOK); 1324 else 1325 ip->i_din2 = uma_zalloc(uma_ufs2, M_WAITOK); 1326 ffs_load_inode(bp, ip, fs, ino); 1327 if (DOINGSOFTDEP(vp)) 1328 softdep_load_inodeblock(ip); 1329 else 1330 ip->i_effnlink = ip->i_nlink; 1331 bqrelse(bp); 1332 1333 /* 1334 * Initialize the vnode from the inode, check for aliases. 1335 * Note that the underlying vnode may have changed. 1336 */ 1337 error = ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp); 1338 if (error) { 1339 vput(vp); 1340 *vpp = NULL; 1341 return (error); 1342 } 1343 /* 1344 * Finish inode initialization. 1345 */ 1346 VREF(ip->i_devvp); 1347 /* 1348 * Set up a generation number for this inode if it does not 1349 * already have one. This should only happen on old filesystems. 1350 */ 1351 if (ip->i_gen == 0) { 1352 ip->i_gen = arc4random() / 2 + 1; 1353 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 1354 ip->i_flag |= IN_MODIFIED; 1355 DIP(ip, i_gen) = ip->i_gen; 1356 } 1357 } 1358 /* 1359 * Ensure that uid and gid are correct. This is a temporary 1360 * fix until fsck has been changed to do the update. 1361 */ 1362 if (fs->fs_magic == FS_UFS1_MAGIC && /* XXX */ 1363 fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */ 1364 ip->i_uid = ip->i_din1->di_ouid; /* XXX */ 1365 ip->i_gid = ip->i_din1->di_ogid; /* XXX */ 1366 } /* XXX */ 1367 1368#ifdef MAC 1369 if ((mp->mnt_flag & MNT_MULTILABEL) && ip->i_mode) { 1370 /* 1371 * If this vnode is already allocated, and we're running 1372 * multi-label, attempt to perform a label association 1373 * from the extended attributes on the inode. 1374 */ 1375 error = mac_associate_vnode_extattr(mp, vp); 1376 if (error) { 1377 /* ufs_inactive will release ip->i_devvp ref. */ 1378 vput(vp); 1379 *vpp = NULL; 1380 return (error); 1381 } 1382 } 1383#endif 1384 1385 *vpp = vp; 1386 return (0); 1387} 1388 1389/* 1390 * File handle to vnode 1391 * 1392 * Have to be really careful about stale file handles: 1393 * - check that the inode number is valid 1394 * - call ffs_vget() to get the locked inode 1395 * - check for an unallocated inode (i_mode == 0) 1396 * - check that the given client host has export rights and return 1397 * those rights via. exflagsp and credanonp 1398 */ 1399int 1400ffs_fhtovp(mp, fhp, vpp) 1401 struct mount *mp; 1402 struct fid *fhp; 1403 struct vnode **vpp; 1404{ 1405 struct ufid *ufhp; 1406 struct fs *fs; 1407 1408 ufhp = (struct ufid *)fhp; 1409 fs = VFSTOUFS(mp)->um_fs; 1410 if (ufhp->ufid_ino < ROOTINO || 1411 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg) 1412 return (ESTALE); 1413 return (ufs_fhtovp(mp, ufhp, vpp)); 1414} 1415 1416/* 1417 * Vnode pointer to File handle 1418 */ 1419/* ARGSUSED */ 1420int 1421ffs_vptofh(vp, fhp) 1422 struct vnode *vp; 1423 struct fid *fhp; 1424{ 1425 struct inode *ip; 1426 struct ufid *ufhp; 1427 1428 ip = VTOI(vp); 1429 ufhp = (struct ufid *)fhp; 1430 ufhp->ufid_len = sizeof(struct ufid); 1431 ufhp->ufid_ino = ip->i_number; 1432 ufhp->ufid_gen = ip->i_gen; 1433 return (0); 1434} 1435 1436/* 1437 * Initialize the filesystem. 1438 */ 1439static int 1440ffs_init(vfsp) 1441 struct vfsconf *vfsp; 1442{ 1443 1444 softdep_initialize(); 1445 return (ufs_init(vfsp)); 1446} 1447 1448/* 1449 * Undo the work of ffs_init(). 