ffs_vfsops.c revision 1.316
1 /* $NetBSD: ffs_vfsops.c,v 1.316 2015/02/14 13:43:28 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 1.316 2015/02/14 13:43:28 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); 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, fs_sbsize; 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 /* 718 * Step 1: invalidate all cached meta-data. 719 */ 720 devvp = ump->um_devvp; 721 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 722 error = vinvalbuf(devvp, 0, cred, l, 0, 0); 723 VOP_UNLOCK(devvp); 724 if (error) 725 panic("ffs_reload: dirty1"); 726 727 /* 728 * Step 2: re-read superblock from disk. XXX: We don't handle 729 * possibility that superblock moved. Which implies that we don't 730 * want its size to change either. 731 */ 732 fs = ump->um_fs; 733 fs_sbsize = fs->fs_sbsize; 734 error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs_sbsize, 735 NOCRED, 0, &bp); 736 if (error) 737 return (error); 738 newfs = kmem_alloc(fs_sbsize, KM_SLEEP); 739 memcpy(newfs, bp->b_data, fs_sbsize); 740 741 #ifdef FFS_EI 742 if (ump->um_flags & UFS_NEEDSWAP) { 743 ffs_sb_swap((struct fs*)bp->b_data, newfs); 744 fs->fs_flags |= FS_SWAPPED; 745 } else 746 #endif 747 fs->fs_flags &= ~FS_SWAPPED; 748 749 /* We don't want the superblock size to change. */ 750 if (newfs->fs_sbsize != fs_sbsize) { 751 brelse(bp, 0); 752 kmem_free(newfs, fs_sbsize); 753 return (EINVAL); 754 } 755 if ((newfs->fs_magic != FS_UFS1_MAGIC && 756 newfs->fs_magic != FS_UFS2_MAGIC)) { 757 brelse(bp, 0); 758 kmem_free(newfs, fs_sbsize); 759 return (EIO); /* XXX needs translation */ 760 } 761 if (!ffs_superblock_validate(newfs)) { 762 brelse(bp, 0); 763 kmem_free(newfs, fs_sbsize); 764 return (EINVAL); 765 } 766 767 /* Store off old fs_sblockloc for fs_oldfscompat_read. */ 768 sblockloc = fs->fs_sblockloc; 769 /* 770 * Copy pointer fields back into superblock before copying in XXX 771 * new superblock. These should really be in the ufsmount. XXX 772 * Note that important parameters (eg fs_ncg) are unchanged. 773 */ 774 newfs->fs_csp = fs->fs_csp; 775 newfs->fs_maxcluster = fs->fs_maxcluster; 776 newfs->fs_contigdirs = fs->fs_contigdirs; 777 newfs->fs_ronly = fs->fs_ronly; 778 newfs->fs_active = fs->fs_active; 779 memcpy(fs, newfs, (u_int)fs_sbsize); 780 brelse(bp, 0); 781 kmem_free(newfs, fs_sbsize); 782 783 /* Recheck for apple UFS filesystem */ 784 ump->um_flags &= ~UFS_ISAPPLEUFS; 785 /* First check to see if this is tagged as an Apple UFS filesystem 786 * in the disklabel 787 */ 788 if (getdiskinfo(devvp, &dkw) == 0 && 789 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0) 790 ump->um_flags |= UFS_ISAPPLEUFS; 791 #ifdef APPLE_UFS 792 else { 793 /* Manually look for an apple ufs label, and if a valid one 794 * is found, then treat it like an Apple UFS filesystem anyway 795 * 796 * EINVAL is most probably a blocksize or alignment problem, 797 * it is unlikely that this is an Apple UFS filesystem then. 798 */ 799 error = bread(devvp, 800 (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE), 801 APPLEUFS_LABEL_SIZE, cred, 0, &bp); 802 if (error && error != EINVAL) { 803 return error; 804 } 805 if (error == 0) { 806 error = ffs_appleufs_validate(fs->fs_fsmnt, 807 (struct appleufslabel *)bp->b_data, NULL); 808 if (error == 0) 809 ump->um_flags |= UFS_ISAPPLEUFS; 810 brelse(bp, 0); 811 } 812 bp = NULL; 813 } 814 #else 815 if (ump->um_flags & UFS_ISAPPLEUFS) 816 return (EIO); 817 #endif 818 819 if (UFS_MPISAPPLEUFS(ump)) { 820 /* see comment about NeXT below */ 821 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN; 822 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ; 823 mp->mnt_iflag |= IMNT_DTYPE; 824 } else { 825 ump->um_maxsymlinklen = fs->fs_maxsymlinklen; 826 ump->um_dirblksiz = UFS_DIRBLKSIZ; 827 if (ump->um_maxsymlinklen > 0) 828 mp->mnt_iflag |= IMNT_DTYPE; 829 else 830 mp->mnt_iflag &= ~IMNT_DTYPE; 831 } 832 ffs_oldfscompat_read(fs, ump, sblockloc); 833 834 mutex_enter(&ump->um_lock); 835 ump->um_maxfilesize = fs->fs_maxfilesize; 836 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) { 837 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n", 838 mp->mnt_stat.f_mntonname, fs->fs_flags, 839 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 840 if ((mp->mnt_flag & MNT_FORCE) == 0) { 841 mutex_exit(&ump->um_lock); 842 return (EINVAL); 843 } 844 } 845 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 846 fs->fs_pendingblocks = 0; 847 fs->fs_pendinginodes = 0; 848 } 849 mutex_exit(&ump->um_lock); 850 851 ffs_statvfs(mp, &mp->mnt_stat); 852 /* 853 * Step 3: re-read summary information from disk. 854 */ 855 blks = howmany(fs->fs_cssize, fs->fs_fsize); 856 space = fs->fs_csp; 857 for (i = 0; i < blks; i += fs->fs_frag) { 858 bsize = fs->fs_bsize; 859 if (i + fs->fs_frag > blks) 860 bsize = (blks - i) * fs->fs_fsize; 861 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize, 862 NOCRED, 0, &bp); 863 if (error) { 864 return (error); 865 } 866 #ifdef FFS_EI 867 if (UFS_FSNEEDSWAP(fs)) 868 ffs_csum_swap((struct csum *)bp->b_data, 869 (struct csum *)space, bsize); 870 else 871 #endif 872 memcpy(space, bp->b_data, (size_t)bsize); 873 space = (char *)space + bsize; 874 brelse(bp, 0); 875 } 876 /* 877 * We no longer know anything about clusters per cylinder group. 878 */ 879 if (fs->fs_contigsumsize > 0) { 880 lp = fs->fs_maxcluster; 881 for (i = 0; i < fs->fs_ncg; i++) 882 *lp++ = fs->fs_contigsumsize; 883 } 884 885 vfs_vnode_iterator_init(mp, &marker); 886 while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) { 887 /* 888 * Step 4: invalidate all inactive vnodes. 