ffs_vfsops.c revision 1.310
1 /* $NetBSD: ffs_vfsops.c,v 1.310 2015/02/14 07:11:34 maxv Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Wasabi Systems, Inc, and by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1991, 1993, 1994 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ffs_vfsops.c 8.31 (Berkeley) 5/20/95 61 */ 62 63 #include <sys/cdefs.h> 64 __KERNEL_RCSID(0, "$NetBSD: ffs_vfsops.c,v 1.310 2015/02/14 07:11:34 maxv Exp $"); 65 66 #if defined(_KERNEL_OPT) 67 #include "opt_ffs.h" 68 #include "opt_quota.h" 69 #include "opt_wapbl.h" 70 #endif 71 72 #include <sys/param.h> 73 #include <sys/systm.h> 74 #include <sys/namei.h> 75 #include <sys/proc.h> 76 #include <sys/kernel.h> 77 #include <sys/vnode.h> 78 #include <sys/socket.h> 79 #include <sys/mount.h> 80 #include <sys/buf.h> 81 #include <sys/device.h> 82 #include <sys/disk.h> 83 #include <sys/mbuf.h> 84 #include <sys/file.h> 85 #include <sys/disklabel.h> 86 #include <sys/ioctl.h> 87 #include <sys/errno.h> 88 #include <sys/kmem.h> 89 #include <sys/pool.h> 90 #include <sys/lock.h> 91 #include <sys/sysctl.h> 92 #include <sys/conf.h> 93 #include <sys/kauth.h> 94 #include <sys/wapbl.h> 95 #include <sys/fstrans.h> 96 #include <sys/module.h> 97 98 #include <miscfs/genfs/genfs.h> 99 #include <miscfs/specfs/specdev.h> 100 101 #include <ufs/ufs/quota.h> 102 #include <ufs/ufs/ufsmount.h> 103 #include <ufs/ufs/inode.h> 104 #include <ufs/ufs/dir.h> 105 #include <ufs/ufs/ufs_extern.h> 106 #include <ufs/ufs/ufs_bswap.h> 107 #include <ufs/ufs/ufs_wapbl.h> 108 109 #include <ufs/ffs/fs.h> 110 #include <ufs/ffs/ffs_extern.h> 111 112 MODULE(MODULE_CLASS_VFS, ffs, NULL); 113 114 static int 115 ffs_vfs_fsync(vnode_t *, int); 116 117 static int 118 ffs_superblock_validate(struct fs *fs, u_int32_t fs_sbsize, int32_t fs_bsize); 119 120 static struct sysctllog *ffs_sysctl_log; 121 122 static kauth_listener_t ffs_snapshot_listener; 123 124 /* how many times ffs_init() was called */ 125 int ffs_initcount = 0; 126 127 #ifdef DEBUG_FFS_MOUNT 128 #define DPRINTF(a) printf a 129 #else 130 #define DPRINTF(a) do {} while (/*CONSTCOND*/0) 131 #endif 132 133 extern const struct vnodeopv_desc ffs_vnodeop_opv_desc; 134 extern const struct vnodeopv_desc ffs_specop_opv_desc; 135 extern const struct vnodeopv_desc ffs_fifoop_opv_desc; 136 137 const struct vnodeopv_desc * const ffs_vnodeopv_descs[] = { 138 &ffs_vnodeop_opv_desc, 139 &ffs_specop_opv_desc, 140 &ffs_fifoop_opv_desc, 141 NULL, 142 }; 143 144 struct vfsops ffs_vfsops = { 145 .vfs_name = MOUNT_FFS, 146 .vfs_min_mount_data = sizeof (struct ufs_args), 147 .vfs_mount = ffs_mount, 148 .vfs_start = ufs_start, 149 .vfs_unmount = ffs_unmount, 150 .vfs_root = ufs_root, 151 .vfs_quotactl = ufs_quotactl, 152 .vfs_statvfs = ffs_statvfs, 153 .vfs_sync = ffs_sync, 154 .vfs_vget = ufs_vget, 155 .vfs_loadvnode = ffs_loadvnode, 156 .vfs_fhtovp = ffs_fhtovp, 157 .vfs_vptofh = ffs_vptofh, 158 .vfs_init = ffs_init, 159 .vfs_reinit = ffs_reinit, 160 .vfs_done = ffs_done, 161 .vfs_mountroot = ffs_mountroot, 162 .vfs_snapshot = ffs_snapshot, 163 .vfs_extattrctl = ffs_extattrctl, 164 .vfs_suspendctl = ffs_suspendctl, 165 .vfs_renamelock_enter = genfs_renamelock_enter, 166 .vfs_renamelock_exit = genfs_renamelock_exit, 167 .vfs_fsync = ffs_vfs_fsync, 168 .vfs_opv_descs = ffs_vnodeopv_descs 169 }; 170 171 static const struct genfs_ops ffs_genfsops = { 172 .gop_size = ffs_gop_size, 173 .gop_alloc = ufs_gop_alloc, 174 .gop_write = genfs_gop_write, 175 .gop_markupdate = ufs_gop_markupdate, 176 }; 177 178 static const struct ufs_ops ffs_ufsops = { 179 .uo_itimes = ffs_itimes, 180 .uo_update = ffs_update, 181 .uo_truncate = ffs_truncate, 182 .uo_valloc = ffs_valloc, 183 .uo_vfree = ffs_vfree, 184 .uo_balloc = ffs_balloc, 185 .uo_snapgone = ffs_snapgone, 186 }; 187 188 static int 189 ffs_snapshot_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 190 void *arg0, void *arg1, void *arg2, void *arg3) 191 { 192 vnode_t *vp = arg2; 193 int result = KAUTH_RESULT_DEFER; 194 195 if (action != KAUTH_SYSTEM_FS_SNAPSHOT) 196 return result; 197 198 if (VTOI(vp)->i_uid == kauth_cred_geteuid(cred)) 199 result = KAUTH_RESULT_ALLOW; 200 201 return result; 202 } 203 204 static int 205 ffs_modcmd(modcmd_t cmd, void *arg) 206 { 207 int error; 208 209 #if 0 210 extern int doasyncfree; 211 #endif 212 #ifdef UFS_EXTATTR 213 extern int ufs_extattr_autocreate; 214 #endif 215 extern int ffs_log_changeopt; 216 217 switch (cmd) { 218 case MODULE_CMD_INIT: 219 error = vfs_attach(&ffs_vfsops); 220 if (error != 0) 221 break; 222 223 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 224 CTLFLAG_PERMANENT, 225 CTLTYPE_NODE, "ffs", 226 SYSCTL_DESCR("Berkeley Fast File System"), 227 NULL, 0, NULL, 0, 228 CTL_VFS, 1, CTL_EOL); 229 /* 230 * @@@ should we even bother with these first three? 231 */ 232 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 233 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 234 CTLTYPE_INT, "doclusterread", NULL, 235 sysctl_notavail, 0, NULL, 0, 236 CTL_VFS, 1, FFS_CLUSTERREAD, CTL_EOL); 237 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 238 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 239 CTLTYPE_INT, "doclusterwrite", NULL, 240 sysctl_notavail, 0, NULL, 0, 241 CTL_VFS, 1, FFS_CLUSTERWRITE, CTL_EOL); 242 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 243 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 244 CTLTYPE_INT, "doreallocblks", NULL, 245 sysctl_notavail, 0, NULL, 0, 246 CTL_VFS, 1, FFS_REALLOCBLKS, CTL_EOL); 247 #if 0 248 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 249 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 250 CTLTYPE_INT, "doasyncfree", 251 SYSCTL_DESCR("Release dirty blocks asynchronously"), 252 NULL, 0, &doasyncfree, 0, 253 CTL_VFS, 1, FFS_ASYNCFREE, CTL_EOL); 254 #endif 255 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 256 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 257 CTLTYPE_INT, "log_changeopt", 258 SYSCTL_DESCR("Log changes in optimization strategy"), 259 NULL, 0, &ffs_log_changeopt, 0, 260 CTL_VFS, 1, FFS_LOG_CHANGEOPT, CTL_EOL); 261 #ifdef UFS_EXTATTR 262 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 263 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 264 CTLTYPE_INT, "extattr_autocreate", 265 SYSCTL_DESCR("Size of attribute for " 266 "backing file autocreation"), 267 NULL, 0, &ufs_extattr_autocreate, 0, 268 CTL_VFS, 1, FFS_EXTATTR_AUTOCREATE, CTL_EOL); 269 270 #endif /* UFS_EXTATTR */ 271 272 ffs_snapshot_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM, 273 ffs_snapshot_cb, NULL); 274 if (ffs_snapshot_listener == NULL) 275 printf("ffs_modcmd: can't listen on system scope.\n"); 276 277 break; 278 case MODULE_CMD_FINI: 279 error = vfs_detach(&ffs_vfsops); 280 if (error != 0) 281 break; 282 sysctl_teardown(&ffs_sysctl_log); 283 if (ffs_snapshot_listener != NULL) 284 kauth_unlisten_scope(ffs_snapshot_listener); 285 break; 286 default: 287 error = ENOTTY; 288 break; 289 } 290 291 return (error); 292 } 293 294 pool_cache_t ffs_inode_cache; 295 pool_cache_t ffs_dinode1_cache; 296 pool_cache_t ffs_dinode2_cache; 297 298 static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t); 299 static void ffs_oldfscompat_write(struct fs *, struct ufsmount *); 300 301 /* 302 * Called by main() when ffs is going to be mounted as root. 303 */ 304 305 int 306 ffs_mountroot(void) 307 { 308 struct fs *fs; 309 struct mount *mp; 310 struct lwp *l = curlwp; /* XXX */ 311 struct ufsmount *ump; 312 int error; 313 314 if (device_class(root_device) != DV_DISK) 315 return (ENODEV); 316 317 if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp))) { 318 vrele(rootvp); 319 return (error); 320 } 321 322 /* 323 * We always need to be able to mount the root file system. 