ffs_vfsops.c revision 138290
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 138290 2004-12-01 23:16:38Z phk $"); 34 35#include "opt_mac.h" 36#include "opt_quota.h" 37#include "opt_ufs.h" 38#include "opt_ffs.h" 39 40#include <sys/param.h> 41#include <sys/systm.h> 42#include <sys/namei.h> 43#include <sys/proc.h> 44#include <sys/kernel.h> 45#include <sys/mac.h> 46#include <sys/vnode.h> 47#include <sys/mount.h> 48#include <sys/bio.h> 49#include <sys/buf.h> 50#include <sys/conf.h> 51#include <sys/fcntl.h> 52#include <sys/disk.h> 53#include <sys/malloc.h> 54#include <sys/mutex.h> 55 56#include <ufs/ufs/extattr.h> 57#include <ufs/ufs/quota.h> 58#include <ufs/ufs/ufsmount.h> 59#include <ufs/ufs/inode.h> 60#include <ufs/ufs/ufs_extern.h> 61 62#include <ufs/ffs/fs.h> 63#include <ufs/ffs/ffs_extern.h> 64 65#include <vm/vm.h> 66#include <vm/uma.h> 67#include <vm/vm_page.h> 68 69#include <geom/geom.h> 70#include <geom/geom_vfs.h> 71 72uma_zone_t uma_inode, uma_ufs1, uma_ufs2; 73 74static int ffs_sbupdate(struct ufsmount *, int); 75static int ffs_reload(struct mount *, struct thread *); 76static int ffs_mountfs(struct vnode *, struct mount *, struct thread *); 77static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, 78 ufs2_daddr_t); 79static void ffs_oldfscompat_write(struct fs *, struct ufsmount *); 80static void ffs_ifree(struct ufsmount *ump, struct inode *ip); 81static vfs_init_t ffs_init; 82static vfs_uninit_t ffs_uninit; 83static vfs_extattrctl_t ffs_extattrctl; 84static vfs_omount_t ffs_omount; 85 86static struct vfsops ufs_vfsops = { 87 .vfs_extattrctl = ffs_extattrctl, 88 .vfs_fhtovp = ffs_fhtovp, 89 .vfs_init = ffs_init, 90 .vfs_omount = ffs_omount, 91 .vfs_quotactl = ufs_quotactl, 92 .vfs_root = ufs_root, 93 .vfs_statfs = ffs_statfs, 94 .vfs_sync = ffs_sync, 95 .vfs_uninit = ffs_uninit, 96 .vfs_unmount = ffs_unmount, 97 .vfs_vget = ffs_vget, 98 .vfs_vptofh = ffs_vptofh, 99}; 100 101VFS_SET(ufs_vfsops, ufs, 0); 102 103static b_strategy_t ffs_geom_strategy; 104 105static struct buf_ops ffs_ops = { 106 .bop_name = "FFS", 107 .bop_write = bufwrite, 108 .bop_strategy = ffs_geom_strategy, 109}; 110 111/* 112 * ffs_omount 113 * 114 * Called when mounting local physical media 115 * 116 * PARAMETERS: 117 * mountroot 118 * mp mount point structure 119 * path path to mount point 120 * data <unused> 121 * ndp <unused> 122 * p process (user credentials check [statfs]) 123 * 124 * mount 125 * mp mount point structure 126 * path path to mount point 127 * data pointer to argument struct in user space 128 * ndp mount point namei() return (used for 129 * credentials on reload), reused to look 130 * up block device. 131 * p process (user credentials check) 132 * 133 * RETURNS: 0 Success 134 * !0 error number (errno.h) 135 * 136 * LOCK STATE: 137 * 138 * ENTRY 139 * mount point is locked 140 * EXIT 141 * mount point is locked 142 * 143 * NOTES: 144 * A NULL path can be used for a flag since the mount 145 * system call will fail with EFAULT in copyinstr in 146 * namei() if it is a genuine NULL from the user. 147 */ 148static int 149ffs_omount(struct mount *mp, char *path, caddr_t data, struct thread *td) 150{ 151 size_t size; 152 struct vnode *devvp, *rootvp; 153 struct ufs_args args; 154 struct ufsmount *ump = 0; 155 struct fs *fs; 156 int error, flags; 157 mode_t accessmode; 158 struct nameidata ndp; 159 160 if (uma_inode == NULL) { 161 uma_inode = uma_zcreate("FFS inode", 162 sizeof(struct inode), NULL, NULL, NULL, NULL, 163 UMA_ALIGN_PTR, 0); 164 uma_ufs1 = uma_zcreate("FFS1 dinode", 165 sizeof(struct ufs1_dinode), NULL, NULL, NULL, NULL, 166 UMA_ALIGN_PTR, 0); 167 uma_ufs2 = uma_zcreate("FFS2 dinode", 168 sizeof(struct ufs2_dinode), NULL, NULL, NULL, NULL, 169 UMA_ALIGN_PTR, 0); 170 } 171 if ((mp->mnt_flag & MNT_ROOTFS) && mp->mnt_data == NULL) { 172 if ((error = bdevvp(rootdev, &rootvp))) { 173 printf("ffs_mountroot: can't find rootvp\n"); 174 return (error); 175 } 176 177 if ((error = ffs_mountfs(rootvp, mp, td)) != 0) 178 return (error); 179 return (0); 180 } 181 182 /* 183 * Get mount options, if any. 184 */ 185 if (data != NULL) { 186 error = copyin(data, (caddr_t)&args, sizeof args); 187 if (error) 188 return (error); 189 } else { 190 memset(&args, 0, sizeof args); 191 } 192 193 /* 194 * If updating, check whether changing from read-only to 195 * read/write; if there is no device name, that's all we do. 196 */ 197 if (mp->mnt_flag & MNT_UPDATE) { 198 ump = VFSTOUFS(mp); 199 fs = ump->um_fs; 200 devvp = ump->um_devvp; 201 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 202 if ((error = vn_start_write(NULL, &mp, V_WAIT)) != 0) 203 return (error); 204 /* 205 * Flush any dirty data. 206 */ 207 if ((error = VFS_SYNC(mp, MNT_WAIT, 208 td->td_ucred, td)) != 0) { 209 vn_finished_write(mp); 210 return (error); 211 } 212 /* 213 * Check for and optionally get rid of files open 214 * for writing. 