889 */ 890 if (vrecycle(vp)) 891 continue; 892 /* 893 * Step 5: invalidate all cached file data. 894 */ 895 if (vn_lock(vp, LK_EXCLUSIVE)) { 896 vrele(vp); 897 continue; 898 } 899 if (vinvalbuf(vp, 0, cred, l, 0, 0)) 900 panic("ffs_reload: dirty2"); 901 /* 902 * Step 6: re-read inode data for all active vnodes. 903 */ 904 ip = VTOI(vp); 905 error = bread(devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)), 906 (int)fs->fs_bsize, NOCRED, 0, &bp); 907 if (error) { 908 vput(vp); 909 break; 910 } 911 ffs_load_inode(bp, ip, fs, ip->i_number); 912 brelse(bp, 0); 913 vput(vp); 914 } 915 vfs_vnode_iterator_destroy(marker); 916 return (error); 917 } 918 919 /* 920 * Possible superblock locations ordered from most to least likely. 921 */ 922 static const int sblock_try[] = SBLOCKSEARCH; 923 924 925 static int 926 ffs_superblock_validate(struct fs *fs) 927 { 928 int32_t i, fs_bshift = 0, fs_fshift = 0, fs_frag; 929 930 /* Check the superblock size */ 931 if (fs->fs_sbsize > SBLOCKSIZE || fs->fs_sbsize < sizeof(struct fs)) 932 return 0; 933 934 /* Check the file system blocksize */ 935 if (fs->fs_bsize > MAXBSIZE || fs->fs_bsize < MINBSIZE) 936 return 0; 937 if (!powerof2(fs->fs_bsize)) 938 return 0; 939 940 /* Check the size of frag blocks */ 941 if (!powerof2(fs->fs_fsize)) 942 return 0; 943 944 if (fs->fs_size == 0) 945 return 0; 946 if (fs->fs_cssize == 0) 947 return 0; 948 949 /* Block size cannot be smaller than fragment size */ 950 if (fs->fs_bsize < fs->fs_fsize) 951 return 0; 952 953 /* Compute fs_bshift and ensure it is consistent */ 954 for (i = fs->fs_bsize; i > 1; i >>= 1) 955 fs_bshift++; 956 if (fs->fs_bshift != fs_bshift) 957 return 0; 958 959 /* Compute fs_fshift and ensure it is consistent */ 960 for (i = fs->fs_fsize; i > 1; i >>= 1) 961 fs_fshift++; 962 if (fs->fs_fshift != fs_fshift) 963 return 0; 964 965 /* Now that the shifts are sanitized, we can use the ffs_ API */ 966 967 /* Check the number of frag blocks */ 968 if ((fs_frag = ffs_numfrags(fs, fs->fs_bsize)) > MAXFRAG) 969 return 0; 970 if (fs->fs_frag != fs_frag) 971 return 0; 972 973 return 1; 974 } 975 976 /* 977 * Common code for mount and mountroot 978 */ 979 int 980 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l) 981 { 982 struct ufsmount *ump = NULL; 983 struct buf *bp = NULL; 984 struct fs *fs = NULL; 985 dev_t dev; 986 struct dkwedge_info dkw; 987 void *space; 988 daddr_t sblockloc = 0; 989 int blks, fstype = 0; 990 int error, i, bsize, ronly, bset = 0; 991 #ifdef FFS_EI 992 int needswap = 0; /* keep gcc happy */ 993 #endif 994 int32_t *lp; 995 kauth_cred_t cred; 996 u_int32_t fs_sbsize = 8192; /* keep gcc happy*/ 997 u_int32_t allocsbsize; 998 999 dev = devvp->v_rdev; 1000 cred = l ? l->l_cred : NOCRED; 1001 1002 /* Flush out any old buffers remaining from a previous use. */ 1003 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 1004 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0); 1005 VOP_UNLOCK(devvp); 1006 if (error) { 1007 DPRINTF(("%s: vinvalbuf %d\n", __func__, error)); 1008 return error; 1009 } 1010 1011 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 1012 1013 error = fstrans_mount(mp); 1014 if (error) { 1015 DPRINTF(("%s: fstrans_mount %d\n", __func__, error)); 1016 return error; 1017 } 1018 1019 ump = kmem_zalloc(sizeof(*ump), KM_SLEEP); 1020 mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE); 1021 error = ffs_snapshot_init(ump); 1022 if (error) { 1023 DPRINTF(("%s: ffs_snapshot_init %d\n", __func__, error)); 1024 goto out; 1025 } 1026 ump->um_ops = &ffs_ufsops; 1027 1028 #ifdef WAPBL 1029 sbagain: 1030 #endif 1031 /* 1032 * Try reading the superblock in each of its possible locations. 1033 */ 1034 for (i = 0; ; i++) { 1035 daddr_t fsblockloc; 1036 1037 if (bp != NULL) { 1038 brelse(bp, BC_NOCACHE); 1039 bp = NULL; 1040 } 1041 if (sblock_try[i] == -1) { 1042 DPRINTF(("%s: sblock_try\n", __func__)); 1043 error = EINVAL; 1044 fs = NULL; 1045 goto out; 1046 } 1047 1048 error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, 1049 cred, 0, &bp); 1050 if (error) { 1051 DPRINTF(("%s: bread@0x%x %d\n", __func__, 1052 sblock_try[i] / DEV_BSIZE, error)); 1053 fs = NULL; 1054 goto out; 1055 } 1056 fs = (struct fs*)bp->b_data; 1057 1058 fsblockloc = sblockloc = sblock_try[i]; 1059 DPRINTF(("%s: fs_magic 0x%x\n", __func__, fs->fs_magic)); 1060 1061 /* 1062 * Swap: here, we swap fs->fs_sbsize in order to get the correct 1063 * size to read the superblock. Once read, we swap the whole 1064 * superblock structure. 1065 */ 1066 if (fs->fs_magic == FS_UFS1_MAGIC) { 1067 fs_sbsize = fs->fs_sbsize; 1068 fstype = UFS1; 1069 #ifdef FFS_EI 1070 needswap = 0; 1071 } else if (fs->fs_magic == FS_UFS1_MAGIC_SWAPPED) { 1072 fs_sbsize = bswap32(fs->fs_sbsize); 1073 fstype = UFS1; 1074 needswap = 1; 1075 #endif 1076 } else if (fs->fs_magic == FS_UFS2_MAGIC) { 1077 fs_sbsize = fs->fs_sbsize; 1078 fstype = UFS2; 1079 #ifdef FFS_EI 1080 needswap = 0; 1081 } else if (fs->fs_magic == FS_UFS2_MAGIC_SWAPPED) { 1082 fs_sbsize = bswap32(fs->fs_sbsize); 1083 fstype = UFS2; 1084 needswap = 1; 1085 #endif 1086 } else 1087 continue; 1088 1089 /* fs->fs_sblockloc isn't defined for old filesystems */ 1090 if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) { 1091 if (sblockloc == SBLOCK_UFS2) 1092 /* 1093 * This is likely to be the first alternate 1094 * in a filesystem with 64k blocks. 