324 */ 325 mp->mnt_flag |= MNT_FORCE; 326 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) { 327 vfs_unbusy(mp, false, NULL); 328 vfs_destroy(mp); 329 return (error); 330 } 331 mp->mnt_flag &= ~MNT_FORCE; 332 mountlist_append(mp); 333 ump = VFSTOUFS(mp); 334 fs = ump->um_fs; 335 memset(fs->fs_fsmnt, 0, sizeof(fs->fs_fsmnt)); 336 (void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0); 337 (void)ffs_statvfs(mp, &mp->mnt_stat); 338 vfs_unbusy(mp, false, NULL); 339 setrootfstime((time_t)fs->fs_time); 340 return (0); 341 } 342 343 /* 344 * VFS Operations. 345 * 346 * mount system call 347 */ 348 int 349 ffs_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 350 { 351 struct lwp *l = curlwp; 352 struct vnode *devvp = NULL; 353 struct ufs_args *args = data; 354 struct ufsmount *ump = NULL; 355 struct fs *fs; 356 int error = 0, flags, update; 357 mode_t accessmode; 358 359 if (args == NULL) { 360 DPRINTF(("%s: NULL args\n", __func__)); 361 return EINVAL; 362 } 363 if (*data_len < sizeof(*args)) { 364 DPRINTF(("%s: bad size args %zu != %zu\n", 365 __func__, *data_len, sizeof(*args))); 366 return EINVAL; 367 } 368 369 if (mp->mnt_flag & MNT_GETARGS) { 370 ump = VFSTOUFS(mp); 371 if (ump == NULL) { 372 DPRINTF(("%s: no ump\n", __func__)); 373 return EIO; 374 } 375 args->fspec = NULL; 376 *data_len = sizeof *args; 377 return 0; 378 } 379 380 update = mp->mnt_flag & MNT_UPDATE; 381 382 /* Check arguments */ 383 if (args->fspec != NULL) { 384 /* 385 * Look up the name and verify that it's sane. 386 */ 387 error = namei_simple_user(args->fspec, 388 NSM_FOLLOW_NOEMULROOT, &devvp); 389 if (error != 0) { 390 DPRINTF(("%s: namei_simple_user %d\n", __func__, 391 error)); 392 return error; 393 } 394 395 if (!update) { 396 /* 397 * Be sure this is a valid block device 398 */ 399 if (devvp->v_type != VBLK) { 400 DPRINTF(("%s: non block device %d\n", 401 __func__, devvp->v_type)); 402 error = ENOTBLK; 403 } else if (bdevsw_lookup(devvp->v_rdev) == NULL) { 404 DPRINTF(("%s: can't find block device 0x%jx\n", 405 __func__, devvp->v_rdev)); 406 error = ENXIO; 407 } 408 } else { 409 /* 410 * Be sure we're still naming the same device 411 * used for our initial mount 412 */ 413 ump = VFSTOUFS(mp); 414 if (devvp != ump->um_devvp) { 415 if (devvp->v_rdev != ump->um_devvp->v_rdev) { 416 DPRINTF(("%s: wrong device 0x%jx" 417 " != 0x%jx\n", __func__, 418 (uintmax_t)devvp->v_rdev, 419 (uintmax_t)ump->um_devvp->v_rdev)); 420 error = EINVAL; 421 } else { 422 vrele(devvp); 423 devvp = ump->um_devvp; 424 vref(devvp); 425 } 426 } 427 } 428 } else { 429 if (!update) { 430 /* New mounts must have a filename for the device */ 431 DPRINTF(("%s: no filename for mount\n", __func__)); 432 return EINVAL; 433 } else { 434 /* Use the extant mount */ 435 ump = VFSTOUFS(mp); 436 devvp = ump->um_devvp; 437 vref(devvp); 438 } 439 } 440 441 /* 442 * If mount by non-root, then verify that user has necessary 443 * permissions on the device. 444 * 445 * Permission to update a mount is checked higher, so here we presume 446 * updating the mount is okay (for example, as far as securelevel goes) 447 * which leaves us with the normal check. 448 */ 449 if (error == 0) { 450 accessmode = VREAD; 451 if (update ? 452 (mp->mnt_iflag & IMNT_WANTRDWR) != 0 : 453 (mp->mnt_flag & MNT_RDONLY) == 0) 454 accessmode |= VWRITE; 455 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 456 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 457 KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, 458 KAUTH_ARG(accessmode)); 459 if (error) { 460 DPRINTF(("%s: kauth %d\n", __func__, error)); 461 } 462 VOP_UNLOCK(devvp); 463 } 464 465 if (error) { 466 vrele(devvp); 467 return (error); 468 } 469 470 #ifdef WAPBL 471 /* WAPBL can only be enabled on a r/w mount. */ 472 if ((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) { 473 mp->mnt_flag &= ~MNT_LOG; 474 } 475 #else /* !WAPBL */ 476 mp->mnt_flag &= ~MNT_LOG; 477 #endif /* !WAPBL */ 478 479 if (!update) { 480 int xflags; 481 482 if (mp->mnt_flag & MNT_RDONLY) 483 xflags = FREAD; 484 else 485 xflags = FREAD | FWRITE; 486 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 487 error = VOP_OPEN(devvp, xflags, FSCRED); 488 VOP_UNLOCK(devvp); 489 if (error) { 490 DPRINTF(("%s: VOP_OPEN %d\n", __func__, error)); 491 goto fail; 492 } 493 error = ffs_mountfs(devvp, mp, l); 494 if (error) { 495 DPRINTF(("%s: ffs_mountfs %d\n", __func__, error)); 496 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 497 (void)VOP_CLOSE(devvp, xflags, NOCRED); 498 VOP_UNLOCK(devvp); 499 goto fail; 500 } 501 502 ump = VFSTOUFS(mp); 503 fs = ump->um_fs; 504 } else { 505 /* 506 * Update the mount. 507 */ 508 509 /* 510 * The initial mount got a reference on this 511 * device, so drop the one obtained via 512 * namei(), above. 513 */ 514 vrele(devvp); 515 516 ump = VFSTOUFS(mp); 517 fs = ump->um_fs; 518 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 519 /* 520 * Changing from r/w to r/o 521 */ 522 flags = WRITECLOSE; 523 if (mp->mnt_flag & MNT_FORCE) 524 flags |= FORCECLOSE; 525 error = ffs_flushfiles(mp, flags, l); 526 if (error == 0) 527 error = UFS_WAPBL_BEGIN(mp); 528 if (error == 0 && 529 ffs_cgupdate(ump, MNT_WAIT) == 0 && 530 fs->fs_clean & FS_WASCLEAN) { 531 if (mp->mnt_flag & MNT_SOFTDEP) 532 fs->fs_flags &= ~FS_DOSOFTDEP; 533 fs->fs_clean = FS_ISCLEAN; 534 (void) ffs_sbupdate(ump, MNT_WAIT); 535 } 536 if (error) { 537 DPRINTF(("%s: wapbl %d\n", __func__, error)); 538 return error; 539 } 540 UFS_WAPBL_END(mp); 541 } 542 543 #ifdef WAPBL 544 if ((mp->mnt_flag & MNT_LOG) == 0) { 545 error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE); 546 if (error) { 547 DPRINTF(("%s: ffs_wapbl_stop %d\n", 548 __func__, error)); 549 return error; 550 } 551 } 552 #endif /* WAPBL */ 553 554 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 555 /* 556 * Finish change from r/w to r/o 557 */ 558 fs->fs_ronly = 1; 559 fs->fs_fmod = 0; 560 } 561 562 if (mp->mnt_flag & MNT_RELOAD) { 563 error = ffs_reload(mp, l->l_cred, l); 564 if (error) { 565 DPRINTF(("%s: ffs_reload %d\n", 566 __func__, error)); 567 return error; 568 } 569 } 570 571 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) { 572 /* 573 * Changing from read-only to read/write 574 */ 575 #ifndef QUOTA2 576 if (fs->fs_flags & FS_DOQUOTA2) { 577 ump->um_flags |= UFS_QUOTA2; 578 uprintf("%s: options QUOTA2 not enabled%s\n", 579 mp->mnt_stat.f_mntonname, 580 (mp->mnt_flag & MNT_FORCE) ? "" : 581 ", not mounting"); 582 DPRINTF(("%s: ffs_quota2 %d\n", 583 __func__, EINVAL)); 584 return EINVAL; 585 } 586 #endif 587 fs->fs_ronly = 0; 588 fs->fs_clean <<= 1; 589 fs->fs_fmod = 1; 590 #ifdef WAPBL 591 if (fs->fs_flags & FS_DOWAPBL) { 592 const char *nm = mp->mnt_stat.f_mntonname; 593 if (!mp->mnt_wapbl_replay) { 594 printf("%s: log corrupted;" 595 " replay cancelled\n", nm); 596 return EFTYPE; 597 } 598 printf("%s: replaying log to disk\n", nm); 599 error = wapbl_replay_write(mp->mnt_wapbl_replay, 600 devvp); 601 if (error) { 602 DPRINTF(( 603 "%s: %s: wapbl_replay_write %d\n", 604 __func__, nm, error)); 605 return error; 606 } 607 wapbl_replay_stop(mp->mnt_wapbl_replay); 608 fs->fs_clean = FS_WASCLEAN; 609 } 610 #endif /* WAPBL */ 611 if (fs->fs_snapinum[0] != 0) 612 ffs_snapshot_mount(mp); 613 } 614 615 #ifdef WAPBL 616 error = ffs_wapbl_start(mp); 617 if (error) { 618 DPRINTF(("%s: ffs_wapbl_start %d\n", 619 __func__, error)); 620 return error; 621 } 622 #endif /* WAPBL */ 623 624 #ifdef QUOTA2 625 if (!fs->fs_ronly) { 626 error = ffs_quota2_mount(mp); 627 if (error) { 628 DPRINTF(("%s: ffs_quota2_mount %d\n", 629 __func__, error)); 630 return error; 631 } 632 } 633 #endif 634 635 if ((mp->mnt_flag & MNT_DISCARD) && !