215 */ 216 flags = WRITECLOSE; 217 if (mp->mnt_flag & MNT_FORCE) 218 flags |= FORCECLOSE; 219 if (mp->mnt_flag & MNT_SOFTDEP) { 220 error = softdep_flushfiles(mp, flags, td); 221 } else { 222 error = ffs_flushfiles(mp, flags, td); 223 } 224 if (error) { 225 vn_finished_write(mp); 226 return (error); 227 } 228 if (fs->fs_pendingblocks != 0 || 229 fs->fs_pendinginodes != 0) { 230 printf("%s: %s: blocks %jd files %d\n", 231 fs->fs_fsmnt, "update error", 232 (intmax_t)fs->fs_pendingblocks, 233 fs->fs_pendinginodes); 234 fs->fs_pendingblocks = 0; 235 fs->fs_pendinginodes = 0; 236 } 237 fs->fs_ronly = 1; 238 if ((fs->fs_flags & (FS_UNCLEAN | FS_NEEDSFSCK)) == 0) 239 fs->fs_clean = 1; 240 if ((error = ffs_sbupdate(ump, MNT_WAIT)) != 0) { 241 fs->fs_ronly = 0; 242 fs->fs_clean = 0; 243 vn_finished_write(mp); 244 return (error); 245 } 246 vn_finished_write(mp); 247 DROP_GIANT(); 248 g_topology_lock(); 249 g_access(ump->um_cp, 0, -1, 0); 250 g_topology_unlock(); 251 PICKUP_GIANT(); 252 } 253 if ((mp->mnt_flag & MNT_RELOAD) && 254 (error = ffs_reload(mp, td)) != 0) 255 return (error); 256 if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { 257 /* 258 * If upgrade to read-write by non-root, then verify 259 * that user has necessary permissions on the device. 260 */ 261 if (suser(td)) { 262 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 263 if ((error = VOP_ACCESS(devvp, VREAD | VWRITE, 264 td->td_ucred, td)) != 0) { 265 VOP_UNLOCK(devvp, 0, td); 266 return (error); 267 } 268 VOP_UNLOCK(devvp, 0, td); 269 } 270 fs->fs_flags &= ~FS_UNCLEAN; 271 if (fs->fs_clean == 0) { 272 fs->fs_flags |= FS_UNCLEAN; 273 if ((mp->mnt_flag & MNT_FORCE) || 274 ((fs->fs_flags & FS_NEEDSFSCK) == 0 && 275 (fs->fs_flags & FS_DOSOFTDEP))) { 276 printf("WARNING: %s was not %s\n", 277 fs->fs_fsmnt, "properly dismounted"); 278 } else { 279 printf( 280"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", 281 fs->fs_fsmnt); 282 return (EPERM); 283 } 284 } 285 DROP_GIANT(); 286 g_topology_lock(); 287 /* 288 * If we're the root device, we may not have an E count 289 * yet, get it now. 290 */ 291 if (ump->um_cp->ace == 0) 292 error = g_access(ump->um_cp, 0, 1, 1); 293 else 294 error = g_access(ump->um_cp, 0, 1, 0); 295 g_topology_unlock(); 296 PICKUP_GIANT(); 297 if (error) 298 return (error); 299 if ((error = vn_start_write(NULL, &mp, V_WAIT)) != 0) 300 return (error); 301 fs->fs_ronly = 0; 302 fs->fs_clean = 0; 303 if ((error = ffs_sbupdate(ump, MNT_WAIT)) != 0) { 304 vn_finished_write(mp); 305 return (error); 306 } 307 /* check to see if we need to start softdep */ 308 if ((fs->fs_flags & FS_DOSOFTDEP) && 309 (error = softdep_mount(devvp, mp, fs, td->td_ucred))){ 310 vn_finished_write(mp); 311 return (error); 312 } 313 if (fs->fs_snapinum[0] != 0) 314 ffs_snapshot_mount(mp); 315 vn_finished_write(mp); 316 } 317 /* 318 * Soft updates is incompatible with "async", 319 * so if we are doing softupdates stop the user 320 * from setting the async flag in an update. 321 * Softdep_mount() clears it in an initial mount 322 * or ro->rw remount. 323 */ 324 if (mp->mnt_flag & MNT_SOFTDEP) 325 mp->mnt_flag &= ~MNT_ASYNC; 326 /* 327 * If not updating name, process export requests. 328 */ 329 if (args.fspec == 0) 330 return (vfs_export(mp, &args.export)); 331 /* 332 * If this is a snapshot request, take the snapshot. 333 */ 334 if (mp->mnt_flag & MNT_SNAPSHOT) 335 return (ffs_snapshot(mp, args.fspec)); 336 } 337 338 /* 339 * Not an update, or updating the name: look up the name 340 * and verify that it refers to a sensible disk device. 341 */ 342 NDINIT(&ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, td); 343 if ((error = namei(&ndp)) != 0) 344 return (error); 345 NDFREE(&ndp, NDF_ONLY_PNBUF); 346 devvp = ndp.ni_vp; 347 if (!vn_isdisk(devvp, &error)) { 348 vrele(devvp); 349 return (error); 350 } 351 352 /* 353 * If mount by non-root, then verify that user has necessary 354 * permissions on the device. 355 */ 356 if (suser(td)) { 357 accessmode = VREAD; 358 if ((mp->mnt_flag & MNT_RDONLY) == 0) 359 accessmode |= VWRITE; 360 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 361 if ((error = VOP_ACCESS(devvp, accessmode, td->td_ucred, td))!= 0){ 362 vput(devvp); 363 return (error); 364 } 365 VOP_UNLOCK(devvp, 0, td); 366 } 367 368 if (mp->mnt_flag & MNT_UPDATE) { 369 /* 370 * Update only 371 * 372 * If it's not the same vnode, or at least the same device 373 * then it's not correct. 374 */ 375 376 if (devvp->v_rdev != ump->um_devvp->v_rdev) 377 error = EINVAL; /* needs translation */ 378 vrele(devvp); 379 if (error) 380 return (error); 381 } else { 382 /* 383 * New mount 384 * 385 * We need the name for the mount point (also used for 386 * "last mounted on") copied in. If an error occurs, 387 * the mount point is discarded by the upper level code. 388 * Note that vfs_mount() populates f_mntonname for us. 389 */ 390 if ((error = ffs_mountfs(devvp, mp, td)) != 0) { 391 vrele(devvp); 392 return (error); 393 } 394 } 395 /* 396 * Save "mounted from" device name info for mount point (NULL pad). 397 */ 398 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size); 399 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 400 return (0); 401} 402 403/* 404 * Reload all incore data for a filesystem (used after running fsck on 405 * the root filesystem and finding things to fix). The filesystem must 406 * be mounted read-only. 407 * 408 * Things to do to update the mount: 409 * 1) invalidate all cached meta-data. 410 * 2) re-read superblock from disk. 411 * 3) re-read summary information from disk. 412 * 4) invalidate all inactive vnodes. 413 * 5) invalidate all cached file data. 414 * 6) re-read inode data for all active vnodes. 