1095 * Don't use it. 1096 */ 1097 continue; 1098 fsblockloc = sblockloc; 1099 } else { 1100 fsblockloc = fs->fs_sblockloc; 1101 #ifdef FFS_EI 1102 if (needswap) 1103 fsblockloc = bswap64(fsblockloc); 1104 #endif 1105 } 1106 1107 /* Check we haven't found an alternate superblock */ 1108 if (fsblockloc != sblockloc) 1109 continue; 1110 1111 /* Check the superblock size */ 1112 if (fs_sbsize > SBLOCKSIZE || fs_sbsize < sizeof(struct fs)) 1113 continue; 1114 fs = kmem_alloc((u_long)fs_sbsize, KM_SLEEP); 1115 memcpy(fs, bp->b_data, fs_sbsize); 1116 1117 /* Swap the whole superblock structure, if necessary. */ 1118 #ifdef FFS_EI 1119 if (needswap) { 1120 ffs_sb_swap((struct fs*)bp->b_data, fs); 1121 fs->fs_flags |= FS_SWAPPED; 1122 } else 1123 #endif 1124 fs->fs_flags &= ~FS_SWAPPED; 1125 1126 /* 1127 * Now that everything is swapped, the superblock is ready to 1128 * be sanitized. 1129 */ 1130 if (!ffs_superblock_validate(fs)) { 1131 kmem_free(fs, fs_sbsize); 1132 continue; 1133 } 1134 1135 /* Ok seems to be a good superblock */ 1136 break; 1137 } 1138 1139 ump->um_fs = fs; 1140 1141 #ifdef WAPBL 1142 if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) { 1143 error = ffs_wapbl_replay_start(mp, fs, devvp); 1144 if (error && (mp->mnt_flag & MNT_FORCE) == 0) { 1145 DPRINTF(("%s: ffs_wapbl_replay_start %d\n", __func__, 1146 error)); 1147 goto out; 1148 } 1149 if (!error) { 1150 if (!ronly) { 1151 /* XXX fsmnt may be stale. */ 1152 printf("%s: replaying log to disk\n", 1153 fs->fs_fsmnt); 1154 error = wapbl_replay_write(mp->mnt_wapbl_replay, 1155 devvp); 1156 if (error) { 1157 DPRINTF(("%s: wapbl_replay_write %d\n", 1158 __func__, error)); 1159 goto out; 1160 } 1161 wapbl_replay_stop(mp->mnt_wapbl_replay); 1162 fs->fs_clean = FS_WASCLEAN; 1163 } else { 1164 /* XXX fsmnt may be stale */ 1165 printf("%s: replaying log to memory\n", 1166 fs->fs_fsmnt); 1167 } 1168 1169 /* Force a re-read of the superblock */ 1170 brelse(bp, BC_INVAL); 1171 bp = NULL; 1172 kmem_free(fs, fs_sbsize); 1173 fs = NULL; 1174 goto sbagain; 1175 } 1176 } 1177 #else /* !WAPBL */ 1178 if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) { 1179 error = EPERM; 1180 DPRINTF(("%s: no force %d\n", __func__, error)); 1181 goto out; 1182 } 1183 #endif /* !WAPBL */ 1184 1185 ffs_oldfscompat_read(fs, ump, sblockloc); 1186 ump->um_maxfilesize = fs->fs_maxfilesize; 1187 1188 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) { 1189 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n", 1190 mp->mnt_stat.f_mntonname, fs->fs_flags, 1191 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 1192 if ((mp->mnt_flag & MNT_FORCE) == 0) { 1193 error = EINVAL; 1194 DPRINTF(("%s: no force %d\n", __func__, error)); 1195 goto out; 1196 } 1197 } 1198 1199 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 1200 fs->fs_pendingblocks = 0; 1201 fs->fs_pendinginodes = 0; 1202 } 1203 1204 ump->um_fstype = fstype; 1205 if (fs->fs_sbsize < SBLOCKSIZE) 1206 brelse(bp, BC_INVAL); 1207 else 1208 brelse(bp, 0); 1209 bp = NULL; 1210 1211 /* First check to see if this is tagged as an Apple UFS filesystem 1212 * in the disklabel 1213 */ 1214 if (getdiskinfo(devvp, &dkw) == 0 && 1215 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0) 1216 ump->um_flags |= UFS_ISAPPLEUFS; 1217 #ifdef APPLE_UFS 1218 else { 1219 /* Manually look for an apple ufs label, and if a valid one 1220 * is found, then treat it like an Apple UFS filesystem anyway 1221 */ 1222 error = bread(devvp, 1223 (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE), 1224 APPLEUFS_LABEL_SIZE, cred, 0, &bp); 1225 if (error) { 1226 DPRINTF(("%s: apple bread@0x%jx %d\n", __func__, 1227 (intmax_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE), 1228 error)); 1229 goto out; 1230 } 1231 error = ffs_appleufs_validate(fs->fs_fsmnt, 1232 (struct appleufslabel *)bp->b_data, NULL); 1233 if (error == 0) 1234 ump->um_flags |= UFS_ISAPPLEUFS; 1235 brelse(bp, 0); 1236 bp = NULL; 1237 } 1238 #else 1239 if (ump->um_flags & UFS_ISAPPLEUFS) { 1240 DPRINTF(("%s: bad apple\n", __func__)); 1241 error = EINVAL; 1242 goto out; 1243 } 1244 #endif 1245 1246 #if 0 1247 /* 1248 * XXX This code changes the behaviour of mounting dirty filesystems, to 1249 * XXX require "mount -f ..." to mount them. This doesn't match what 1250 * XXX mount(8) describes and is disabled for now. 1251 */ 1252 /* 1253 * If the file system is not clean, don't allow it to be mounted 1254 * unless MNT_FORCE is specified. (Note: MNT_FORCE is always set 1255 * for the root file system.) 1256 */ 1257 if (fs->fs_flags & FS_DOWAPBL) { 1258 /* 1259 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL 1260 * bit is set, although there's a window in unmount where it 1261 * could be FS_ISCLEAN 1262 */ 1263 if ((mp->mnt_flag & MNT_FORCE) == 0 && 1264 (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) { 1265 error = EPERM; 1266 goto out; 1267 } 1268 } else 1269 if ((fs->fs_clean & FS_ISCLEAN) == 0 && 1270 (mp->mnt_flag & MNT_FORCE) == 0) { 1271 error = EPERM; 1272 goto out; 1273 } 1274 #endif 1275 1276 /* 1277 * Verify that we can access the last block in the fs 1278 * if we're mounting read/write. 