1450 */ 1451static int 1452ffs_uninit(vfsp) 1453 struct vfsconf *vfsp; 1454{ 1455 int ret; 1456 1457 ret = ufs_uninit(vfsp); 1458 softdep_uninitialize(); 1459 return (ret); 1460} 1461 1462/* 1463 * Write a superblock and associated information back to disk. 1464 */ 1465static int 1466ffs_sbupdate(mp, waitfor) 1467 struct ufsmount *mp; 1468 int waitfor; 1469{ 1470 struct fs *fs = mp->um_fs; 1471 struct buf *bp; 1472 int blks; 1473 void *space; 1474 int i, size, error, allerror = 0; 1475 1476 if (fs->fs_ronly == 1 && 1477 (mp->um_mountp->mnt_flag & (MNT_RDONLY | MNT_UPDATE)) != 1478 (MNT_RDONLY | MNT_UPDATE)) 1479 panic("ffs_sbupdate: write read-only filesystem"); 1480 /* 1481 * First write back the summary information. 1482 */ 1483 blks = howmany(fs->fs_cssize, fs->fs_fsize); 1484 space = fs->fs_csp; 1485 for (i = 0; i < blks; i += fs->fs_frag) { 1486 size = fs->fs_bsize; 1487 if (i + fs->fs_frag > blks) 1488 size = (blks - i) * fs->fs_fsize; 1489 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i), 1490 size, 0, 0, 0); 1491 bcopy(space, bp->b_data, (u_int)size); 1492 space = (char *)space + size; 1493 if (waitfor != MNT_WAIT) 1494 bawrite(bp); 1495 else if ((error = bwrite(bp)) != 0) 1496 allerror = error; 1497 } 1498 /* 1499 * Now write back the superblock itself. If any errors occurred 1500 * up to this point, then fail so that the superblock avoids 1501 * being written out as clean. 1502 */ 1503 if (allerror) 1504 return (allerror); 1505 if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_sblockloc != SBLOCK_UFS1 && 1506 (fs->fs_flags & FS_FLAGS_UPDATED) == 0) { 1507 printf("%s: correcting fs_sblockloc from %jd to %d\n", 1508 fs->fs_fsmnt, fs->fs_sblockloc, SBLOCK_UFS1); 1509 fs->fs_sblockloc = SBLOCK_UFS1; 1510 } 1511 if (fs->fs_magic == FS_UFS2_MAGIC && fs->fs_sblockloc != SBLOCK_UFS2 && 1512 (fs->fs_flags & FS_FLAGS_UPDATED) == 0) { 1513 printf("%s: correcting fs_sblockloc from %jd to %d\n", 1514 fs->fs_fsmnt, fs->fs_sblockloc, SBLOCK_UFS2); 1515 fs->fs_sblockloc = SBLOCK_UFS2; 1516 } 1517 bp = getblk(mp->um_devvp, btodb(fs->fs_sblockloc), (int)fs->fs_sbsize, 1518 0, 0, 0); 1519 fs->fs_fmod = 0; 1520 fs->fs_time = time_second; 1521 bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize); 1522 ffs_oldfscompat_write((struct fs *)bp->b_data, mp); 1523 if (waitfor != MNT_WAIT) 1524 bawrite(bp); 1525 else if ((error = bwrite(bp)) != 0) 1526 allerror = error; 1527 return (allerror); 1528} 1529 1530static int 1531ffs_extattrctl(struct mount *mp, int cmd, struct vnode *filename_vp, 1532 int attrnamespace, const char *attrname, struct thread *td) 1533{ 1534 1535#ifdef UFS_EXTATTR 1536 return (ufs_extattrctl(mp, cmd, filename_vp, attrnamespace, 1537 attrname, td)); 1538#else 1539 return (vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, 1540 attrname, td)); 1541#endif 1542} 1543 1544static void 1545ffs_ifree(struct ufsmount *ump, struct inode *ip) 1546{ 1547 1548 if (ump->um_fstype == UFS1 && ip->i_din1 != NULL) 1549 uma_zfree(uma_ufs1, ip->i_din1); 1550 else if (ip->i_din2 != NULL) 1551 uma_zfree(uma_ufs2, ip->i_din2); 1552 uma_zfree(uma_inode, ip); 1553} 1554