415 */ 416static int 417ffs_reload(struct mount *mp, struct thread *td) 418{ 419 struct vnode *vp, *nvp, *devvp; 420 struct inode *ip; 421 void *space; 422 struct buf *bp; 423 struct fs *fs, *newfs; 424 ufs2_daddr_t sblockloc; 425 int i, blks, size, error; 426 int32_t *lp; 427 428 if ((mp->mnt_flag & MNT_RDONLY) == 0) 429 return (EINVAL); 430 /* 431 * Step 1: invalidate all cached meta-data. 432 */ 433 devvp = VFSTOUFS(mp)->um_devvp; 434 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, td); 435 if (vinvalbuf(devvp, 0, td->td_ucred, td, 0, 0) != 0) 436 panic("ffs_reload: dirty1"); 437 vfs_object_create(devvp, td, td->td_ucred); 438 VOP_UNLOCK(devvp, 0, td); 439 440 /* 441 * Step 2: re-read superblock from disk. 442 */ 443 fs = VFSTOUFS(mp)->um_fs; 444 if ((error = bread(devvp, btodb(fs->fs_sblockloc), fs->fs_sbsize, 445 NOCRED, &bp)) != 0) 446 return (error); 447 newfs = (struct fs *)bp->b_data; 448 if ((newfs->fs_magic != FS_UFS1_MAGIC && 449 newfs->fs_magic != FS_UFS2_MAGIC) || 450 newfs->fs_bsize > MAXBSIZE || 451 newfs->fs_bsize < sizeof(struct fs)) { 452 brelse(bp); 453 return (EIO); /* XXX needs translation */ 454 } 455 /* 456 * Copy pointer fields back into superblock before copying in XXX 457 * new superblock. These should really be in the ufsmount. XXX 458 * Note that important parameters (eg fs_ncg) are unchanged. 459 */ 460 newfs->fs_csp = fs->fs_csp; 461 newfs->fs_maxcluster = fs->fs_maxcluster; 462 newfs->fs_contigdirs = fs->fs_contigdirs; 463 newfs->fs_active = fs->fs_active; 464 /* The file system is still read-only. */ 465 newfs->fs_ronly = 1; 466 sblockloc = fs->fs_sblockloc; 467 bcopy(newfs, fs, (u_int)fs->fs_sbsize); 468 brelse(bp); 469 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 470 ffs_oldfscompat_read(fs, VFSTOUFS(mp), sblockloc); 471 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 472 printf("%s: reload pending error: blocks %jd files %d\n", 473 fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, 474 fs->fs_pendinginodes); 475 fs->fs_pendingblocks = 0; 476 fs->fs_pendinginodes = 0; 477 } 478 479 /* 480 * Step 3: re-read summary information from disk. 481 */ 482 blks = howmany(fs->fs_cssize, fs->fs_fsize); 483 space = fs->fs_csp; 484 for (i = 0; i < blks; i += fs->fs_frag) { 485 size = fs->fs_bsize; 486 if (i + fs->fs_frag > blks) 487 size = (blks - i) * fs->fs_fsize; 488 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 489 NOCRED, &bp); 490 if (error) 491 return (error); 492 bcopy(bp->b_data, space, (u_int)size); 493 space = (char *)space + size; 494 brelse(bp); 495 } 496 /* 497 * We no longer know anything about clusters per cylinder group. 498 */ 499 if (fs->fs_contigsumsize > 0) { 500 lp = fs->fs_maxcluster; 501 for (i = 0; i < fs->fs_ncg; i++) 502 *lp++ = fs->fs_contigsumsize; 503 } 504 505loop: 506 MNT_ILOCK(mp); 507 MNT_VNODE_FOREACH(vp, mp, nvp) { 508 VI_LOCK(vp); 509 if (vp->v_iflag & VI_XLOCK) { 510 VI_UNLOCK(vp); 511 continue; 512 } 513 MNT_IUNLOCK(mp); 514 /* 515 * Step 4: invalidate all inactive vnodes. 516 */ 517 if (vp->v_usecount == 0) { 518 vgonel(vp, td); 519 goto loop; 520 } 521 /* 522 * Step 5: invalidate all cached file data. 523 */ 524 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) { 525 goto loop; 526 } 527 if (vinvalbuf(vp, 0, td->td_ucred, td, 0, 0)) 528 panic("ffs_reload: dirty2"); 529 /* 530 * Step 6: re-read inode data for all active vnodes. 531 */ 532 ip = VTOI(vp); 533 error = 534 bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), 535 (int)fs->fs_bsize, NOCRED, &bp); 536 if (error) { 537 VOP_UNLOCK(vp, 0, td); 538 vrele(vp); 539 return (error); 540 } 541 ffs_load_inode(bp, ip, fs, ip->i_number); 542 ip->i_effnlink = ip->i_nlink; 543 brelse(bp); 544 VOP_UNLOCK(vp, 0, td); 545 vrele(vp); 546 MNT_ILOCK(mp); 547 } 548 MNT_IUNLOCK(mp); 549 return (0); 550} 551 552/* 553 * Possible superblock locations ordered from most to least likely. 554 */ 555static int sblock_try[] = SBLOCKSEARCH; 556 557/* 558 * Common code for mount and mountroot 559 */ 560static int 561ffs_mountfs(devvp, mp, td) 562 struct vnode *devvp; 563 struct mount *mp; 564 struct thread *td; 565{ 566 struct ufsmount *ump; 567 struct buf *bp; 568 struct fs *fs; 569 struct cdev *dev; 570 void *space; 571 ufs2_daddr_t sblockloc; 572 int error, i, blks, size, ronly; 573 int32_t *lp; 574 struct ucred *cred; 575 size_t strsize; 576 struct g_consumer *cp; 577 578 dev = devvp->v_rdev; 579 cred = td ? td->td_ucred : NOCRED; 580 581 vfs_object_create(devvp, td, td->td_ucred); 582 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 583#if 0 584 /* 585 * XXX: check filesystem permissions, they may be more strict 586 * XXX: than what geom enforces. 587 * XXX: But since we're root, they wouldn't matter, would they ? 588 */ 589 error = VOP_ACCESS(devvp, ronly ? FREAD : FREAD | FWRITE, FSCRED, td); 590 if (error) { 591 VOP_UNLOCK(devvp, 0, td); 592 return (error); 593 } 594#endif 595 DROP_GIANT(); 596 g_topology_lock(); 597 error = g_vfs_open(devvp, &cp, "ffs", ronly ? 0 : 1); 598 599 /* 600 * If we are a root mount, drop the E flag so fsck can do its magic. 601 * We will pick it up again when we remounte R/W. 602 */ 603 if (error == 0 && ronly && (mp->mnt_flag & MNT_ROOTFS)) 604 error = g_access(cp, 0, 0, -1); 605 g_topology_unlock(); 606 PICKUP_GIANT(); 607 VOP_UNLOCK(devvp, 0, td); 608 if (error) 609 return (error); 610 if (devvp->v_rdev->si_iosize_max != 0) 611 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max; 612 if (mp->mnt_iosize_max > MAXPHYS) 613 mp->mnt_iosize_max = MAXPHYS; 614 615 devvp->v_bufobj.bo_private = cp; 616 devvp->v_bufobj.bo_ops = &ffs_ops; 617 618 bp = NULL; 619 ump = NULL; 620 fs = NULL; 621 sblockloc = 0; 622 /* 623 * Try reading the superblock in each of its possible locations. 