1279 */ 1280 1281 if (!ronly) { 1282 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_size - 1), 1283 fs->fs_fsize, cred, 0, &bp); 1284 if (error) { 1285 DPRINTF(("%s: bread@0x%jx %d\n", __func__, 1286 (intmax_t)FFS_FSBTODB(fs, fs->fs_size - 1), 1287 error)); 1288 bset = BC_INVAL; 1289 goto out; 1290 } 1291 if (bp->b_bcount != fs->fs_fsize) { 1292 DPRINTF(("%s: bcount %x != fsize %x\n", __func__, 1293 bp->b_bcount, fs->fs_fsize)); 1294 error = EINVAL; 1295 } 1296 brelse(bp, BC_INVAL); 1297 bp = NULL; 1298 } 1299 1300 fs->fs_ronly = ronly; 1301 /* Don't bump fs_clean if we're replaying journal */ 1302 if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN))) { 1303 if (ronly == 0) { 1304 fs->fs_clean <<= 1; 1305 fs->fs_fmod = 1; 1306 } 1307 } 1308 1309 bsize = fs->fs_cssize; 1310 blks = howmany(bsize, fs->fs_fsize); 1311 if (fs->fs_contigsumsize > 0) 1312 bsize += fs->fs_ncg * sizeof(int32_t); 1313 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1314 allocsbsize = bsize; 1315 space = kmem_alloc((u_long)allocsbsize, KM_SLEEP); 1316 fs->fs_csp = space; 1317 1318 for (i = 0; i < blks; i += fs->fs_frag) { 1319 bsize = fs->fs_bsize; 1320 if (i + fs->fs_frag > blks) 1321 bsize = (blks - i) * fs->fs_fsize; 1322 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize, 1323 cred, 0, &bp); 1324 if (error) { 1325 DPRINTF(("%s: bread@0x%jx %d\n", __func__, 1326 (intmax_t)FFS_FSBTODB(fs, fs->fs_csaddr + i), 1327 error)); 1328 goto out1; 1329 } 1330 #ifdef FFS_EI 1331 if (needswap) 1332 ffs_csum_swap((struct csum *)bp->b_data, 1333 (struct csum *)space, bsize); 1334 else 1335 #endif 1336 memcpy(space, bp->b_data, (u_int)bsize); 1337 1338 space = (char *)space + bsize; 1339 brelse(bp, 0); 1340 bp = NULL; 1341 } 1342 if (fs->fs_contigsumsize > 0) { 1343 fs->fs_maxcluster = lp = space; 1344 for (i = 0; i < fs->fs_ncg; i++) 1345 *lp++ = fs->fs_contigsumsize; 1346 space = lp; 1347 } 1348 bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1349 fs->fs_contigdirs = space; 1350 space = (char *)space + bsize; 1351 memset(fs->fs_contigdirs, 0, bsize); 1352 /* Compatibility for old filesystems - XXX */ 1353 if (fs->fs_avgfilesize <= 0) 1354 fs->fs_avgfilesize = AVFILESIZ; 1355 if (fs->fs_avgfpdir <= 0) 1356 fs->fs_avgfpdir = AFPDIR; 1357 fs->fs_active = NULL; 1358 mp->mnt_data = ump; 1359 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev; 1360 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS); 1361 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0]; 1362 mp->mnt_stat.f_namemax = FFS_MAXNAMLEN; 1363 if (UFS_MPISAPPLEUFS(ump)) { 1364 /* NeXT used to keep short symlinks in the inode even 1365 * when using FS_42INODEFMT. In that case fs->fs_maxsymlinklen 1366 * is probably -1, but we still need to be able to identify 1367 * short symlinks. 1368 */ 1369 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN; 1370 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ; 1371 mp->mnt_iflag |= IMNT_DTYPE; 1372 } else { 1373 ump->um_maxsymlinklen = fs->fs_maxsymlinklen; 1374 ump->um_dirblksiz = UFS_DIRBLKSIZ; 1375 if (ump->um_maxsymlinklen > 0) 1376 mp->mnt_iflag |= IMNT_DTYPE; 1377 else 1378 mp->mnt_iflag &= ~IMNT_DTYPE; 1379 } 1380 mp->mnt_fs_bshift = fs->fs_bshift; 1381 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */ 1382 mp->mnt_flag |= MNT_LOCAL; 1383 mp->mnt_iflag |= IMNT_MPSAFE; 1384 #ifdef FFS_EI 1385 if (needswap) 1386 ump->um_flags |= UFS_NEEDSWAP; 1387 #endif 1388 ump->um_mountp = mp; 1389 ump->um_dev = dev; 1390 ump->um_devvp = devvp; 1391 ump->um_nindir = fs->fs_nindir; 1392 ump->um_lognindir = ffs(fs->fs_nindir) - 1; 1393 ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT; 1394 ump->um_seqinc = fs->fs_frag; 1395 for (i = 0; i < MAXQUOTAS; i++) 1396 ump->um_quotas[i] = NULLVP; 1397 spec_node_setmountedfs(devvp, mp); 1398 if (ronly == 0 && fs->fs_snapinum[0] != 0) 1399 ffs_snapshot_mount(mp); 1400 #ifdef WAPBL 1401 if (!ronly) { 1402 KDASSERT(fs->fs_ronly == 0); 1403 /* 1404 * ffs_wapbl_start() needs mp->mnt_stat initialised if it 1405 * needs to create a new log file in-filesystem. 1406 */ 1407 error = ffs_statvfs(mp, &mp->mnt_stat); 1408 if (error) { 1409 DPRINTF(("%s: ffs_statvfs %d\n", __func__, error)); 1410 goto out1; 1411 } 1412 1413 error = ffs_wapbl_start(mp); 1414 if (error) { 1415 DPRINTF(("%s: ffs_wapbl_start %d\n", __func__, error)); 1416 goto out1; 1417 } 1418 } 1419 #endif /* WAPBL */ 1420 if (ronly == 0) { 1421 #ifdef QUOTA2 1422 error = ffs_quota2_mount(mp); 1423 if (error) { 1424 DPRINTF(("%s: ffs_quota2_mount %d\n", __func__, error)); 1425 goto out1; 1426 } 1427 #else 1428 if (fs->fs_flags & FS_DOQUOTA2) { 1429 ump->um_flags |= UFS_QUOTA2; 1430 uprintf("%s: options QUOTA2 not enabled%s\n", 1431 mp->mnt_stat.f_mntonname, 1432 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 1433 if ((mp->mnt_flag & MNT_FORCE) == 0) { 1434 error = EINVAL; 1435 DPRINTF(("%s: quota disabled %d\n", __func__, 1436 error)); 1437 goto out1; 1438 } 1439 } 1440 #endif 1441 } 1442 1443 if (mp->mnt_flag & MNT_DISCARD) 1444 ump->um_discarddata = ffs_discard_init(devvp, fs); 1445 1446 return (0); 1447 out1: 1448 kmem_free(fs->fs_csp, allocsbsize); 1449 out: 1450 #ifdef WAPBL 1451 if (mp->mnt_wapbl_replay) { 1452 wapbl_replay_stop(mp->mnt_wapbl_replay); 1453 wapbl_replay_free(mp->mnt_wapbl_replay); 1454 mp->mnt_wapbl_replay = 0; 1455 } 1456 #endif 1457 1458 fstrans_unmount(mp); 1459 if (fs) 1460 kmem_free(fs, fs->fs_sbsize); 1461 spec_node_setmountedfs(devvp, NULL); 1462 if (bp) 1463 brelse(bp, bset); 1464 if (ump) { 1465 if (ump->um_oldfscompat) 1466 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t)); 1467 mutex_destroy(&ump->um_lock); 1468 kmem_free(ump, sizeof(*ump)); 1469 mp->mnt_data = NULL; 1470 } 1471 return (error); 1472 } 1473 1474 /* 1475 * Sanity checks for loading old filesystem superblocks. 