(ump->um_discarddata)) 636 ump->um_discarddata = ffs_discard_init(devvp, fs); 637 638 if (args->fspec == NULL) 639 return 0; 640 } 641 642 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, 643 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l); 644 if (error == 0) 645 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, 646 sizeof(fs->fs_fsmnt)); 647 else { 648 DPRINTF(("%s: set_statvfs_info %d\n", __func__, error)); 649 } 650 fs->fs_flags &= ~FS_DOSOFTDEP; 651 if (fs->fs_fmod != 0) { /* XXX */ 652 int err; 653 654 fs->fs_fmod = 0; 655 if (fs->fs_clean & FS_WASCLEAN) 656 fs->fs_time = time_second; 657 else { 658 printf("%s: file system not clean (fs_clean=%#x); " 659 "please fsck(8)\n", mp->mnt_stat.f_mntfromname, 660 fs->fs_clean); 661 printf("%s: lost blocks %" PRId64 " files %d\n", 662 mp->mnt_stat.f_mntfromname, fs->fs_pendingblocks, 663 fs->fs_pendinginodes); 664 } 665 err = UFS_WAPBL_BEGIN(mp); 666 if (err == 0) { 667 (void) ffs_cgupdate(ump, MNT_WAIT); 668 UFS_WAPBL_END(mp); 669 } 670 } 671 if ((mp->mnt_flag & MNT_SOFTDEP) != 0) { 672 printf("%s: `-o softdep' is no longer supported, " 673 "consider `-o log'\n", mp->mnt_stat.f_mntfromname); 674 mp->mnt_flag &= ~MNT_SOFTDEP; 675 } 676 677 return (error); 678 679 fail: 680 vrele(devvp); 681 return (error); 682 } 683 684 /* 685 * Reload all incore data for a filesystem (used after running fsck on 686 * the root filesystem and finding things to fix). The filesystem must 687 * be mounted read-only. 688 * 689 * Things to do to update the mount: 690 * 1) invalidate all cached meta-data. 691 * 2) re-read superblock from disk. 692 * 3) re-read summary information from disk. 693 * 4) invalidate all inactive vnodes. 694 * 5) invalidate all cached file data. 695 * 6) re-read inode data for all active vnodes. 696 */ 697 int 698 ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l) 699 { 700 struct vnode *vp, *devvp; 701 struct inode *ip; 702 void *space; 703 struct buf *bp; 704 struct fs *fs, *newfs; 705 struct dkwedge_info dkw; 706 int i, bsize, blks, error; 707 int32_t *lp; 708 struct ufsmount *ump; 709 daddr_t sblockloc; 710 struct vnode_iterator *marker; 711 712 if ((mp->mnt_flag & MNT_RDONLY) == 0) 713 return (EINVAL); 714 715 ump = VFSTOUFS(mp); 716 /* 717 * Step 1: invalidate all cached meta-data. 718 */ 719 devvp = ump->um_devvp; 720 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 721 error = vinvalbuf(devvp, 0, cred, l, 0, 0); 722 VOP_UNLOCK(devvp); 723 if (error) 724 panic("ffs_reload: dirty1"); 725 /* 726 * Step 2: re-read superblock from disk. 727 */ 728 fs = ump->um_fs; 729 730 /* XXX we don't handle possibility that superblock moved. */ 731 error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs->fs_sbsize, 732 NOCRED, 0, &bp); 733 if (error) { 734 return (error); 735 } 736 newfs = kmem_alloc(fs->fs_sbsize, KM_SLEEP); 737 memcpy(newfs, bp->b_data, fs->fs_sbsize); 738 #ifdef FFS_EI 739 if (ump->um_flags & UFS_NEEDSWAP) { 740 ffs_sb_swap((struct fs*)bp->b_data, newfs); 741 fs->fs_flags |= FS_SWAPPED; 742 } else 743 #endif 744 fs->fs_flags &= ~FS_SWAPPED; 745 746 if ((newfs->fs_magic != FS_UFS1_MAGIC && 747 newfs->fs_magic != FS_UFS2_MAGIC)) { 748 brelse(bp, 0); 749 kmem_free(newfs, fs->fs_sbsize); 750 return (EIO); /* XXX needs translation */ 751 } 752 if (!ffs_superblock_validate(newfs, newfs->fs_sbsize, newfs->fs_bsize)) { 753 brelse(bp, 0); 754 kmem_free(newfs, fs->fs_sbsize); 755 return (EINVAL); 756 } 757 758 /* Store off old fs_sblockloc for fs_oldfscompat_read. */ 759 sblockloc = fs->fs_sblockloc; 760 /* 761 * Copy pointer fields back into superblock before copying in XXX 762 * new superblock. These should really be in the ufsmount. XXX 763 * Note that important parameters (eg fs_ncg) are unchanged. 764 */ 765 newfs->fs_csp = fs->fs_csp; 766 newfs->fs_maxcluster = fs->fs_maxcluster; 767 newfs->fs_contigdirs = fs->fs_contigdirs; 768 newfs->fs_ronly = fs->fs_ronly; 769 newfs->fs_active = fs->fs_active; 770 memcpy(fs, newfs, (u_int)fs->fs_sbsize); 771 brelse(bp, 0); 772 kmem_free(newfs, fs->fs_sbsize); 773 774 /* Recheck for apple UFS filesystem */ 775 ump->um_flags &= ~UFS_ISAPPLEUFS; 776 /* First check to see if this is tagged as an Apple UFS filesystem 777 * in the disklabel 778 */ 779 if (getdiskinfo(devvp, &dkw) == 0 && 780 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0) 781 ump->um_flags |= UFS_ISAPPLEUFS; 782 #ifdef APPLE_UFS 783 else { 784 /* Manually look for an apple ufs label, and if a valid one 785 * is found, then treat it like an Apple UFS filesystem anyway 786 * 787 * EINVAL is most probably a blocksize or alignment problem, 788 * it is unlikely that this is an Apple UFS filesystem then. 789 */ 790 error = bread(devvp, 791 (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE), 792 APPLEUFS_LABEL_SIZE, cred, 0, &bp); 793 if (error && error != EINVAL) { 794 return error; 795 } 796 if (error == 0) { 797 error = ffs_appleufs_validate(fs->fs_fsmnt, 798 (struct appleufslabel *)bp->b_data, NULL); 799 if (error == 0) 800 ump->um_flags |= UFS_ISAPPLEUFS; 801 brelse(bp, 0); 802 } 803 bp = NULL; 804 } 805 #else 806 if (ump->um_flags & UFS_ISAPPLEUFS) 807 return (EIO); 808 #endif 809 810 if (UFS_MPISAPPLEUFS(ump)) { 811 /* see comment about NeXT below */ 812 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN; 813 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ; 814 mp->mnt_iflag |= IMNT_DTYPE; 815 } else { 816 ump->um_maxsymlinklen = fs->fs_maxsymlinklen; 817 ump->um_dirblksiz = UFS_DIRBLKSIZ; 818 if (ump->um_maxsymlinklen > 0) 819 mp->mnt_iflag |= IMNT_DTYPE; 820 else 821 mp->mnt_iflag &= ~IMNT_DTYPE; 822 } 823 ffs_oldfscompat_read(fs, ump, sblockloc); 824 825 mutex_enter(&ump->um_lock); 826 ump->um_maxfilesize = fs->fs_maxfilesize; 827 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) { 828 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n", 829 mp->mnt_stat.f_mntonname, fs->fs_flags, 830 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 831 if ((mp->mnt_flag & MNT_FORCE) == 0) { 832 mutex_exit(&ump->um_lock); 833 return (EINVAL); 834 } 835 } 836 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 837 fs->fs_pendingblocks = 0; 838 fs->fs_pendinginodes = 0; 839 } 840 mutex_exit(&ump->um_lock); 841 842 ffs_statvfs(mp, &mp->mnt_stat); 843 /* 844 * Step 3: re-read summary information from disk. 845 */ 846 blks = howmany(fs->fs_cssize, fs->fs_fsize); 847 space = fs->fs_csp; 848 for (i = 0; i < blks; i += fs->fs_frag) { 849 bsize = fs->fs_bsize; 850 if (i + fs->fs_frag > blks) 851 bsize = (blks - i) * fs->fs_fsize; 852 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize, 853 NOCRED, 0, &bp); 854 if (error) { 855 return (error); 856 } 857 #ifdef FFS_EI 858 if (UFS_FSNEEDSWAP(fs)) 859 ffs_csum_swap((struct csum *)bp->b_data, 860 (struct csum *)space, bsize); 861 else 862 #endif 863 memcpy(space, bp->b_data, (size_t)bsize); 864 space = (char *)space + bsize; 865 brelse(bp, 0); 866 } 867 /* 868 * We no longer know anything about clusters per cylinder group. 