624 */ 625 for (i = 0; sblock_try[i] != -1; i++) { 626 if ((error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, 627 cred, &bp)) != 0) 628 goto out; 629 fs = (struct fs *)bp->b_data; 630 sblockloc = sblock_try[i]; 631 if ((fs->fs_magic == FS_UFS1_MAGIC || 632 (fs->fs_magic == FS_UFS2_MAGIC && 633 (fs->fs_sblockloc == sblockloc || 634 (fs->fs_old_flags & FS_FLAGS_UPDATED) == 0))) && 635 fs->fs_bsize <= MAXBSIZE && 636 fs->fs_bsize >= sizeof(struct fs)) 637 break; 638 brelse(bp); 639 bp = NULL; 640 } 641 if (sblock_try[i] == -1) { 642 error = EINVAL; /* XXX needs translation */ 643 goto out; 644 } 645 fs->fs_fmod = 0; 646 fs->fs_flags &= ~FS_INDEXDIRS; /* no support for directory indicies */ 647 fs->fs_flags &= ~FS_UNCLEAN; 648 if (fs->fs_clean == 0) { 649 fs->fs_flags |= FS_UNCLEAN; 650 if (ronly || (mp->mnt_flag & MNT_FORCE) || 651 ((fs->fs_flags & FS_NEEDSFSCK) == 0 && 652 (fs->fs_flags & FS_DOSOFTDEP))) { 653 printf( 654"WARNING: %s was not properly dismounted\n", 655 fs->fs_fsmnt); 656 } else { 657 printf( 658"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n", 659 fs->fs_fsmnt); 660 error = EPERM; 661 goto out; 662 } 663 if ((fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) && 664 (mp->mnt_flag & MNT_FORCE)) { 665 printf("%s: lost blocks %jd files %d\n", fs->fs_fsmnt, 666 (intmax_t)fs->fs_pendingblocks, 667 fs->fs_pendinginodes); 668 fs->fs_pendingblocks = 0; 669 fs->fs_pendinginodes = 0; 670 } 671 } 672 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 673 printf("%s: mount pending error: blocks %jd files %d\n", 674 fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, 675 fs->fs_pendinginodes); 676 fs->fs_pendingblocks = 0; 677 fs->fs_pendinginodes = 0; 678 } 679 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK | M_ZERO); 680 ump->um_cp = cp; 681 ump->um_bo = &devvp->v_bufobj; 682 ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT, M_WAITOK); 683 if (fs->fs_magic == FS_UFS1_MAGIC) { 684 ump->um_fstype = UFS1; 685 ump->um_balloc = ffs_balloc_ufs1; 686 } else { 687 ump->um_fstype = UFS2; 688 ump->um_balloc = ffs_balloc_ufs2; 689 } 690 ump->um_blkatoff = ffs_blkatoff; 691 ump->um_truncate = ffs_truncate; 692 ump->um_update = ffs_update; 693 ump->um_valloc = ffs_valloc; 694 ump->um_vfree = ffs_vfree; 695 ump->um_ifree = ffs_ifree; 696 bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize); 697 if (fs->fs_sbsize < SBLOCKSIZE) 698 bp->b_flags |= B_INVAL | B_NOCACHE; 699 brelse(bp); 700 bp = NULL; 701 fs = ump->um_fs; 702 ffs_oldfscompat_read(fs, ump, sblockloc); 703 fs->fs_ronly = ronly; 704 size = fs->fs_cssize; 705 blks = howmany(size, fs->fs_fsize); 706 if (fs->fs_contigsumsize > 0) 707 size += fs->fs_ncg * sizeof(int32_t); 708 size += fs->fs_ncg * sizeof(u_int8_t); 709 space = malloc((u_long)size, M_UFSMNT, M_WAITOK); 710 fs->fs_csp = space; 711 for (i = 0; i < blks; i += fs->fs_frag) { 712 size = fs->fs_bsize; 713 if (i + fs->fs_frag > blks) 714 size = (blks - i) * fs->fs_fsize; 715 if ((error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size, 716 cred, &bp)) != 0) { 717 free(fs->fs_csp, M_UFSMNT); 718 goto out; 719 } 720 bcopy(bp->b_data, space, (u_int)size); 721 space = (char *)space + size; 722 brelse(bp); 723 bp = NULL; 724 } 725 if (fs->fs_contigsumsize > 0) { 726 fs->fs_maxcluster = lp = space; 727 for (i = 0; i < fs->fs_ncg; i++) 728 *lp++ = fs->fs_contigsumsize; 729 space = lp; 730 } 731 size = fs->fs_ncg * sizeof(u_int8_t); 732 fs->fs_contigdirs = (u_int8_t *)space; 733 bzero(fs->fs_contigdirs, size); 734 fs->fs_active = NULL; 735 mp->mnt_data = (qaddr_t)ump; 736 mp->mnt_stat.f_fsid.val[0] = fs->fs_id[0]; 737 mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1]; 738 if (fs->fs_id[0] == 0 || fs->fs_id[1] == 0 || 739 vfs_getvfs(&mp->mnt_stat.f_fsid)) 740 vfs_getnewfsid(mp); 741 mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen; 742 mp->mnt_flag |= MNT_LOCAL; 743 if ((fs->fs_flags & FS_MULTILABEL) != 0) 744#ifdef MAC 745 mp->mnt_flag |= MNT_MULTILABEL; 746#else 747 printf( 748"WARNING: %s: multilabel flag on fs but no MAC support\n", 749 fs->fs_fsmnt); 750#endif 751 if ((fs->fs_flags & FS_ACLS) != 0) 752#ifdef UFS_ACL 753 mp->mnt_flag |= MNT_ACLS; 754#else 755 printf( 756"WARNING: %s: ACLs flag on fs but no ACLs support\n", 757 fs->fs_fsmnt); 758#endif 759 ump->um_mountp = mp; 760 ump->um_dev = dev; 761 ump->um_devvp = devvp; 762 ump->um_nindir = fs->fs_nindir; 763 ump->um_bptrtodb = fs->fs_fsbtodb; 764 ump->um_seqinc = fs->fs_frag; 765 for (i = 0; i < MAXQUOTAS; i++) 766 ump->um_quotas[i] = NULLVP; 767#ifdef UFS_EXTATTR 768 ufs_extattr_uepm_init(&ump->um_extattr); 769#endif 770 /* 771 * Set FS local "last mounted on" information (NULL pad) 772 */ 773 copystr( mp->mnt_stat.f_mntonname, /* mount point*/ 774 fs->fs_fsmnt, /* copy area*/ 775 sizeof(fs->fs_fsmnt) - 1, /* max size*/ 776 &strsize); /* real size*/ 777 bzero( fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize); 778 779 if( mp->mnt_flag & MNT_ROOTFS) { 780 /* 781 * Root mount; update timestamp in mount structure. 782 * this will be used by the common root mount code 783 * to update the system clock. 