1476 * See ffs_oldfscompat_write below for unwound actions. 1477 * 1478 * XXX - Parts get retired eventually. 1479 * Unfortunately new bits get added. 1480 */ 1481 static void 1482 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc) 1483 { 1484 off_t maxfilesize; 1485 int32_t *extrasave; 1486 1487 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1488 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1489 return; 1490 1491 if (!ump->um_oldfscompat) 1492 ump->um_oldfscompat = kmem_alloc(512 + 3*sizeof(int32_t), 1493 KM_SLEEP); 1494 1495 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512); 1496 extrasave = ump->um_oldfscompat; 1497 extrasave += 512/sizeof(int32_t); 1498 extrasave[0] = fs->fs_old_npsect; 1499 extrasave[1] = fs->fs_old_interleave; 1500 extrasave[2] = fs->fs_old_trackskew; 1501 1502 /* These fields will be overwritten by their 1503 * original values in fs_oldfscompat_write, so it is harmless 1504 * to modify them here. 1505 */ 1506 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir; 1507 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree; 1508 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree; 1509 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree; 1510 1511 fs->fs_maxbsize = fs->fs_bsize; 1512 fs->fs_time = fs->fs_old_time; 1513 fs->fs_size = fs->fs_old_size; 1514 fs->fs_dsize = fs->fs_old_dsize; 1515 fs->fs_csaddr = fs->fs_old_csaddr; 1516 fs->fs_sblockloc = sblockloc; 1517 1518 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL); 1519 1520 if (fs->fs_old_postblformat == FS_42POSTBLFMT) { 1521 fs->fs_old_nrpos = 8; 1522 fs->fs_old_npsect = fs->fs_old_nsect; 1523 fs->fs_old_interleave = 1; 1524 fs->fs_old_trackskew = 0; 1525 } 1526 1527 if (fs->fs_old_inodefmt < FS_44INODEFMT) { 1528 fs->fs_maxfilesize = (u_quad_t) 1LL << 39; 1529 fs->fs_qbmask = ~fs->fs_bmask; 1530 fs->fs_qfmask = ~fs->fs_fmask; 1531 } 1532 1533 maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1; 1534 if (fs->fs_maxfilesize > maxfilesize) 1535 fs->fs_maxfilesize = maxfilesize; 1536 1537 /* Compatibility for old filesystems */ 1538 if (fs->fs_avgfilesize <= 0) 1539 fs->fs_avgfilesize = AVFILESIZ; 1540 if (fs->fs_avgfpdir <= 0) 1541 fs->fs_avgfpdir = AFPDIR; 1542 1543 #if 0 1544 if (bigcgs) { 1545 fs->fs_save_cgsize = fs->fs_cgsize; 1546 fs->fs_cgsize = fs->fs_bsize; 1547 } 1548 #endif 1549 } 1550 1551 /* 1552 * Unwinding superblock updates for old filesystems. 1553 * See ffs_oldfscompat_read above for details. 1554 * 1555 * XXX - Parts get retired eventually. 1556 * Unfortunately new bits get added. 1557 */ 1558 static void 1559 ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump) 1560 { 1561 int32_t *extrasave; 1562 1563 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1564 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1565 return; 1566 1567 fs->fs_old_time = fs->fs_time; 1568 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir; 1569 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree; 1570 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree; 1571 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree; 1572 fs->fs_old_flags = fs->fs_flags; 1573 1574 #if 0 1575 if (bigcgs) { 1576 fs->fs_cgsize = fs->fs_save_cgsize; 1577 } 1578 #endif 1579 1580 memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512); 1581 extrasave = ump->um_oldfscompat; 1582 extrasave += 512/sizeof(int32_t); 1583 fs->fs_old_npsect = extrasave[0]; 1584 fs->fs_old_interleave = extrasave[1]; 1585 fs->fs_old_trackskew = extrasave[2]; 1586 1587 } 1588 1589 /* 1590 * unmount vfs operation 1591 */ 1592 int 1593 ffs_unmount(struct mount *mp, int mntflags) 1594 { 1595 struct lwp *l = curlwp; 1596 struct ufsmount *ump = VFSTOUFS(mp); 1597 struct fs *fs = ump->um_fs; 1598 int error, flags; 1599 u_int32_t bsize; 1600 #ifdef WAPBL 1601 extern int doforce; 1602 #endif 1603 1604 if (ump->um_discarddata) { 1605 ffs_discard_finish(ump->um_discarddata, mntflags); 1606 ump->um_discarddata = NULL; 1607 } 1608 1609 flags = 0; 1610 if (mntflags & MNT_FORCE) 1611 flags |= FORCECLOSE; 1612 if ((error = ffs_flushfiles(mp, flags, l)) != 0) 1613 return (error); 1614 error = UFS_WAPBL_BEGIN(mp); 1615 if (error == 0) 1616 if (fs->fs_ronly == 0 && 1617 ffs_cgupdate(ump, MNT_WAIT) == 0 && 1618 fs->fs_clean & FS_WASCLEAN) { 1619 fs->fs_clean = FS_ISCLEAN; 1620 fs->fs_fmod = 0; 1621 (void) ffs_sbupdate(ump, MNT_WAIT); 1622 } 1623 if (error == 0) 1624 UFS_WAPBL_END(mp); 1625 #ifdef WAPBL 1626 KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl)); 1627 if (mp->mnt_wapbl_replay) { 1628 KDASSERT(fs->fs_ronly); 1629 wapbl_replay_stop(mp->mnt_wapbl_replay); 1630 wapbl_replay_free(mp->mnt_wapbl_replay); 1631 mp->mnt_wapbl_replay = 0; 1632 } 1633 error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE)); 1634 if (error) { 1635 return error; 1636 } 1637 #endif /* WAPBL */ 1638 1639 if (ump->um_devvp->v_type != VBAD) 1640 spec_node_setmountedfs(ump->um_devvp, NULL); 1641 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1642 (void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE, 1643 NOCRED); 1644 vput(ump->um_devvp); 1645 1646 bsize = fs->fs_cssize; 1647 if (fs->fs_contigsumsize > 0) 1648 bsize += fs->fs_ncg * sizeof(int32_t); 1649 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1650 kmem_free(fs->fs_csp, bsize); 1651 1652 kmem_free(fs, fs->fs_sbsize); 1653 if (ump->um_oldfscompat != NULL) 1654 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t)); 1655 mutex_destroy(&ump->um_lock); 1656 ffs_snapshot_fini(ump); 1657 kmem_free(ump, sizeof(*ump)); 1658 mp->mnt_data = NULL; 1659 mp->mnt_flag &= ~MNT_LOCAL; 1660 fstrans_unmount(mp); 1661 return (0); 1662 } 1663 1664 /* 1665 * Flush out all the files in a filesystem. 