869 */ 870 if (fs->fs_contigsumsize > 0) { 871 lp = fs->fs_maxcluster; 872 for (i = 0; i < fs->fs_ncg; i++) 873 *lp++ = fs->fs_contigsumsize; 874 } 875 876 vfs_vnode_iterator_init(mp, &marker); 877 while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) { 878 /* 879 * Step 4: invalidate all inactive vnodes. 880 */ 881 if (vrecycle(vp)) 882 continue; 883 /* 884 * Step 5: invalidate all cached file data. 885 */ 886 if (vn_lock(vp, LK_EXCLUSIVE)) { 887 vrele(vp); 888 continue; 889 } 890 if (vinvalbuf(vp, 0, cred, l, 0, 0)) 891 panic("ffs_reload: dirty2"); 892 /* 893 * Step 6: re-read inode data for all active vnodes. 894 */ 895 ip = VTOI(vp); 896 error = bread(devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)), 897 (int)fs->fs_bsize, NOCRED, 0, &bp); 898 if (error) { 899 vput(vp); 900 break; 901 } 902 ffs_load_inode(bp, ip, fs, ip->i_number); 903 brelse(bp, 0); 904 vput(vp); 905 } 906 vfs_vnode_iterator_destroy(marker); 907 return (error); 908 } 909 910 /* 911 * Possible superblock locations ordered from most to least likely. 912 */ 913 static const int sblock_try[] = SBLOCKSEARCH; 914 915 916 static int 917 ffs_superblock_validate(struct fs *fs, u_int32_t fs_sbsize, int32_t fs_bsize) 918 { 919 /* Check the superblock size */ 920 if (fs_sbsize > SBLOCKSIZE || fs_sbsize < sizeof(struct fs)) 921 return 0; 922 923 /* Check the file system blocksize */ 924 if (fs_bsize > MAXBSIZE || fs_bsize < MINBSIZE) 925 return 0; 926 if (!powerof2(fs_bsize)) 927 return 0; 928 929 /* Check the size of frag blocks */ 930 if (!powerof2(fs->fs_fsize)) 931 return 0; 932 933 if (fs->fs_size == 0) 934 return 0; 935 if (fs->fs_cssize == 0) 936 return 0; 937 938 /* Block size cannot be smaller than fragment size */ 939 if (fs_bsize < fs->fs_fsize) 940 return 0; 941 942 return 1; 943 } 944 945 /* 946 * Common code for mount and mountroot 947 */ 948 int 949 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l) 950 { 951 struct ufsmount *ump = NULL; 952 struct buf *bp = NULL; 953 struct fs *fs = NULL; 954 dev_t dev; 955 struct dkwedge_info dkw; 956 void *space; 957 daddr_t sblockloc = 0; 958 int blks, fstype = 0; 959 int error, i, bsize, ronly, bset = 0; 960 #ifdef FFS_EI 961 int needswap = 0; /* keep gcc happy */ 962 #endif 963 int32_t *lp; 964 kauth_cred_t cred; 965 u_int32_t fs_sbsize = 8192; /* keep gcc happy*/ 966 u_int32_t allocsbsize; 967 968 dev = devvp->v_rdev; 969 cred = l ? l->l_cred : NOCRED; 970 971 /* Flush out any old buffers remaining from a previous use. */ 972 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 973 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0); 974 VOP_UNLOCK(devvp); 975 if (error) { 976 DPRINTF(("%s: vinvalbuf %d\n", __func__, error)); 977 return error; 978 } 979 980 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 981 982 error = fstrans_mount(mp); 983 if (error) { 984 DPRINTF(("%s: fstrans_mount %d\n", __func__, error)); 985 return error; 986 } 987 988 ump = kmem_zalloc(sizeof(*ump), KM_SLEEP); 989 mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE); 990 error = ffs_snapshot_init(ump); 991 if (error) { 992 DPRINTF(("%s: ffs_snapshot_init %d\n", __func__, error)); 993 goto out; 994 } 995 ump->um_ops = &ffs_ufsops; 996 997 #ifdef WAPBL 998 sbagain: 999 #endif 1000 /* 1001 * Try reading the superblock in each of its possible locations. 1002 */ 1003 for (i = 0; ; i++) { 1004 daddr_t fsblockloc; 1005 int32_t fs_bsize; 1006 1007 if (bp != NULL) { 1008 brelse(bp, BC_NOCACHE); 1009 bp = NULL; 1010 } 1011 if (sblock_try[i] == -1) { 1012 DPRINTF(("%s: sblock_try\n", __func__)); 1013 error = EINVAL; 1014 fs = NULL; 1015 goto out; 1016 } 1017 1018 error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, 1019 cred, 0, &bp); 1020 if (error) { 1021 DPRINTF(("%s: bread@0x%x %d\n", __func__, 1022 sblock_try[i] / DEV_BSIZE, error)); 1023 fs = NULL; 1024 goto out; 1025 } 1026 fs = (struct fs*)bp->b_data; 1027 1028 fsblockloc = sblockloc = sblock_try[i]; 1029 DPRINTF(("%s: fs_magic 0x%x\n", __func__, fs->fs_magic)); 1030 if (fs->fs_magic == FS_UFS1_MAGIC) { 1031 fs_sbsize = fs->fs_sbsize; 1032 fstype = UFS1; 1033 fs_bsize = fs->fs_bsize; 1034 #ifdef FFS_EI 1035 needswap = 0; 1036 } else if (fs->fs_magic == FS_UFS1_MAGIC_SWAPPED) { 1037 fs_sbsize = bswap32(fs->fs_sbsize); 1038 fstype = UFS1; 1039 fs_bsize = bswap32(fs->fs_bsize); 1040 needswap = 1; 1041 #endif 1042 } else if (fs->fs_magic == FS_UFS2_MAGIC) { 1043 fs_sbsize = fs->fs_sbsize; 1044 fstype = UFS2; 1045 fs_bsize = fs->fs_bsize; 1046 #ifdef FFS_EI 1047 needswap = 0; 1048 } else if (fs->fs_magic == FS_UFS2_MAGIC_SWAPPED) { 1049 fs_sbsize = bswap32(fs->fs_sbsize); 1050 fstype = UFS2; 1051 fs_bsize = bswap32(fs->fs_bsize); 1052 needswap = 1; 1053 #endif 1054 } else 1055 continue; 1056 1057 /* fs->fs_sblockloc isn't defined for old filesystems */ 1058 if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) { 1059 if (sblockloc == SBLOCK_UFS2) 1060 /* 1061 * This is likely to be the first alternate 1062 * in a filesystem with 64k blocks. 1063 * Don't use it. 1064 */ 1065 continue; 1066 fsblockloc = sblockloc; 1067 } else { 1068 fsblockloc = fs->fs_sblockloc; 1069 #ifdef FFS_EI 1070 if (needswap) 1071 fsblockloc = bswap64(fsblockloc); 1072 #endif 1073 } 1074 1075 /* Check we haven't found an alternate superblock */ 1076 if (fsblockloc != sblockloc) 1077 continue; 1078 1079 if (!ffs_superblock_validate(fs, fs_sbsize, fs_bsize)) 1080 continue; 1081 1082 /* Ok seems to be a good superblock */ 1083 break; 1084 } 1085 1086 fs = kmem_alloc((u_long)fs_sbsize, KM_SLEEP); 1087 memcpy(fs, bp->b_data, fs_sbsize); 1088 ump->um_fs = fs; 1089 1090 #ifdef FFS_EI 1091 if (needswap) { 1092 ffs_sb_swap((struct fs*)bp->b_data, fs); 1093 fs->fs_flags |= FS_SWAPPED; 1094 } else 1095 #endif 1096 fs->fs_flags &= ~FS_SWAPPED; 1097 1098 #ifdef WAPBL 1099 if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) { 1100 error = ffs_wapbl_replay_start(mp, fs, devvp); 1101 if (error && (mp->mnt_flag & MNT_FORCE) == 0) { 1102 DPRINTF(("%s: ffs_wapbl_replay_start %d\n", __func__, 1103 error)); 1104 goto out; 1105 } 1106 if (!error) { 1107 if (!ronly) { 1108 /* XXX fsmnt may be stale. */ 1109 printf("%s: replaying log to disk\n", 1110 fs->fs_fsmnt); 1111 error = wapbl_replay_write(mp->mnt_wapbl_replay, 1112 devvp); 1113 if (error) { 1114 DPRINTF(("%s: wapbl_replay_write %d\n", 1115 __func__, error)); 1116 goto out; 1117 } 1118 wapbl_replay_stop(mp->mnt_wapbl_replay); 1119 fs->fs_clean = FS_WASCLEAN; 1120 } else { 1121 /* XXX fsmnt may be stale */ 1122 printf("%s: replaying log to memory\n", 1123 fs->fs_fsmnt); 1124 } 1125 1126 /* Force a re-read of the superblock */ 1127 brelse(bp, BC_INVAL); 1128 bp = NULL; 1129 kmem_free(fs, fs_sbsize); 1130 fs = NULL; 1131 goto sbagain; 1132 } 1133 } 1134 #else /* !WAPBL */ 1135 if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) { 1136 error = EPERM; 1137 DPRINTF(("%s: no force %d\n", __func__, error)); 1138 goto out; 1139 } 1140 #endif /* !WAPBL */ 1141 1142 ffs_oldfscompat_read(fs, ump, sblockloc); 1143 ump->um_maxfilesize = fs->fs_maxfilesize; 1144 1145 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) { 1146 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n", 1147 mp->mnt_stat.f_mntonname, fs->fs_flags, 1148 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 1149 if ((mp->mnt_flag & MNT_FORCE) == 0) { 1150 error = EINVAL; 1151 DPRINTF(("%s: no force %d\n", __func__, error)); 1152 goto out; 1153 } 1154 } 1155 1156 