784 */ 785 mp->mnt_time = fs->fs_time; 786 } 787 788 if (ronly == 0) { 789 if ((fs->fs_flags & FS_DOSOFTDEP) && 790 (error = softdep_mount(devvp, mp, fs, cred)) != 0) { 791 free(fs->fs_csp, M_UFSMNT); 792 goto out; 793 } 794 if (fs->fs_snapinum[0] != 0) 795 ffs_snapshot_mount(mp); 796 fs->fs_fmod = 1; 797 fs->fs_clean = 0; 798 (void) ffs_sbupdate(ump, MNT_WAIT); 799 } 800 /* 801 * Initialize filesystem stat information in mount struct. 802 */ 803 (void)VFS_STATFS(mp, &mp->mnt_stat, td); 804#ifdef UFS_EXTATTR 805#ifdef UFS_EXTATTR_AUTOSTART 806 /* 807 * 808 * Auto-starting does the following: 809 * - check for /.attribute in the fs, and extattr_start if so 810 * - for each file in .attribute, enable that file with 811 * an attribute of the same name. 812 * Not clear how to report errors -- probably eat them. 813 * This would all happen while the filesystem was busy/not 814 * available, so would effectively be "atomic". 815 */ 816 (void) ufs_extattr_autostart(mp, td); 817#endif /* !UFS_EXTATTR_AUTOSTART */ 818#endif /* !UFS_EXTATTR */ 819 return (0); 820out: 821 if (bp) 822 brelse(bp); 823 if (cp != NULL) { 824 DROP_GIANT(); 825 g_topology_lock(); 826 g_wither_geom_close(cp->geom, ENXIO); 827 g_topology_unlock(); 828 PICKUP_GIANT(); 829 } 830 if (ump) { 831 free(ump->um_fs, M_UFSMNT); 832 free(ump, M_UFSMNT); 833 mp->mnt_data = (qaddr_t)0; 834 } 835 return (error); 836} 837 838#include <sys/sysctl.h> 839int bigcgs = 0; 840SYSCTL_INT(_debug, OID_AUTO, bigcgs, CTLFLAG_RW, &bigcgs, 0, ""); 841 842/* 843 * Sanity checks for loading old filesystem superblocks. 844 * See ffs_oldfscompat_write below for unwound actions. 845 * 846 * XXX - Parts get retired eventually. 847 * Unfortunately new bits get added. 848 */ 849static void 850ffs_oldfscompat_read(fs, ump, sblockloc) 851 struct fs *fs; 852 struct ufsmount *ump; 853 ufs2_daddr_t sblockloc; 854{ 855 off_t maxfilesize; 856 857 /* 858 * If not yet done, update fs_flags location and value of fs_sblockloc. 859 */ 860 if ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0) { 861 fs->fs_flags = fs->fs_old_flags; 862 fs->fs_old_flags |= FS_FLAGS_UPDATED; 863 fs->fs_sblockloc = sblockloc; 864 } 865 /* 866 * If not yet done, update UFS1 superblock with new wider fields. 867 */ 868 if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_maxbsize != fs->fs_bsize) { 869 fs->fs_maxbsize = fs->fs_bsize; 870 fs->fs_time = fs->fs_old_time; 871 fs->fs_size = fs->fs_old_size; 872 fs->fs_dsize = fs->fs_old_dsize; 873 fs->fs_csaddr = fs->fs_old_csaddr; 874 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir; 875 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree; 876 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree; 877 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree; 878 } 879 if (fs->fs_magic == FS_UFS1_MAGIC && 880 fs->fs_old_inodefmt < FS_44INODEFMT) { 881 fs->fs_maxfilesize = (u_quad_t) 1LL << 39; 882 fs->fs_qbmask = ~fs->fs_bmask; 883 fs->fs_qfmask = ~fs->fs_fmask; 884 } 885 if (fs->fs_magic == FS_UFS1_MAGIC) { 886 ump->um_savedmaxfilesize = fs->fs_maxfilesize; 887 maxfilesize = (u_int64_t)0x40000000 * fs->fs_bsize - 1; 888 if (fs->fs_maxfilesize > maxfilesize) 889 fs->fs_maxfilesize = maxfilesize; 890 } 891 /* Compatibility for old filesystems */ 892 if (fs->fs_avgfilesize <= 0) 893 fs->fs_avgfilesize = AVFILESIZ; 894 if (fs->fs_avgfpdir <= 0) 895 fs->fs_avgfpdir = AFPDIR; 896 if (bigcgs) { 897 fs->fs_save_cgsize = fs->fs_cgsize; 898 fs->fs_cgsize = fs->fs_bsize; 899 } 900} 901 902/* 903 * Unwinding superblock updates for old filesystems. 904 * See ffs_oldfscompat_read above for details. 905 * 906 * XXX - Parts get retired eventually. 907 * Unfortunately new bits get added. 908 */ 909static void 910ffs_oldfscompat_write(fs, ump) 911 struct fs *fs; 912 struct ufsmount *ump; 913{ 914 915 /* 916 * Copy back UFS2 updated fields that UFS1 inspects. 917 */ 918 if (fs->fs_magic == FS_UFS1_MAGIC) { 919 fs->fs_old_time = fs->fs_time; 920 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir; 921 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree; 922 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree; 923 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree; 924 fs->fs_maxfilesize = ump->um_savedmaxfilesize; 925 } 926 if (bigcgs) { 927 fs->fs_cgsize = fs->fs_save_cgsize; 928 fs->fs_save_cgsize = 0; 929 } 930} 931 932/* 933 * unmount system call 934 */ 935int 936ffs_unmount(mp, mntflags, td) 937 struct mount *mp; 938 int mntflags; 939 struct thread *td; 940{ 941 struct ufsmount *ump = VFSTOUFS(mp); 942 struct fs *fs; 943 int error, flags; 944 945 flags = 0; 946 if (mntflags & MNT_FORCE) { 947 flags |= FORCECLOSE; 948 } 949#ifdef UFS_EXTATTR 950 if ((error = ufs_extattr_stop(mp, td))) { 951 if (error != EOPNOTSUPP) 952 printf("ffs_unmount: ufs_extattr_stop returned %d\n", 953 error); 954 } else { 955 ufs_extattr_uepm_destroy(&ump->um_extattr); 956 } 957#endif 958 if (mp->mnt_flag & MNT_SOFTDEP) { 959 if ((error = softdep_flushfiles(mp, flags, td)) != 0) 960 return (error); 961 } else { 962 if ((error = ffs_flushfiles(mp, flags, td)) != 0) 963 return (error); 964 } 965 fs = ump->um_fs; 966 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 967 printf("%s: unmount pending error: blocks %jd files %d\n", 968 fs->fs_fsmnt, (intmax_t)fs->fs_pendingblocks, 969 fs->fs_pendinginodes); 970 fs->fs_pendingblocks = 0; 971 fs->fs_pendinginodes = 0; 972 } 973 if (fs->fs_ronly == 0) { 974 fs->fs_clean = fs->fs_flags & (FS_UNCLEAN|FS_NEEDSFSCK) ? 