1666 */ 1667 int 1668 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l) 1669 { 1670 extern int doforce; 1671 struct ufsmount *ump; 1672 int error; 1673 1674 if (!doforce) 1675 flags &= ~FORCECLOSE; 1676 ump = VFSTOUFS(mp); 1677 #ifdef QUOTA 1678 if ((error = quota1_umount(mp, flags)) != 0) 1679 return (error); 1680 #endif 1681 #ifdef QUOTA2 1682 if ((error = quota2_umount(mp, flags)) != 0) 1683 return (error); 1684 #endif 1685 #ifdef UFS_EXTATTR 1686 if (ump->um_fstype == UFS1) { 1687 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_STARTED) 1688 ufs_extattr_stop(mp, l); 1689 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_INITIALIZED) 1690 ufs_extattr_uepm_destroy(&ump->um_extattr); 1691 mp->mnt_flag &= ~MNT_EXTATTR; 1692 } 1693 #endif 1694 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0) 1695 return (error); 1696 ffs_snapshot_unmount(mp); 1697 /* 1698 * Flush all the files. 1699 */ 1700 error = vflush(mp, NULLVP, flags); 1701 if (error) 1702 return (error); 1703 /* 1704 * Flush filesystem metadata. 1705 */ 1706 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1707 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0); 1708 VOP_UNLOCK(ump->um_devvp); 1709 if (flags & FORCECLOSE) /* XXXDBJ */ 1710 error = 0; 1711 1712 #ifdef WAPBL 1713 if (error) 1714 return error; 1715 if (mp->mnt_wapbl) { 1716 error = wapbl_flush(mp->mnt_wapbl, 1); 1717 if (flags & FORCECLOSE) 1718 error = 0; 1719 } 1720 #endif 1721 1722 return (error); 1723 } 1724 1725 /* 1726 * Get file system statistics. 1727 */ 1728 int 1729 ffs_statvfs(struct mount *mp, struct statvfs *sbp) 1730 { 1731 struct ufsmount *ump; 1732 struct fs *fs; 1733 1734 ump = VFSTOUFS(mp); 1735 fs = ump->um_fs; 1736 mutex_enter(&ump->um_lock); 1737 sbp->f_bsize = fs->fs_bsize; 1738 sbp->f_frsize = fs->fs_fsize; 1739 sbp->f_iosize = fs->fs_bsize; 1740 sbp->f_blocks = fs->fs_dsize; 1741 sbp->f_bfree = ffs_blkstofrags(fs, fs->fs_cstotal.cs_nbfree) + 1742 fs->fs_cstotal.cs_nffree + FFS_DBTOFSB(fs, fs->fs_pendingblocks); 1743 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t) 1744 fs->fs_minfree) / (u_int64_t) 100; 1745 if (sbp->f_bfree > sbp->f_bresvd) 1746 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd; 1747 else 1748 sbp->f_bavail = 0; 1749 sbp->f_files = fs->fs_ncg * fs->fs_ipg - UFS_ROOTINO; 1750 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes; 1751 sbp->f_favail = sbp->f_ffree; 1752 sbp->f_fresvd = 0; 1753 mutex_exit(&ump->um_lock); 1754 copy_statvfs_info(sbp, mp); 1755 1756 return (0); 1757 } 1758 1759 struct ffs_sync_ctx { 1760 int waitfor; 1761 bool is_suspending; 1762 }; 1763 1764 static bool 1765 ffs_sync_selector(void *cl, struct vnode *vp) 1766 { 1767 struct ffs_sync_ctx *c = cl; 1768 struct inode *ip; 1769 1770 ip = VTOI(vp); 1771 /* 1772 * Skip the vnode/inode if inaccessible. 1773 */ 1774 if (ip == NULL || vp->v_type == VNON) 1775 return false; 1776 1777 /* 1778 * We deliberately update inode times here. This will 1779 * prevent a massive queue of updates accumulating, only 1780 * to be handled by a call to unmount. 1781 * 1782 * XXX It would be better to have the syncer trickle these 1783 * out. Adjustment needed to allow registering vnodes for 1784 * sync when the vnode is clean, but the inode dirty. Or 1785 * have ufs itself trickle out inode updates. 1786 * 1787 * If doing a lazy sync, we don't care about metadata or 1788 * data updates, because they are handled by each vnode's 1789 * synclist entry. In this case we are only interested in 1790 * writing back modified inodes. 1791 */ 1792 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | 1793 IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 && 1794 (c->waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) && 1795 UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) 1796 return false; 1797 1798 if (vp->v_type == VBLK && c->is_suspending) 1799 return false; 1800 1801 return true; 1802 } 1803 1804 /* 1805 * Go through the disk queues to initiate sandbagged IO; 1806 * go through the inodes to write those that have been modified; 1807 * initiate the writing of the super block if it has been modified. 1808 * 1809 * Note: we are always called with the filesystem marked `MPBUSY'. 1810 */ 1811 int 1812 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred) 1813 { 1814 struct vnode *vp; 1815 struct ufsmount *ump = VFSTOUFS(mp); 1816 struct fs *fs; 1817 struct vnode_iterator *marker; 1818 int error, allerror = 0; 1819 bool is_suspending; 1820 struct ffs_sync_ctx ctx; 1821 1822 fs = ump->um_fs; 1823 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */ 1824 printf("fs = %s\n", fs->fs_fsmnt); 1825 panic("update: rofs mod"); 1826 } 1827 1828 fstrans_start(mp, FSTRANS_SHARED); 1829 is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING); 1830 /* 1831 * Write back each (modified) inode. 