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 1157 fs->fs_pendingblocks = 0; 1158 fs->fs_pendinginodes = 0; 1159 } 1160 1161 ump->um_fstype = fstype; 1162 if (fs->fs_sbsize < SBLOCKSIZE) 1163 brelse(bp, BC_INVAL); 1164 else 1165 brelse(bp, 0); 1166 bp = NULL; 1167 1168 /* First check to see if this is tagged as an Apple UFS filesystem 1169 * in the disklabel 1170 */ 1171 if (getdiskinfo(devvp, &dkw) == 0 && 1172 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0) 1173 ump->um_flags |= UFS_ISAPPLEUFS; 1174 #ifdef APPLE_UFS 1175 else { 1176 /* Manually look for an apple ufs label, and if a valid one 1177 * is found, then treat it like an Apple UFS filesystem anyway 1178 */ 1179 error = bread(devvp, 1180 (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE), 1181 APPLEUFS_LABEL_SIZE, cred, 0, &bp); 1182 if (error) { 1183 DPRINTF(("%s: apple bread@0x%jx %d\n", __func__, 1184 (intmax_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE), 1185 error)); 1186 goto out; 1187 } 1188 error = ffs_appleufs_validate(fs->fs_fsmnt, 1189 (struct appleufslabel *)bp->b_data, NULL); 1190 if (error == 0) 1191 ump->um_flags |= UFS_ISAPPLEUFS; 1192 brelse(bp, 0); 1193 bp = NULL; 1194 } 1195 #else 1196 if (ump->um_flags & UFS_ISAPPLEUFS) { 1197 DPRINTF(("%s: bad apple\n", __func__)); 1198 error = EINVAL; 1199 goto out; 1200 } 1201 #endif 1202 1203 #if 0 1204 /* 1205 * XXX This code changes the behaviour of mounting dirty filesystems, to 1206 * XXX require "mount -f ..." to mount them. This doesn't match what 1207 * XXX mount(8) describes and is disabled for now. 1208 */ 1209 /* 1210 * If the file system is not clean, don't allow it to be mounted 1211 * unless MNT_FORCE is specified. (Note: MNT_FORCE is always set 1212 * for the root file system.) 1213 */ 1214 if (fs->fs_flags & FS_DOWAPBL) { 1215 /* 1216 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL 1217 * bit is set, although there's a window in unmount where it 1218 * could be FS_ISCLEAN 1219 */ 1220 if ((mp->mnt_flag & MNT_FORCE) == 0 && 1221 (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) { 1222 error = EPERM; 1223 goto out; 1224 } 1225 } else 1226 if ((fs->fs_clean & FS_ISCLEAN) == 0 && 1227 (mp->mnt_flag & MNT_FORCE) == 0) { 1228 error = EPERM; 1229 goto out; 1230 } 1231 #endif 1232 1233 /* 1234 * Verify that we can access the last block in the fs 1235 * if we're mounting read/write. 1236 */ 1237 1238 if (!ronly) { 1239 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_size - 1), 1240 fs->fs_fsize, cred, 0, &bp); 1241 if (error) { 1242 DPRINTF(("%s: bread@0x%jx %d\n", __func__, 1243 (intmax_t)FFS_FSBTODB(fs, fs->fs_size - 1), 1244 error)); 1245 bset = BC_INVAL; 1246 goto out; 1247 } 1248 if (bp->b_bcount != fs->fs_fsize) { 1249 DPRINTF(("%s: bcount %x != fsize %x\n", __func__, 1250 bp->b_bcount, fs->fs_fsize)); 1251 error = EINVAL; 1252 } 1253 brelse(bp, BC_INVAL); 1254 bp = NULL; 1255 } 1256 1257 fs->fs_ronly = ronly; 1258 /* Don't bump fs_clean if we're replaying journal */ 1259 if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN))) { 1260 if (ronly == 0) { 1261 fs->fs_clean <<= 1; 1262 fs->fs_fmod = 1; 1263 } 1264 } 1265 1266 bsize = fs->fs_cssize; 1267 blks = howmany(bsize, fs->fs_fsize); 1268 if (fs->fs_contigsumsize > 0) 1269 bsize += fs->fs_ncg * sizeof(int32_t); 1270 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1271 allocsbsize = bsize; 1272 space = kmem_alloc((u_long)allocsbsize, KM_SLEEP); 1273 fs->fs_csp = space; 1274 1275 for (i = 0; i < blks; i += fs->fs_frag) { 1276 bsize = fs->fs_bsize; 1277 if (i + fs->fs_frag > blks) 1278 bsize = (blks - i) * fs->fs_fsize; 1279 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize, 1280 cred, 0, &bp); 1281 if (error) { 1282 DPRINTF(("%s: bread@0x%jx %d\n", __func__, 1283 (intmax_t)FFS_FSBTODB(fs, fs->fs_csaddr + i), 1284 error)); 1285 goto out1; 1286 } 1287 #ifdef FFS_EI 1288 if (needswap) 1289 ffs_csum_swap((struct csum *)bp->b_data, 1290 (struct csum *)space, bsize); 1291 else 1292 #endif 1293 memcpy(space, bp->b_data, (u_int)bsize); 1294 1295 space = (char *)space + bsize; 1296 brelse(bp, 0); 1297 bp = NULL; 1298 } 1299 if (fs->fs_contigsumsize > 0) { 1300 fs->fs_maxcluster = lp = space; 1301 for (i = 0; i < fs->fs_ncg; i++) 1302 *lp++ = fs->fs_contigsumsize; 1303 space = lp; 1304 } 1305 bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1306 fs->fs_contigdirs = space; 1307 space = (char *)space + bsize; 1308 memset(fs->fs_contigdirs, 0, bsize); 1309 /* Compatibility for old filesystems - XXX */ 1310 if (fs->fs_avgfilesize <= 0) 1311 fs->fs_avgfilesize = AVFILESIZ; 1312 if (fs->fs_avgfpdir <= 0) 1313 fs->fs_avgfpdir = AFPDIR; 1314 fs->fs_active = NULL; 1315 mp->mnt_data = ump; 1316 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev; 1317 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS); 1318 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0]; 1319 mp->mnt_stat.f_namemax = FFS_MAXNAMLEN; 1320 if (UFS_MPISAPPLEUFS(ump)) { 1321 /* NeXT used to keep short symlinks in the inode even 1322 * when using FS_42INODEFMT. In that case fs->fs_maxsymlinklen 1323 * is probably -1, but we still need to be able to identify 1324 * short symlinks. 1325 */ 1326 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN; 1327 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ; 1328 mp->mnt_iflag |= IMNT_DTYPE; 1329 } else { 1330 ump->um_maxsymlinklen = fs->fs_maxsymlinklen; 1331 ump->um_dirblksiz = UFS_DIRBLKSIZ; 1332 if (ump->um_maxsymlinklen > 0) 1333 mp->mnt_iflag |= IMNT_DTYPE; 1334 else 1335 mp->mnt_iflag &= ~IMNT_DTYPE; 1336 } 1337 mp->mnt_fs_bshift = fs->fs_bshift; 1338 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */ 1339 mp->mnt_flag |= MNT_LOCAL; 1340 mp->mnt_iflag |= IMNT_MPSAFE; 1341 #ifdef FFS_EI 1342 if (needswap) 1343 ump->um_flags |= UFS_NEEDSWAP; 1344 #endif 1345 ump->um_mountp = mp; 1346 ump->um_dev = dev; 1347 ump->um_devvp = devvp; 1348 ump->um_nindir = fs->fs_nindir; 1349 ump->um_lognindir = ffs(fs->fs_nindir) - 1; 1350 ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT; 1351 ump->um_seqinc = fs->fs_frag; 1352 for (i = 0; i < MAXQUOTAS; i++) 1353 ump->um_quotas[i] = NULLVP; 1354 spec_node_setmountedfs(devvp, mp); 1355 if (ronly == 0 && fs->fs_snapinum[0] != 0) 1356 ffs_snapshot_mount(mp); 1357 #ifdef WAPBL 1358 if (!ronly) { 1359 KDASSERT(fs->fs_ronly == 0); 1360 /* 1361 * ffs_wapbl_start() needs mp->mnt_stat initialised if it 1362 * needs to create a new log file in-filesystem. 1363 */ 1364 error = ffs_statvfs(mp, &mp->mnt_stat); 1365 if (error) { 1366 DPRINTF(("%s: ffs_statvfs %d\n", __func__, error)); 1367 goto out1; 1368 } 1369 1370 error = ffs_wapbl_start(mp); 1371 if (error) { 1372 DPRINTF(("%s: ffs_wapbl_start %d\n", __func__, error)); 1373 goto out1; 1374 } 1375 } 1376 #endif /* WAPBL */ 1377 if (ronly == 0) { 1378 #ifdef QUOTA2 1379 error = ffs_quota2_mount(mp); 1380 if (error) { 1381 DPRINTF(("%s: ffs_quota2_mount %d\n", __func__, error)); 1382 goto out1; 1383 } 1384 #else 1385 if (fs->fs_flags & FS_DOQUOTA2) { 1386 ump->um_flags |= UFS_QUOTA2; 1387 uprintf("%s: options QUOTA2 not enabled%s\n", 1388 mp->mnt_stat.f_mntonname, 1389 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 1390 if ((mp->mnt_flag & MNT_FORCE) == 0) { 1391 error = EINVAL; 1392 DPRINTF(("%s: quota disabled %d\n", __func__, 1393 error)); 1394 goto out1; 1395 } 1396 } 1397 #endif 1398 } 1399 1400 if (mp->mnt_flag & MNT_DISCARD) 1401 ump->um_discarddata = ffs_discard_init(devvp, fs); 1402 1403 return (0); 1404 out1: 1405 kmem_free(fs->fs_csp, allocsbsize); 1406 out: 1407 #ifdef WAPBL 1408 if (mp->mnt_wapbl_replay) { 1409 wapbl_replay_stop(mp->mnt_wapbl_replay); 1410 wapbl_replay_free(mp->mnt_wapbl_replay); 1411 mp->mnt_wapbl_replay = 0; 1412 } 1413 #endif 1414 1415 fstrans_unmount(mp); 1416 if (fs) 1417 kmem_free(fs, fs->fs_sbsize); 1418 spec_node_setmountedfs(devvp, NULL); 1419 if (bp) 1420 brelse(bp, bset); 1421 if (ump) { 1422 if (ump->um_oldfscompat) 1423 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t)); 1424 mutex_destroy(&ump->um_lock); 1425 kmem_free(ump, sizeof(*ump)); 1426 mp->mnt_data = NULL; 1427 } 1428 return (error); 1429 } 1430 1431 /* 1432 * Sanity checks for loading old filesystem superblocks. 