0 : 1; 975 error = ffs_sbupdate(ump, MNT_WAIT); 976 if (error) { 977 fs->fs_clean = 0; 978 return (error); 979 } 980 } 981 vinvalbuf(ump->um_devvp, V_SAVE, NOCRED, td, 0, 0); 982 DROP_GIANT(); 983 g_topology_lock(); 984 g_wither_geom_close(ump->um_cp->geom, ENXIO); 985 g_topology_unlock(); 986 PICKUP_GIANT(); 987 vrele(ump->um_devvp); 988 free(fs->fs_csp, M_UFSMNT); 989 free(fs, M_UFSMNT); 990 free(ump, M_UFSMNT); 991 mp->mnt_data = (qaddr_t)0; 992 mp->mnt_flag &= ~MNT_LOCAL; 993 return (error); 994} 995 996/* 997 * Flush out all the files in a filesystem. 998 */ 999int 1000ffs_flushfiles(mp, flags, td) 1001 struct mount *mp; 1002 int flags; 1003 struct thread *td; 1004{ 1005 struct ufsmount *ump; 1006 int error; 1007 1008 ump = VFSTOUFS(mp); 1009#ifdef QUOTA 1010 if (mp->mnt_flag & MNT_QUOTA) { 1011 int i; 1012 error = vflush(mp, 0, SKIPSYSTEM|flags, td); 1013 if (error) 1014 return (error); 1015 for (i = 0; i < MAXQUOTAS; i++) { 1016 if (ump->um_quotas[i] == NULLVP) 1017 continue; 1018 quotaoff(td, mp, i); 1019 } 1020 /* 1021 * Here we fall through to vflush again to ensure 1022 * that we have gotten rid of all the system vnodes. 1023 */ 1024 } 1025#endif 1026 ASSERT_VOP_LOCKED(ump->um_devvp, "ffs_flushfiles"); 1027 if (ump->um_devvp->v_vflag & VV_COPYONWRITE) { 1028 if ((error = vflush(mp, 0, SKIPSYSTEM | flags, td)) != 0) 1029 return (error); 1030 ffs_snapshot_unmount(mp); 1031 /* 1032 * Here we fall through to vflush again to ensure 1033 * that we have gotten rid of all the system vnodes. 1034 */ 1035 } 1036 /* 1037 * Flush all the files. 1038 */ 1039 if ((error = vflush(mp, 0, flags, td)) != 0) 1040 return (error); 1041 /* 1042 * Flush filesystem metadata. 1043 */ 1044 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY, td); 1045 error = VOP_FSYNC(ump->um_devvp, td->td_ucred, MNT_WAIT, td); 1046 VOP_UNLOCK(ump->um_devvp, 0, td); 1047 return (error); 1048} 1049 1050/* 1051 * Get filesystem statistics. 1052 */ 1053int 1054ffs_statfs(mp, sbp, td) 1055 struct mount *mp; 1056 struct statfs *sbp; 1057 struct thread *td; 1058{ 1059 struct ufsmount *ump; 1060 struct fs *fs; 1061 1062 ump = VFSTOUFS(mp); 1063 fs = ump->um_fs; 1064 if (fs->fs_magic != FS_UFS1_MAGIC && fs->fs_magic != FS_UFS2_MAGIC) 1065 panic("ffs_statfs"); 1066 sbp->f_version = STATFS_VERSION; 1067 sbp->f_bsize = fs->fs_fsize; 1068 sbp->f_iosize = fs->fs_bsize; 1069 sbp->f_blocks = fs->fs_dsize; 1070 sbp->f_bfree = fs->fs_cstotal.cs_nbfree * fs->fs_frag + 1071 fs->fs_cstotal.cs_nffree + dbtofsb(fs, fs->fs_pendingblocks); 1072 sbp->f_bavail = freespace(fs, fs->fs_minfree) + 1073 dbtofsb(fs, fs->fs_pendingblocks); 1074 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO; 1075 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes; 1076 sbp->f_namemax = NAME_MAX; 1077 if (sbp != &mp->mnt_stat) { 1078 sbp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 1079 sbp->f_type = mp->mnt_vfc->vfc_typenum; 1080 sbp->f_syncwrites = mp->mnt_stat.f_syncwrites; 1081 sbp->f_asyncwrites = mp->mnt_stat.f_asyncwrites; 1082 sbp->f_syncreads = mp->mnt_stat.f_syncreads; 1083 sbp->f_asyncreads = mp->mnt_stat.f_asyncreads; 1084 sbp->f_owner = mp->mnt_stat.f_owner; 1085 sbp->f_fsid = mp->mnt_stat.f_fsid; 1086 bcopy((caddr_t)mp->mnt_stat.f_fstypename, 1087 (caddr_t)&sbp->f_fstypename[0], MFSNAMELEN); 1088 bcopy((caddr_t)mp->mnt_stat.f_mntonname, 1089 (caddr_t)&sbp->f_mntonname[0], MNAMELEN); 1090 bcopy((caddr_t)mp->mnt_stat.f_mntfromname, 1091 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN); 1092 } 1093 return (0); 1094} 1095 1096/* 1097 * Go through the disk queues to initiate sandbagged IO; 1098 * go through the inodes to write those that have been modified; 1099 * initiate the writing of the super block if it has been modified. 1100 * 1101 * Note: we are always called with the filesystem marked `MPBUSY'. 1102 */ 1103int 1104ffs_sync(mp, waitfor, cred, td) 1105 struct mount *mp; 1106 int waitfor; 1107 struct ucred *cred; 1108 struct thread *td; 1109{ 1110 struct vnode *nvp, *vp, *devvp; 1111 struct inode *ip; 1112 struct ufsmount *ump = VFSTOUFS(mp); 1113 struct fs *fs; 1114 int error, count, wait, lockreq, allerror = 0; 1115 struct bufobj *bo; 1116 1117 fs = ump->um_fs; 1118 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */ 1119 printf("fs = %s\n", fs->fs_fsmnt); 1120 panic("ffs_sync: rofs mod"); 1121 } 1122 /* 1123 * Write back each (modified) inode. 1124 */ 1125 wait = 0; 1126 lockreq = LK_EXCLUSIVE | LK_NOWAIT; 1127 if (waitfor == MNT_WAIT) { 1128 wait = 1; 1129 lockreq = LK_EXCLUSIVE; 1130 } 1131 lockreq |= LK_INTERLOCK; 1132 MNT_ILOCK(mp); 1133loop: 1134 MNT_VNODE_FOREACH(vp, mp, nvp) { 1135 /* 1136 * Depend on the mntvnode_slock to keep things stable enough 1137 * for a quick test. Since there might be hundreds of 1138 * thousands of vnodes, we cannot afford even a subroutine 1139 * call unless there's a good chance that we have work to do. 1140 */ 1141 VI_LOCK(vp); 1142 if (vp->v_iflag & VI_XLOCK) { 1143 VI_UNLOCK(vp); 1144 continue; 1145 } 1146 ip = VTOI(vp); 1147 if (vp->v_type == VNON || ((ip->i_flag & 1148 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 && 1149 vp->v_bufobj.bo_dirty.bv_cnt == 0)) { 1150 VI_UNLOCK(vp); 1151 continue; 1152 } 1153 MNT_IUNLOCK(mp); 1154 if ((error = vget(vp, lockreq, td)) != 0) { 1155 MNT_ILOCK(mp); 1156 if (error == ENOENT) 1157 goto loop; 1158 continue; 1159 } 1160 if ((error = VOP_FSYNC(vp, cred, waitfor, td)) != 0) 1161 allerror = error; 1162 VOP_UNLOCK(vp, 0, td); 1163 vrele(vp); 1164 MNT_ILOCK(mp); 1165 } 1166 MNT_IUNLOCK(mp); 1167 /* 1168 * Force stale filesystem control information to be flushed. 