1832 */ 1833 vfs_vnode_iterator_init(mp, &marker); 1834 1835 ctx.waitfor = waitfor; 1836 ctx.is_suspending = is_suspending; 1837 while ((vp = vfs_vnode_iterator_next(marker, ffs_sync_selector, &ctx))) 1838 { 1839 error = vn_lock(vp, LK_EXCLUSIVE); 1840 if (error) { 1841 vrele(vp); 1842 continue; 1843 } 1844 if (waitfor == MNT_LAZY) { 1845 error = UFS_WAPBL_BEGIN(vp->v_mount); 1846 if (!error) { 1847 error = ffs_update(vp, NULL, NULL, 1848 UPDATE_CLOSE); 1849 UFS_WAPBL_END(vp->v_mount); 1850 } 1851 } else { 1852 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG | 1853 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0); 1854 } 1855 if (error) 1856 allerror = error; 1857 vput(vp); 1858 } 1859 vfs_vnode_iterator_destroy(marker); 1860 1861 /* 1862 * Force stale file system control information to be flushed. 1863 */ 1864 if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 || 1865 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) { 1866 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1867 if ((error = VOP_FSYNC(ump->um_devvp, cred, 1868 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG, 1869 0, 0)) != 0) 1870 allerror = error; 1871 VOP_UNLOCK(ump->um_devvp); 1872 } 1873 #if defined(QUOTA) || defined(QUOTA2) 1874 qsync(mp); 1875 #endif 1876 /* 1877 * Write back modified superblock. 1878 */ 1879 if (fs->fs_fmod != 0) { 1880 fs->fs_fmod = 0; 1881 fs->fs_time = time_second; 1882 error = UFS_WAPBL_BEGIN(mp); 1883 if (error) 1884 allerror = error; 1885 else { 1886 if ((error = ffs_cgupdate(ump, waitfor))) 1887 allerror = error; 1888 UFS_WAPBL_END(mp); 1889 } 1890 } 1891 1892 #ifdef WAPBL 1893 if (mp->mnt_wapbl) { 1894 error = wapbl_flush(mp->mnt_wapbl, 0); 1895 if (error) 1896 allerror = error; 1897 } 1898 #endif 1899 1900 fstrans_done(mp); 1901 return (allerror); 1902 } 1903 1904 /* 1905 * Read an inode from disk and initialize this vnode / inode pair. 1906 * Caller assures no other thread will try to load this inode. 1907 */ 1908 int 1909 ffs_loadvnode(struct mount *mp, struct vnode *vp, 1910 const void *key, size_t key_len, const void **new_key) 1911 { 1912 ino_t ino; 1913 struct fs *fs; 1914 struct inode *ip; 1915 struct ufsmount *ump; 1916 struct buf *bp; 1917 dev_t dev; 1918 int error; 1919 1920 KASSERT(key_len == sizeof(ino)); 1921 memcpy(&ino, key, key_len); 1922 ump = VFSTOUFS(mp); 1923 dev = ump->um_dev; 1924 fs = ump->um_fs; 1925 1926 /* Read in the disk contents for the inode. */ 1927 error = bread(ump->um_devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ino)), 1928 (int)fs->fs_bsize, NOCRED, 0, &bp); 1929 if (error) 1930 return error; 1931 1932 /* Allocate and initialize inode. */ 1933 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK); 1934 memset(ip, 0, sizeof(struct inode)); 1935 vp->v_tag = VT_UFS; 1936 vp->v_op = ffs_vnodeop_p; 1937 vp->v_vflag |= VV_LOCKSWORK; 1938 vp->v_data = ip; 1939 ip->i_vnode = vp; 1940 ip->i_ump = ump; 1941 ip->i_fs = fs; 1942 ip->i_dev = dev; 1943 ip->i_number = ino; 1944 #if defined(QUOTA) || defined(QUOTA2) 1945 ufsquota_init(ip); 1946 #endif 1947 1948 /* Initialize genfs node. */ 1949 genfs_node_init(vp, &ffs_genfsops); 1950 1951 if (ip->i_ump->um_fstype == UFS1) 1952 ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache, 1953 PR_WAITOK); 1954 else 1955 ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache, 1956 PR_WAITOK); 1957 ffs_load_inode(bp, ip, fs, ino); 1958 brelse(bp, 0); 1959 1960 /* Initialize the vnode from the inode. */ 1961 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp); 1962 1963 /* Finish inode initialization. */ 1964 ip->i_devvp = ump->um_devvp; 1965 vref(ip->i_devvp); 1966 1967 /* 1968 * Ensure that uid and gid are correct. This is a temporary 1969 * fix until fsck has been changed to do the update. 1970 */ 1971 1972 if (fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */ 1973 ip->i_uid = ip->i_ffs1_ouid; /* XXX */ 1974 ip->i_gid = ip->i_ffs1_ogid; /* XXX */ 1975 } /* XXX */ 1976 uvm_vnp_setsize(vp, ip->i_size); 1977 *new_key = &ip->i_number; 1978 return 0; 1979 } 1980 1981 /* 1982 * File handle to vnode 1983 * 1984 * Have to be really careful about stale file handles: 1985 * - check that the inode number is valid 1986 * - call ffs_vget() to get the locked inode 1987 * - check for an unallocated inode (i_mode == 0) 1988 * - check that the given client host has export rights and return 1989 * those rights via. exflagsp and credanonp 1990 */ 1991 int 1992 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp) 1993 { 1994 struct ufid ufh; 1995 struct fs *fs; 1996 1997 if (fhp->fid_len != sizeof(struct ufid)) 1998 return EINVAL; 1999 2000 memcpy(&ufh, fhp, sizeof(ufh)); 2001 fs = VFSTOUFS(mp)->um_fs; 2002 if (ufh.ufid_ino < UFS_ROOTINO || 2003 ufh.ufid_ino >= fs->fs_ncg * fs->fs_ipg) 2004 return (ESTALE); 2005 return (ufs_fhtovp(mp, &ufh, vpp)); 2006 } 2007 2008 /* 2009 * Vnode pointer to File handle 2010 */ 2011 /* ARGSUSED */ 2012 int 2013 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size) 2014 { 2015 struct inode *ip; 2016 struct ufid ufh; 2017 2018 if (*fh_size < sizeof(struct ufid)) { 2019 *fh_size = sizeof(struct ufid); 2020 return E2BIG; 2021 } 2022 ip = VTOI(vp); 2023 *fh_size = sizeof(struct ufid); 2024 memset(&ufh, 0, sizeof(ufh)); 2025 ufh.ufid_len = sizeof(struct ufid); 2026 ufh.ufid_ino = ip->i_number; 2027 ufh.ufid_gen = ip->i_gen; 2028 memcpy(fhp, &ufh, sizeof(ufh)); 2029 return (0); 2030 } 2031 2032 void 2033 ffs_init(void) 2034 { 2035 if (ffs_initcount++ > 0) 2036 return; 2037 2038 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0, 2039 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL); 2040 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0, 