1433 * See ffs_oldfscompat_write below for unwound actions. 1434 * 1435 * XXX - Parts get retired eventually. 1436 * Unfortunately new bits get added. 1437 */ 1438 static void 1439 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc) 1440 { 1441 off_t maxfilesize; 1442 int32_t *extrasave; 1443 1444 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1445 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1446 return; 1447 1448 if (!ump->um_oldfscompat) 1449 ump->um_oldfscompat = kmem_alloc(512 + 3*sizeof(int32_t), 1450 KM_SLEEP); 1451 1452 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512); 1453 extrasave = ump->um_oldfscompat; 1454 extrasave += 512/sizeof(int32_t); 1455 extrasave[0] = fs->fs_old_npsect; 1456 extrasave[1] = fs->fs_old_interleave; 1457 extrasave[2] = fs->fs_old_trackskew; 1458 1459 /* These fields will be overwritten by their 1460 * original values in fs_oldfscompat_write, so it is harmless 1461 * to modify them here. 1462 */ 1463 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir; 1464 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree; 1465 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree; 1466 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree; 1467 1468 fs->fs_maxbsize = fs->fs_bsize; 1469 fs->fs_time = fs->fs_old_time; 1470 fs->fs_size = fs->fs_old_size; 1471 fs->fs_dsize = fs->fs_old_dsize; 1472 fs->fs_csaddr = fs->fs_old_csaddr; 1473 fs->fs_sblockloc = sblockloc; 1474 1475 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL); 1476 1477 if (fs->fs_old_postblformat == FS_42POSTBLFMT) { 1478 fs->fs_old_nrpos = 8; 1479 fs->fs_old_npsect = fs->fs_old_nsect; 1480 fs->fs_old_interleave = 1; 1481 fs->fs_old_trackskew = 0; 1482 } 1483 1484 if (fs->fs_old_inodefmt < FS_44INODEFMT) { 1485 fs->fs_maxfilesize = (u_quad_t) 1LL << 39; 1486 fs->fs_qbmask = ~fs->fs_bmask; 1487 fs->fs_qfmask = ~fs->fs_fmask; 1488 } 1489 1490 maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1; 1491 if (fs->fs_maxfilesize > maxfilesize) 1492 fs->fs_maxfilesize = maxfilesize; 1493 1494 /* Compatibility for old filesystems */ 1495 if (fs->fs_avgfilesize <= 0) 1496 fs->fs_avgfilesize = AVFILESIZ; 1497 if (fs->fs_avgfpdir <= 0) 1498 fs->fs_avgfpdir = AFPDIR; 1499 1500 #if 0 1501 if (bigcgs) { 1502 fs->fs_save_cgsize = fs->fs_cgsize; 1503 fs->fs_cgsize = fs->fs_bsize; 1504 } 1505 #endif 1506 } 1507 1508 /* 1509 * Unwinding superblock updates for old filesystems. 1510 * See ffs_oldfscompat_read above for details. 1511 * 1512 * XXX - Parts get retired eventually. 1513 * Unfortunately new bits get added. 1514 */ 1515 static void 1516 ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump) 1517 { 1518 int32_t *extrasave; 1519 1520 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1521 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1522 return; 1523 1524 fs->fs_old_time = fs->fs_time; 1525 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir; 1526 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree; 1527 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree; 1528 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree; 1529 fs->fs_old_flags = fs->fs_flags; 1530 1531 #if 0 1532 if (bigcgs) { 1533 fs->fs_cgsize = fs->fs_save_cgsize; 1534 } 1535 #endif 1536 1537 memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512); 1538 extrasave = ump->um_oldfscompat; 1539 extrasave += 512/sizeof(int32_t); 1540 fs->fs_old_npsect = extrasave[0]; 1541 fs->fs_old_interleave = extrasave[1]; 1542 fs->fs_old_trackskew = extrasave[2]; 1543 1544 } 1545 1546 /* 1547 * unmount vfs operation 1548 */ 1549 int 1550 ffs_unmount(struct mount *mp, int mntflags) 1551 { 1552 struct lwp *l = curlwp; 1553 struct ufsmount *ump = VFSTOUFS(mp); 1554 struct fs *fs = ump->um_fs; 1555 int error, flags; 1556 u_int32_t bsize; 1557 #ifdef WAPBL 1558 extern int doforce; 1559 #endif 1560 1561 if (ump->um_discarddata) { 1562 ffs_discard_finish(ump->um_discarddata, mntflags); 1563 ump->um_discarddata = NULL; 1564 } 1565 1566 flags = 0; 1567 if (mntflags & MNT_FORCE) 1568 flags |= FORCECLOSE; 1569 if ((error = ffs_flushfiles(mp, flags, l)) != 0) 1570 return (error); 1571 error = UFS_WAPBL_BEGIN(mp); 1572 if (error == 0) 1573 if (fs->fs_ronly == 0 && 1574 ffs_cgupdate(ump, MNT_WAIT) == 0 && 1575 fs->fs_clean & FS_WASCLEAN) { 1576 fs->fs_clean = FS_ISCLEAN; 1577 fs->fs_fmod = 0; 1578 (void) ffs_sbupdate(ump, MNT_WAIT); 1579 } 1580 if (error == 0) 1581 UFS_WAPBL_END(mp); 1582 #ifdef WAPBL 1583 KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl)); 1584 if (mp->mnt_wapbl_replay) { 1585 KDASSERT(fs->fs_ronly); 1586 wapbl_replay_stop(mp->mnt_wapbl_replay); 1587 wapbl_replay_free(mp->mnt_wapbl_replay); 1588 mp->mnt_wapbl_replay = 0; 1589 } 1590 error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE)); 1591 if (error) { 1592 return error; 1593 } 1594 #endif /* WAPBL */ 1595 1596 if (ump->um_devvp->v_type != VBAD) 1597 spec_node_setmountedfs(ump->um_devvp, NULL); 1598 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1599 (void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE, 1600 NOCRED); 1601 vput(ump->um_devvp); 1602 1603 bsize = fs->fs_cssize; 1604 if (fs->fs_contigsumsize > 0) 1605 bsize += fs->fs_ncg * sizeof(int32_t); 1606 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1607 kmem_free(fs->fs_csp, bsize); 1608 1609 kmem_free(fs, fs->fs_sbsize); 1610 if (ump->um_oldfscompat != NULL) 1611 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t)); 1612 mutex_destroy(&ump->um_lock); 1613 ffs_snapshot_fini(ump); 1614 kmem_free(ump, sizeof(*ump)); 1615 mp->mnt_data = NULL; 1616 mp->mnt_flag &= ~MNT_LOCAL; 1617 fstrans_unmount(mp); 1618 return (0); 1619 } 1620 1621 /* 1622 * Flush out all the files in a filesystem. 1623 */ 1624 int 1625 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l) 1626 { 1627 extern int doforce; 1628 struct ufsmount *ump; 1629 int error; 1630 1631 if (!doforce) 1632 flags &= ~FORCECLOSE; 1633 ump = VFSTOUFS(mp); 1634 #ifdef QUOTA 1635 if ((error = quota1_umount(mp, flags)) != 0) 1636 return (error); 1637 #endif 1638 #ifdef QUOTA2 1639 if ((error = quota2_umount(mp, flags)) != 0) 1640 return (error); 1641 #endif 1642 #ifdef UFS_EXTATTR 1643 if (ump->um_fstype == UFS1) { 1644 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_STARTED) 1645 ufs_extattr_stop(mp, l); 1646 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_INITIALIZED) 1647 ufs_extattr_uepm_destroy(&ump->um_extattr); 1648 mp->mnt_flag &= ~MNT_EXTATTR; 1649 } 1650 #endif 1651 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0) 1652 return (error); 1653 ffs_snapshot_unmount(mp); 1654 /* 1655 * Flush all the files. 1656 */ 1657 error = vflush(mp, NULLVP, flags); 1658 if (error) 1659 return (error); 1660 /* 1661 * Flush filesystem metadata. 