1169 */ 1170 if (waitfor == MNT_WAIT) { 1171 if ((error = softdep_flushworklist(ump->um_mountp, &count, td))) 1172 allerror = error; 1173 /* Flushed work items may create new vnodes to clean */ 1174 if (allerror == 0 && count) { 1175 MNT_ILOCK(mp); 1176 goto loop; 1177 } 1178 } 1179#ifdef QUOTA 1180 qsync(mp); 1181#endif 1182 devvp = ump->um_devvp; 1183 VI_LOCK(devvp); 1184 bo = &devvp->v_bufobj; 1185 if (waitfor != MNT_LAZY && 1186 (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0)) { 1187 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY | LK_INTERLOCK, td); 1188 if ((error = VOP_FSYNC(devvp, cred, waitfor, td)) != 0) 1189 allerror = error; 1190 VOP_UNLOCK(devvp, 0, td); 1191 if (allerror == 0 && waitfor == MNT_WAIT) { 1192 MNT_ILOCK(mp); 1193 goto loop; 1194 } 1195 } else 1196 VI_UNLOCK(devvp); 1197 /* 1198 * Write back modified superblock. 1199 */ 1200 if (fs->fs_fmod != 0 && (error = ffs_sbupdate(ump, waitfor)) != 0) 1201 allerror = error; 1202 return (allerror); 1203} 1204 1205int 1206ffs_vget(mp, ino, flags, vpp) 1207 struct mount *mp; 1208 ino_t ino; 1209 int flags; 1210 struct vnode **vpp; 1211{ 1212 struct thread *td = curthread; /* XXX */ 1213 struct fs *fs; 1214 struct inode *ip; 1215 struct ufsmount *ump; 1216 struct buf *bp; 1217 struct vnode *vp; 1218 struct cdev *dev; 1219 int error; 1220 1221 ump = VFSTOUFS(mp); 1222 dev = ump->um_dev; 1223 1224 /* 1225 * We do not lock vnode creation as it is believed to be too 1226 * expensive for such rare case as simultaneous creation of vnode 1227 * for same ino by different processes. We just allow them to race 1228 * and check later to decide who wins. Let the race begin! 1229 */ 1230 if ((error = ufs_ihashget(dev, ino, flags, vpp)) != 0) 1231 return (error); 1232 if (*vpp != NULL) 1233 return (0); 1234 1235 /* 1236 * If this MALLOC() is performed after the getnewvnode() 1237 * it might block, leaving a vnode with a NULL v_data to be 1238 * found by ffs_sync() if a sync happens to fire right then, 1239 * which will cause a panic because ffs_sync() blindly 1240 * dereferences vp->v_data (as well it should). 1241 */ 1242 ip = uma_zalloc(uma_inode, M_WAITOK); 1243 1244 /* Allocate a new vnode/inode. */ 1245 error = getnewvnode("ufs", mp, &ffs_vnodeops, &vp); 1246 if (error) { 1247 *vpp = NULL; 1248 uma_zfree(uma_inode, ip); 1249 return (error); 1250 } 1251 bzero((caddr_t)ip, sizeof(struct inode)); 1252 /* 1253 * FFS supports recursive locking. 1254 */ 1255 fs = ump->um_fs; 1256 vp->v_vnlock->lk_flags |= LK_CANRECURSE; 1257 vp->v_data = ip; 1258 vp->v_bufobj.bo_bsize = fs->fs_bsize; 1259 ip->i_vnode = vp; 1260 ip->i_ump = ump; 1261 ip->i_fs = fs; 1262 ip->i_dev = dev; 1263 ip->i_number = ino; 1264#ifdef QUOTA 1265 { 1266 int i; 1267 for (i = 0; i < MAXQUOTAS; i++) 1268 ip->i_dquot[i] = NODQUOT; 1269 } 1270#endif 1271 /* 1272 * Exclusively lock the vnode before adding to hash. Note, that we 1273 * must not release nor downgrade the lock (despite flags argument 1274 * says) till it is fully initialized. 1275 */ 1276 lockmgr(vp->v_vnlock, LK_EXCLUSIVE, (struct mtx *)0, td); 1277 1278 /* 1279 * Atomicaly (in terms of ufs_hash operations) check the hash for 1280 * duplicate of vnode being created and add it to the hash. If a 1281 * duplicate vnode was found, it will be vget()ed from hash for us. 1282 */ 1283 if ((error = ufs_ihashins(ip, flags, vpp)) != 0) { 1284 vput(vp); 1285 *vpp = NULL; 1286 return (error); 1287 } 1288 1289 /* We lost the race, then throw away our vnode and return existing */ 1290 if (*vpp != NULL) { 1291 vput(vp); 1292 return (0); 1293 } 1294 1295 /* Read in the disk contents for the inode, copy into the inode. */ 1296 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 1297 (int)fs->fs_bsize, NOCRED, &bp); 1298 if (error) { 1299 /* 1300 * The inode does not contain anything useful, so it would 1301 * be misleading to leave it on its hash chain. With mode 1302 * still zero, it will be unlinked and returned to the free 1303 * list by vput(). 1304 */ 1305 brelse(bp); 1306 vput(vp); 1307 *vpp = NULL; 1308 return (error); 1309 } 1310 if (ip->i_ump->um_fstype == UFS1) 1311 ip->i_din1 = uma_zalloc(uma_ufs1, M_WAITOK); 1312 else 1313 ip->i_din2 = uma_zalloc(uma_ufs2, M_WAITOK); 1314 ffs_load_inode(bp, ip, fs, ino); 1315 if (DOINGSOFTDEP(vp)) 1316 softdep_load_inodeblock(ip); 1317 else 1318 ip->i_effnlink = ip->i_nlink; 1319 bqrelse(bp); 1320 1321 /* 1322 * Initialize the vnode from the inode, check for aliases. 1323 * Note that the underlying vnode may have changed. 1324 */ 1325 error = ufs_vinit(mp, &ffs_fifoops, &vp); 1326 if (error) { 1327 vput(vp); 1328 *vpp = NULL; 1329 return (error); 1330 } 1331 /* 1332 * Finish inode initialization. 1333 */ 1334 VREF(ip->i_devvp); 1335 /* 1336 * Set up a generation number for this inode if it does not 1337 * already have one. This should only happen on old filesystems. 1338 */ 1339 if (ip->i_gen == 0) { 1340 ip->i_gen = arc4random() / 2 + 1; 1341 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 1342 ip->i_flag |= IN_MODIFIED; 1343 DIP_SET(ip, i_gen, ip->i_gen); 1344 } 1345 } 1346 /* 1347 * Ensure that uid and gid are correct. This is a temporary 1348 * fix until fsck has been changed to do the update. 