2041 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL); 2042 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0, 2043 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL); 2044 ufs_init(); 2045 } 2046 2047 void 2048 ffs_reinit(void) 2049 { 2050 2051 ufs_reinit(); 2052 } 2053 2054 void 2055 ffs_done(void) 2056 { 2057 if (--ffs_initcount > 0) 2058 return; 2059 2060 ufs_done(); 2061 pool_cache_destroy(ffs_dinode2_cache); 2062 pool_cache_destroy(ffs_dinode1_cache); 2063 pool_cache_destroy(ffs_inode_cache); 2064 } 2065 2066 /* 2067 * Write a superblock and associated information back to disk. 2068 */ 2069 int 2070 ffs_sbupdate(struct ufsmount *mp, int waitfor) 2071 { 2072 struct fs *fs = mp->um_fs; 2073 struct buf *bp; 2074 int error = 0; 2075 u_int32_t saveflag; 2076 2077 error = ffs_getblk(mp->um_devvp, 2078 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK, 2079 fs->fs_sbsize, false, &bp); 2080 if (error) 2081 return error; 2082 saveflag = fs->fs_flags & FS_INTERNAL; 2083 fs->fs_flags &= ~FS_INTERNAL; 2084 2085 memcpy(bp->b_data, fs, fs->fs_sbsize); 2086 2087 ffs_oldfscompat_write((struct fs *)bp->b_data, mp); 2088 #ifdef FFS_EI 2089 if (mp->um_flags & UFS_NEEDSWAP) 2090 ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data); 2091 #endif 2092 fs->fs_flags |= saveflag; 2093 2094 if (waitfor == MNT_WAIT) 2095 error = bwrite(bp); 2096 else 2097 bawrite(bp); 2098 return (error); 2099 } 2100 2101 int 2102 ffs_cgupdate(struct ufsmount *mp, int waitfor) 2103 { 2104 struct fs *fs = mp->um_fs; 2105 struct buf *bp; 2106 int blks; 2107 void *space; 2108 int i, size, error = 0, allerror = 0; 2109 2110 allerror = ffs_sbupdate(mp, waitfor); 2111 blks = howmany(fs->fs_cssize, fs->fs_fsize); 2112 space = fs->fs_csp; 2113 for (i = 0; i < blks; i += fs->fs_frag) { 2114 size = fs->fs_bsize; 2115 if (i + fs->fs_frag > blks) 2116 size = (blks - i) * fs->fs_fsize; 2117 error = ffs_getblk(mp->um_devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), 2118 FFS_NOBLK, size, false, &bp); 2119 if (error) 2120 break; 2121 #ifdef FFS_EI 2122 if (mp->um_flags & UFS_NEEDSWAP) 2123 ffs_csum_swap((struct csum*)space, 2124 (struct csum*)bp->b_data, size); 2125 else 2126 #endif 2127 memcpy(bp->b_data, space, (u_int)size); 2128 space = (char *)space + size; 2129 if (waitfor == MNT_WAIT) 2130 error = bwrite(bp); 2131 else 2132 bawrite(bp); 2133 } 2134 if (!allerror && error) 2135 allerror = error; 2136 return (allerror); 2137 } 2138 2139 int 2140 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp, 2141 int attrnamespace, const char *attrname) 2142 { 2143 #ifdef UFS_EXTATTR 2144 /* 2145 * File-backed extended attributes are only supported on UFS1. 2146 * UFS2 has native extended attributes. 2147 */ 2148 if (VFSTOUFS(mp)->um_fstype == UFS1) 2149 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname)); 2150 #endif 2151 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname)); 2152 } 2153 2154 int 2155 ffs_suspendctl(struct mount *mp, int cmd) 2156 { 2157 int error; 2158 struct lwp *l = curlwp; 2159 2160 switch (cmd) { 2161 case SUSPEND_SUSPEND: 2162 if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0) 2163 return error; 2164 error = ffs_sync(mp, MNT_WAIT, l->l_proc->p_cred); 2165 if (error == 0) 2166 error = fstrans_setstate(mp, FSTRANS_SUSPENDED); 2167 #ifdef WAPBL 2168 if (error == 0 && mp->mnt_wapbl) 2169 error = wapbl_flush(mp->mnt_wapbl, 1); 2170 #endif 2171 if (error != 0) { 2172 (void) fstrans_setstate(mp, FSTRANS_NORMAL); 2173 return error; 2174 } 2175 return 0; 2176 2177 case SUSPEND_RESUME: 2178 return fstrans_setstate(mp, FSTRANS_NORMAL); 2179 2180 default: 2181 return EINVAL; 2182 } 2183 } 2184 2185 /* 2186 * Synch vnode for a mounted file system. 2187 */ 2188 static int 2189 ffs_vfs_fsync(vnode_t *vp, int flags) 2190 { 2191 int error, i, pflags; 2192 #ifdef WAPBL 2193 struct mount *mp; 2194 #endif 2195 2196 KASSERT(vp->v_type == VBLK); 2197 KASSERT(spec_node_getmountedfs(vp) != NULL); 2198 2199 /* 2200 * Flush all dirty data associated with the vnode. 2201 */ 2202 pflags = PGO_ALLPAGES | PGO_CLEANIT; 2203 if ((flags & FSYNC_WAIT) != 0) 2204 pflags |= PGO_SYNCIO; 2205 mutex_enter(vp->v_interlock); 2206 error = VOP_PUTPAGES(vp, 0, 0, pflags); 2207 if (error) 2208 return error; 2209 2210 #ifdef WAPBL 2211 mp = spec_node_getmountedfs(vp); 2212 if (mp && mp->mnt_wapbl) { 2213 /* 2214 * Don't bother writing out metadata if the syncer is 2215 * making the request. We will let the sync vnode 2216 * write it out in a single burst through a call to 2217 * VFS_SYNC(). 2218 */ 2219 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0) 2220 return 0; 2221 2222 /* 2223 * Don't flush the log if the vnode being flushed 2224 * contains no dirty buffers that could be in the log. 2225 */ 2226 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 2227 error = wapbl_flush(mp->mnt_wapbl, 0); 2228 if (error) 2229 return error; 2230 } 2231 2232 if ((flags & FSYNC_WAIT) != 0) { 2233 mutex_enter(vp->v_interlock); 2234 while (vp->v_numoutput) 2235 cv_wait(&vp->v_cv, vp->v_interlock); 2236 mutex_exit(vp->v_interlock); 2237 } 2238 2239 return 0; 2240 } 2241 #endif /* WAPBL */ 2242 2243 error = vflushbuf(vp, flags); 2244 if (error == 0 && (flags & FSYNC_CACHE) != 0) { 2245 i = 1; 2246 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE, 2247 kauth_cred_get()); 2248 } 2249 2250 return error; 2251 } 2252
Indexes created Thu Oct 02 14:10:14 GMT 2025