1662 */ 1663 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1664 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0); 1665 VOP_UNLOCK(ump->um_devvp); 1666 if (flags & FORCECLOSE) /* XXXDBJ */ 1667 error = 0; 1668 1669 #ifdef WAPBL 1670 if (error) 1671 return error; 1672 if (mp->mnt_wapbl) { 1673 error = wapbl_flush(mp->mnt_wapbl, 1); 1674 if (flags & FORCECLOSE) 1675 error = 0; 1676 } 1677 #endif 1678 1679 return (error); 1680 } 1681 1682 /* 1683 * Get file system statistics. 1684 */ 1685 int 1686 ffs_statvfs(struct mount *mp, struct statvfs *sbp) 1687 { 1688 struct ufsmount *ump; 1689 struct fs *fs; 1690 1691 ump = VFSTOUFS(mp); 1692 fs = ump->um_fs; 1693 mutex_enter(&ump->um_lock); 1694 sbp->f_bsize = fs->fs_bsize; 1695 sbp->f_frsize = fs->fs_fsize; 1696 sbp->f_iosize = fs->fs_bsize; 1697 sbp->f_blocks = fs->fs_dsize; 1698 sbp->f_bfree = ffs_blkstofrags(fs, fs->fs_cstotal.cs_nbfree) + 1699 fs->fs_cstotal.cs_nffree + FFS_DBTOFSB(fs, fs->fs_pendingblocks); 1700 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t) 1701 fs->fs_minfree) / (u_int64_t) 100; 1702 if (sbp->f_bfree > sbp->f_bresvd) 1703 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd; 1704 else 1705 sbp->f_bavail = 0; 1706 sbp->f_files = fs->fs_ncg * fs->fs_ipg - UFS_ROOTINO; 1707 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes; 1708 sbp->f_favail = sbp->f_ffree; 1709 sbp->f_fresvd = 0; 1710 mutex_exit(&ump->um_lock); 1711 copy_statvfs_info(sbp, mp); 1712 1713 return (0); 1714 } 1715 1716 struct ffs_sync_ctx { 1717 int waitfor; 1718 bool is_suspending; 1719 }; 1720 1721 static bool 1722 ffs_sync_selector(void *cl, struct vnode *vp) 1723 { 1724 struct ffs_sync_ctx *c = cl; 1725 struct inode *ip; 1726 1727 ip = VTOI(vp); 1728 /* 1729 * Skip the vnode/inode if inaccessible. 1730 */ 1731 if (ip == NULL || vp->v_type == VNON) 1732 return false; 1733 1734 /* 1735 * We deliberately update inode times here. This will 1736 * prevent a massive queue of updates accumulating, only 1737 * to be handled by a call to unmount. 1738 * 1739 * XXX It would be better to have the syncer trickle these 1740 * out. Adjustment needed to allow registering vnodes for 1741 * sync when the vnode is clean, but the inode dirty. Or 1742 * have ufs itself trickle out inode updates. 1743 * 1744 * If doing a lazy sync, we don't care about metadata or 1745 * data updates, because they are handled by each vnode's 1746 * synclist entry. In this case we are only interested in 1747 * writing back modified inodes. 1748 */ 1749 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | 1750 IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 && 1751 (c->waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) && 1752 UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) 1753 return false; 1754 1755 if (vp->v_type == VBLK && c->is_suspending) 1756 return false; 1757 1758 return true; 1759 } 1760 1761 /* 1762 * Go through the disk queues to initiate sandbagged IO; 1763 * go through the inodes to write those that have been modified; 1764 * initiate the writing of the super block if it has been modified. 1765 * 1766 * Note: we are always called with the filesystem marked `MPBUSY'. 1767 */ 1768 int 1769 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred) 1770 { 1771 struct vnode *vp; 1772 struct ufsmount *ump = VFSTOUFS(mp); 1773 struct fs *fs; 1774 struct vnode_iterator *marker; 1775 int error, allerror = 0; 1776 bool is_suspending; 1777 struct ffs_sync_ctx ctx; 1778 1779 fs = ump->um_fs; 1780 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */ 1781 printf("fs = %s\n", fs->fs_fsmnt); 1782 panic("update: rofs mod"); 1783 } 1784 1785 fstrans_start(mp, FSTRANS_SHARED); 1786 is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING); 1787 /* 1788 * Write back each (modified) inode. 1789 */ 1790 vfs_vnode_iterator_init(mp, &marker); 1791 1792 ctx.waitfor = waitfor; 1793 ctx.is_suspending = is_suspending; 1794 while ((vp = vfs_vnode_iterator_next(marker, ffs_sync_selector, &ctx))) 1795 { 1796 error = vn_lock(vp, LK_EXCLUSIVE); 1797 if (error) { 1798 vrele(vp); 1799 continue; 1800 } 1801 if (waitfor == MNT_LAZY) { 1802 error = UFS_WAPBL_BEGIN(vp->v_mount); 1803 if (!error) { 1804 error = ffs_update(vp, NULL, NULL, 1805 UPDATE_CLOSE); 1806 UFS_WAPBL_END(vp->v_mount); 1807 } 1808 } else { 1809 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG | 1810 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0); 1811 } 1812 if (error) 1813 allerror = error; 1814 vput(vp); 1815 } 1816 vfs_vnode_iterator_destroy(marker); 1817 1818 /* 1819 * Force stale file system control information to be flushed. 1820 */ 1821 if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 || 1822 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) { 1823 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1824 if ((error = VOP_FSYNC(ump->um_devvp, cred, 1825 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG, 1826 0, 0)) != 0) 1827 allerror = error; 1828 VOP_UNLOCK(ump->um_devvp); 1829 } 1830 #if defined(QUOTA) || defined(QUOTA2) 1831 qsync(mp); 1832 #endif 1833 /* 1834 * Write back modified superblock. 1835 */ 1836 if (fs->fs_fmod != 0) { 1837 fs->fs_fmod = 0; 1838 fs->fs_time = time_second; 1839 error = UFS_WAPBL_BEGIN(mp); 1840 if (error) 1841 allerror = error; 1842 else { 1843 if ((error = ffs_cgupdate(ump, waitfor))) 1844 allerror = error; 1845 UFS_WAPBL_END(mp); 1846 } 1847 } 1848 1849 #ifdef WAPBL 1850 if (mp->mnt_wapbl) { 1851 error = wapbl_flush(mp->mnt_wapbl, 0); 1852 if (error) 1853 allerror = error; 1854 } 1855 #endif 1856 1857 fstrans_done(mp); 1858 return (allerror); 1859 } 1860 1861 /* 1862 * Read an inode from disk and initialize this vnode / inode pair. 1863 * Caller assures no other thread will try to load this inode. 1864 */ 1865 int 1866 ffs_loadvnode(struct mount *mp, struct vnode *vp, 1867 const void *key, size_t key_len, const void **new_key) 1868 { 1869 ino_t ino; 1870 struct fs *fs; 1871 struct inode *ip; 1872 struct ufsmount *ump; 1873 struct buf *bp; 1874 dev_t dev; 1875 int error; 1876 1877 KASSERT(key_len == sizeof(ino)); 1878 memcpy(&ino, key, key_len); 1879 ump = VFSTOUFS(mp); 1880 dev = ump->um_dev; 1881 fs = ump->um_fs; 1882 1883 /* Read in the disk contents for the inode. */ 1884 error = bread(ump->um_devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ino)), 1885 (int)fs->fs_bsize, NOCRED, 0, &bp); 1886 if (error) 1887 return error; 1888 1889 /* Allocate and initialize inode. */ 1890 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK); 1891 memset(ip, 0, sizeof(struct inode)); 1892 vp->v_tag = VT_UFS; 1893 vp->v_op = ffs_vnodeop_p; 1894 vp->v_vflag |= VV_LOCKSWORK; 1895 vp->v_data = ip; 1896 ip->i_vnode = vp; 1897 ip->i_ump = ump; 1898 ip->i_fs = fs; 1899 ip->i_dev = dev; 1900 ip->i_number = ino; 1901 #if defined(QUOTA) || defined(QUOTA2) 1902 ufsquota_init(ip); 1903 #endif 1904 1905 /* Initialize genfs node. */ 1906 genfs_node_init(vp, &ffs_genfsops); 1907 1908 if (ip->i_ump->um_fstype == UFS1) 1909 ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache, 1910 PR_WAITOK); 1911 else 1912 ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache, 1913 PR_WAITOK); 1914 ffs_load_inode(bp, ip, fs, ino); 1915 brelse(bp, 0); 1916 1917 /* Initialize the vnode from the inode. */ 1918 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp); 1919 1920 /* Finish inode initialization. */ 1921 ip->i_devvp = ump->um_devvp; 1922 vref(ip->i_devvp); 1923 1924 /* 1925 * Ensure that uid and gid are correct. This is a temporary 1926 * fix until fsck has been changed to do the update. 