1349 */ 1350 if (fs->fs_magic == FS_UFS1_MAGIC && /* XXX */ 1351 fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */ 1352 ip->i_uid = ip->i_din1->di_ouid; /* XXX */ 1353 ip->i_gid = ip->i_din1->di_ogid; /* XXX */ 1354 } /* XXX */ 1355 1356#ifdef MAC 1357 if ((mp->mnt_flag & MNT_MULTILABEL) && ip->i_mode) { 1358 /* 1359 * If this vnode is already allocated, and we're running 1360 * multi-label, attempt to perform a label association 1361 * from the extended attributes on the inode. 1362 */ 1363 error = mac_associate_vnode_extattr(mp, vp); 1364 if (error) { 1365 /* ufs_inactive will release ip->i_devvp ref. */ 1366 vput(vp); 1367 *vpp = NULL; 1368 return (error); 1369 } 1370 } 1371#endif 1372 1373 *vpp = vp; 1374 return (0); 1375} 1376 1377/* 1378 * File handle to vnode 1379 * 1380 * Have to be really careful about stale file handles: 1381 * - check that the inode number is valid 1382 * - call ffs_vget() to get the locked inode 1383 * - check for an unallocated inode (i_mode == 0) 1384 * - check that the given client host has export rights and return 1385 * those rights via. exflagsp and credanonp 1386 */ 1387int 1388ffs_fhtovp(mp, fhp, vpp) 1389 struct mount *mp; 1390 struct fid *fhp; 1391 struct vnode **vpp; 1392{ 1393 struct ufid *ufhp; 1394 struct fs *fs; 1395 1396 ufhp = (struct ufid *)fhp; 1397 fs = VFSTOUFS(mp)->um_fs; 1398 if (ufhp->ufid_ino < ROOTINO || 1399 ufhp->ufid_ino >= fs->fs_ncg * fs->fs_ipg) 1400 return (ESTALE); 1401 return (ufs_fhtovp(mp, ufhp, vpp)); 1402} 1403 1404/* 1405 * Vnode pointer to File handle 1406 */ 1407/* ARGSUSED */ 1408int 1409ffs_vptofh(vp, fhp) 1410 struct vnode *vp; 1411 struct fid *fhp; 1412{ 1413 struct inode *ip; 1414 struct ufid *ufhp; 1415 1416 ip = VTOI(vp); 1417 ufhp = (struct ufid *)fhp; 1418 ufhp->ufid_len = sizeof(struct ufid); 1419 ufhp->ufid_ino = ip->i_number; 1420 ufhp->ufid_gen = ip->i_gen; 1421 return (0); 1422} 1423 1424/* 1425 * Initialize the filesystem. 1426 */ 1427static int 1428ffs_init(vfsp) 1429 struct vfsconf *vfsp; 1430{ 1431 1432 softdep_initialize(); 1433 return (ufs_init(vfsp)); 1434} 1435 1436/* 1437 * Undo the work of ffs_init(). 1438 */ 1439static int 1440ffs_uninit(vfsp) 1441 struct vfsconf *vfsp; 1442{ 1443 int ret; 1444 1445 ret = ufs_uninit(vfsp); 1446 softdep_uninitialize(); 1447 return (ret); 1448} 1449 1450/* 1451 * Write a superblock and associated information back to disk. 1452 */ 1453static int 1454ffs_sbupdate(mp, waitfor) 1455 struct ufsmount *mp; 1456 int waitfor; 1457{ 1458 struct fs *fs = mp->um_fs; 1459 struct buf *bp; 1460 int blks; 1461 void *space; 1462 int i, size, error, allerror = 0; 1463 1464 if (fs->fs_ronly == 1 && 1465 (mp->um_mountp->mnt_flag & (MNT_RDONLY | MNT_UPDATE)) != 1466 (MNT_RDONLY | MNT_UPDATE)) 1467 panic("ffs_sbupdate: write read-only filesystem"); 1468 /* 1469 * First write back the summary information. 1470 */ 1471 blks = howmany(fs->fs_cssize, fs->fs_fsize); 1472 space = fs->fs_csp; 1473 for (i = 0; i < blks; i += fs->fs_frag) { 1474 size = fs->fs_bsize; 1475 if (i + fs->fs_frag > blks) 1476 size = (blks - i) * fs->fs_fsize; 1477 bp = getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i), 1478 size, 0, 0, 0); 1479 bcopy(space, bp->b_data, (u_int)size); 1480 space = (char *)space + size; 1481 if (waitfor != MNT_WAIT) 1482 bawrite(bp); 1483 else if ((error = bwrite(bp)) != 0) 1484 allerror = error; 1485 } 1486 /* 1487 * Now write back the superblock itself. If any errors occurred 1488 * up to this point, then fail so that the superblock avoids 1489 * being written out as clean. 1490 */ 1491 if (allerror) 1492 return (allerror); 1493 if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_sblockloc != SBLOCK_UFS1 && 1494 (fs->fs_flags & FS_FLAGS_UPDATED) == 0) { 1495 printf("%s: correcting fs_sblockloc from %jd to %d\n", 1496 fs->fs_fsmnt, fs->fs_sblockloc, SBLOCK_UFS1); 1497 fs->fs_sblockloc = SBLOCK_UFS1; 1498 } 1499 if (fs->fs_magic == FS_UFS2_MAGIC && fs->fs_sblockloc != SBLOCK_UFS2 && 1500 (fs->fs_flags & FS_FLAGS_UPDATED) == 0) { 1501 printf("%s: correcting fs_sblockloc from %jd to %d\n", 1502 fs->fs_fsmnt, fs->fs_sblockloc, SBLOCK_UFS2); 1503 fs->fs_sblockloc = SBLOCK_UFS2; 1504 } 1505 bp = getblk(mp->um_devvp, btodb(fs->fs_sblockloc), (int)fs->fs_sbsize, 1506 0, 0, 0); 1507 fs->fs_fmod = 0; 1508 fs->fs_time = time_second; 1509 bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize); 1510 ffs_oldfscompat_write((struct fs *)bp->b_data, mp); 1511 if (waitfor != MNT_WAIT) 1512 bawrite(bp); 1513 else if ((error = bwrite(bp)) != 0) 1514 allerror = error; 1515 return (allerror); 1516} 1517 1518static int 1519ffs_extattrctl(struct mount *mp, int cmd, struct vnode *filename_vp, 1520 int attrnamespace, const char *attrname, struct thread *td) 1521{ 1522 1523#ifdef UFS_EXTATTR 1524 return (ufs_extattrctl(mp, cmd, filename_vp, attrnamespace, 1525 attrname, td)); 1526#else 1527 return (vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, 1528 attrname, td)); 1529#endif 1530} 1531 1532static void 1533ffs_ifree(struct ufsmount *ump, struct inode *ip) 1534{ 1535 1536 if (ump->um_fstype == UFS1 && ip->i_din1 != NULL) 1537 uma_zfree(uma_ufs1, ip->i_din1); 1538 else if (ip->i_din2 != NULL) 1539 uma_zfree(uma_ufs2, ip->i_din2); 1540 uma_zfree(uma_inode, ip); 1541} 1542 1543static void 1544ffs_geom_strategy(struct bufobj *bo, struct buf *bp) 1545{ 1546 1547#ifdef SOFTUPDATES 1548 if (bp->b_iocmd == BIO_WRITE && softdep_disk_prewrite(bp)) 1549 return; 1550#endif 1551 g_vfs_strategy(bo, bp); 1552} 1553