1927 */ 1928 1929 if (fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */ 1930 ip->i_uid = ip->i_ffs1_ouid; /* XXX */ 1931 ip->i_gid = ip->i_ffs1_ogid; /* XXX */ 1932 } /* XXX */ 1933 uvm_vnp_setsize(vp, ip->i_size); 1934 *new_key = &ip->i_number; 1935 return 0; 1936 } 1937 1938 /* 1939 * File handle to vnode 1940 * 1941 * Have to be really careful about stale file handles: 1942 * - check that the inode number is valid 1943 * - call ffs_vget() to get the locked inode 1944 * - check for an unallocated inode (i_mode == 0) 1945 * - check that the given client host has export rights and return 1946 * those rights via. exflagsp and credanonp 1947 */ 1948 int 1949 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp) 1950 { 1951 struct ufid ufh; 1952 struct fs *fs; 1953 1954 if (fhp->fid_len != sizeof(struct ufid)) 1955 return EINVAL; 1956 1957 memcpy(&ufh, fhp, sizeof(ufh)); 1958 fs = VFSTOUFS(mp)->um_fs; 1959 if (ufh.ufid_ino < UFS_ROOTINO || 1960 ufh.ufid_ino >= fs->fs_ncg * fs->fs_ipg) 1961 return (ESTALE); 1962 return (ufs_fhtovp(mp, &ufh, vpp)); 1963 } 1964 1965 /* 1966 * Vnode pointer to File handle 1967 */ 1968 /* ARGSUSED */ 1969 int 1970 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size) 1971 { 1972 struct inode *ip; 1973 struct ufid ufh; 1974 1975 if (*fh_size < sizeof(struct ufid)) { 1976 *fh_size = sizeof(struct ufid); 1977 return E2BIG; 1978 } 1979 ip = VTOI(vp); 1980 *fh_size = sizeof(struct ufid); 1981 memset(&ufh, 0, sizeof(ufh)); 1982 ufh.ufid_len = sizeof(struct ufid); 1983 ufh.ufid_ino = ip->i_number; 1984 ufh.ufid_gen = ip->i_gen; 1985 memcpy(fhp, &ufh, sizeof(ufh)); 1986 return (0); 1987 } 1988 1989 void 1990 ffs_init(void) 1991 { 1992 if (ffs_initcount++ > 0) 1993 return; 1994 1995 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0, 1996 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL); 1997 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0, 1998 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL); 1999 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0, 2000 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL); 2001 ufs_init(); 2002 } 2003 2004 void 2005 ffs_reinit(void) 2006 { 2007 2008 ufs_reinit(); 2009 } 2010 2011 void 2012 ffs_done(void) 2013 { 2014 if (--ffs_initcount > 0) 2015 return; 2016 2017 ufs_done(); 2018 pool_cache_destroy(ffs_dinode2_cache); 2019 pool_cache_destroy(ffs_dinode1_cache); 2020 pool_cache_destroy(ffs_inode_cache); 2021 } 2022 2023 /* 2024 * Write a superblock and associated information back to disk. 2025 */ 2026 int 2027 ffs_sbupdate(struct ufsmount *mp, int waitfor) 2028 { 2029 struct fs *fs = mp->um_fs; 2030 struct buf *bp; 2031 int error = 0; 2032 u_int32_t saveflag; 2033 2034 error = ffs_getblk(mp->um_devvp, 2035 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK, 2036 fs->fs_sbsize, false, &bp); 2037 if (error) 2038 return error; 2039 saveflag = fs->fs_flags & FS_INTERNAL; 2040 fs->fs_flags &= ~FS_INTERNAL; 2041 2042 memcpy(bp->b_data, fs, fs->fs_sbsize); 2043 2044 ffs_oldfscompat_write((struct fs *)bp->b_data, mp); 2045 #ifdef FFS_EI 2046 if (mp->um_flags & UFS_NEEDSWAP) 2047 ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data); 2048 #endif 2049 fs->fs_flags |= saveflag; 2050 2051 if (waitfor == MNT_WAIT) 2052 error = bwrite(bp); 2053 else 2054 bawrite(bp); 2055 return (error); 2056 } 2057 2058 int 2059 ffs_cgupdate(struct ufsmount *mp, int waitfor) 2060 { 2061 struct fs *fs = mp->um_fs; 2062 struct buf *bp; 2063 int blks; 2064 void *space; 2065 int i, size, error = 0, allerror = 0; 2066 2067 allerror = ffs_sbupdate(mp, waitfor); 2068 blks = howmany(fs->fs_cssize, fs->fs_fsize); 2069 space = fs->fs_csp; 2070 for (i = 0; i < blks; i += fs->fs_frag) { 2071 size = fs->fs_bsize; 2072 if (i + fs->fs_frag > blks) 2073 size = (blks - i) * fs->fs_fsize; 2074 error = ffs_getblk(mp->um_devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), 2075 FFS_NOBLK, size, false, &bp); 2076 if (error) 2077 break; 2078 #ifdef FFS_EI 2079 if (mp->um_flags & UFS_NEEDSWAP) 2080 ffs_csum_swap((struct csum*)space, 2081 (struct csum*)bp->b_data, size); 2082 else 2083 #endif 2084 memcpy(bp->b_data, space, (u_int)size); 2085 space = (char *)space + size; 2086 if (waitfor == MNT_WAIT) 2087 error = bwrite(bp); 2088 else 2089 bawrite(bp); 2090 } 2091 if (!allerror && error) 2092 allerror = error; 2093 return (allerror); 2094 } 2095 2096 int 2097 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp, 2098 int attrnamespace, const char *attrname) 2099 { 2100 #ifdef UFS_EXTATTR 2101 /* 2102 * File-backed extended attributes are only supported on UFS1. 2103 * UFS2 has native extended attributes. 2104 */ 2105 if (VFSTOUFS(mp)->um_fstype == UFS1) 2106 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname)); 2107 #endif 2108 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname)); 2109 } 2110 2111 int 2112 ffs_suspendctl(struct mount *mp, int cmd) 2113 { 2114 int error; 2115 struct lwp *l = curlwp; 2116 2117 switch (cmd) { 2118 case SUSPEND_SUSPEND: 2119 if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0) 2120 return error; 2121 error = ffs_sync(mp, MNT_WAIT, l->l_proc->p_cred); 2122 if (error == 0) 2123 error = fstrans_setstate(mp, FSTRANS_SUSPENDED); 2124 #ifdef WAPBL 2125 if (error == 0 && mp->mnt_wapbl) 2126 error = wapbl_flush(mp->mnt_wapbl, 1); 2127 #endif 2128 if (error != 0) { 2129 (void) fstrans_setstate(mp, FSTRANS_NORMAL); 2130 return error; 2131 } 2132 return 0; 2133 2134 case SUSPEND_RESUME: 2135 return fstrans_setstate(mp, FSTRANS_NORMAL); 2136 2137 default: 2138 return EINVAL; 2139 } 2140 } 2141 2142 /* 2143 * Synch vnode for a mounted file system. 2144 */ 2145 static int 2146 ffs_vfs_fsync(vnode_t *vp, int flags) 2147 { 2148 int error, i, pflags; 2149 #ifdef WAPBL 2150 struct mount *mp; 2151 #endif 2152 2153 KASSERT(vp->v_type == VBLK); 2154 KASSERT(spec_node_getmountedfs(vp) != NULL); 2155 2156 /* 2157 * Flush all dirty data associated with the vnode. 2158 */ 2159 pflags = PGO_ALLPAGES | PGO_CLEANIT; 2160 if ((flags & FSYNC_WAIT) != 0) 2161 pflags |= PGO_SYNCIO; 2162 mutex_enter(vp->v_interlock); 2163 error = VOP_PUTPAGES(vp, 0, 0, pflags); 2164 if (error) 2165 return error; 2166 2167 #ifdef WAPBL 2168 mp = spec_node_getmountedfs(vp); 2169 if (mp && mp->mnt_wapbl) { 2170 /* 2171 * Don't bother writing out metadata if the syncer is 2172 * making the request. We will let the sync vnode 2173 * write it out in a single burst through a call to 2174 * VFS_SYNC(). 2175 */ 2176 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0) 2177 return 0; 2178 2179 /* 2180 * Don't flush the log if the vnode being flushed 2181 * contains no dirty buffers that could be in the log. 2182 */ 2183 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 2184 error = wapbl_flush(mp->mnt_wapbl, 0); 2185 if (error) 2186 return error; 2187 } 2188 2189 if ((flags & FSYNC_WAIT) != 0) { 2190 mutex_enter(vp->v_interlock); 2191 while (vp->v_numoutput) 2192 cv_wait(&vp->v_cv, vp->v_interlock); 2193 mutex_exit(vp->v_interlock); 2194 } 2195 2196 return 0; 2197 } 2198 #endif /* WAPBL */ 2199 2200 error = vflushbuf(vp, flags); 2201 if (error == 0 && (flags & FSYNC_CACHE) != 0) { 2202 i = 1; 2203 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE, 2204 kauth_cred_get()); 2205 } 2206 2207 return error; 2208 } 2209
Indexes created Thu Oct 02 14:10:14 GMT 2025