ffs_vfsops.c revision 1.331
1 /* $NetBSD: ffs_vfsops.c,v 1.331 2015/05/18 08:07:29 martin 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 1.331 2015/05/18 08:07:29 martin 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 ffs_vfs_fsync(vnode_t *, int); 115 static int ffs_superblock_validate(struct fs *); 116 static int ffs_is_appleufs(struct vnode *, struct fs *); 117 118 static int ffs_init_vnode(struct ufsmount *, struct vnode *, ino_t); 119 static void ffs_deinit_vnode(struct ufsmount *, struct vnode *); 120 121 static struct sysctllog *ffs_sysctl_log; 122 123 static kauth_listener_t ffs_snapshot_listener; 124 125 /* how many times ffs_init() was called */ 126 int ffs_initcount = 0; 127 128 #ifdef DEBUG_FFS_MOUNT 129 #define DPRINTF(_fmt, args...) printf("%s: " _fmt "\n", __func__, ##args) 130 #else 131 #define DPRINTF(_fmt, args...) do {} while (/*CONSTCOND*/0) 132 #endif 133 134 extern const struct vnodeopv_desc ffs_vnodeop_opv_desc; 135 extern const struct vnodeopv_desc ffs_specop_opv_desc; 136 extern const struct vnodeopv_desc ffs_fifoop_opv_desc; 137 138 const struct vnodeopv_desc * const ffs_vnodeopv_descs[] = { 139 &ffs_vnodeop_opv_desc, 140 &ffs_specop_opv_desc, 141 &ffs_fifoop_opv_desc, 142 NULL, 143 }; 144 145 struct vfsops ffs_vfsops = { 146 .vfs_name = MOUNT_FFS, 147 .vfs_min_mount_data = sizeof (struct ufs_args), 148 .vfs_mount = ffs_mount, 149 .vfs_start = ufs_start, 150 .vfs_unmount = ffs_unmount, 151 .vfs_root = ufs_root, 152 .vfs_quotactl = ufs_quotactl, 153 .vfs_statvfs = ffs_statvfs, 154 .vfs_sync = ffs_sync, 155 .vfs_vget = ufs_vget, 156 .vfs_loadvnode = ffs_loadvnode, 157 .vfs_newvnode = ffs_newvnode, 158 .vfs_fhtovp = ffs_fhtovp, 159 .vfs_vptofh = ffs_vptofh, 160 .vfs_init = ffs_init, 161 .vfs_reinit = ffs_reinit, 162 .vfs_done = ffs_done, 163 .vfs_mountroot = ffs_mountroot, 164 .vfs_snapshot = ffs_snapshot, 165 .vfs_extattrctl = ffs_extattrctl, 166 .vfs_suspendctl = ffs_suspendctl, 167 .vfs_renamelock_enter = genfs_renamelock_enter, 168 .vfs_renamelock_exit = genfs_renamelock_exit, 169 .vfs_fsync = ffs_vfs_fsync, 170 .vfs_opv_descs = ffs_vnodeopv_descs 171 }; 172 173 static const struct genfs_ops ffs_genfsops = { 174 .gop_size = ffs_gop_size, 175 .gop_alloc = ufs_gop_alloc, 176 .gop_write = genfs_gop_write, 177 .gop_markupdate = ufs_gop_markupdate, 178 }; 179 180 static const struct ufs_ops ffs_ufsops = { 181 .uo_itimes = ffs_itimes, 182 .uo_update = ffs_update, 183 .uo_truncate = ffs_truncate, 184 .uo_balloc = ffs_balloc, 185 .uo_snapgone = ffs_snapgone, 186 .uo_bufrd = ffs_bufrd, 187 .uo_bufwr = ffs_bufwr, 188 }; 189 190 static int 191 ffs_snapshot_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 192 void *arg0, void *arg1, void *arg2, void *arg3) 193 { 194 vnode_t *vp = arg2; 195 int result = KAUTH_RESULT_DEFER; 196 197 if (action != KAUTH_SYSTEM_FS_SNAPSHOT) 198 return result; 199 200 if (VTOI(vp)->i_uid == kauth_cred_geteuid(cred)) 201 result = KAUTH_RESULT_ALLOW; 202 203 return result; 204 } 205 206 static int 207 ffs_modcmd(modcmd_t cmd, void *arg) 208 { 209 int error; 210 211 #if 0 212 extern int doasyncfree; 213 #endif 214 #ifdef UFS_EXTATTR 215 extern int ufs_extattr_autocreate; 216 #endif 217 extern int ffs_log_changeopt; 218 219 switch (cmd) { 220 case MODULE_CMD_INIT: 221 error = vfs_attach(&ffs_vfsops); 222 if (error != 0) 223 break; 224 225 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 226 CTLFLAG_PERMANENT, 227 CTLTYPE_NODE, "ffs", 228 SYSCTL_DESCR("Berkeley Fast File System"), 229 NULL, 0, NULL, 0, 230 CTL_VFS, 1, CTL_EOL); 231 /* 232 * @@@ should we even bother with these first three? 233 */ 234 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 235 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 236 CTLTYPE_INT, "doclusterread", NULL, 237 sysctl_notavail, 0, NULL, 0, 238 CTL_VFS, 1, FFS_CLUSTERREAD, CTL_EOL); 239 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 240 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 241 CTLTYPE_INT, "doclusterwrite", NULL, 242 sysctl_notavail, 0, NULL, 0, 243 CTL_VFS, 1, FFS_CLUSTERWRITE, CTL_EOL); 244 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 245 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 246 CTLTYPE_INT, "doreallocblks", NULL, 247 sysctl_notavail, 0, NULL, 0, 248 CTL_VFS, 1, FFS_REALLOCBLKS, CTL_EOL); 249 #if 0 250 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 251 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 252 CTLTYPE_INT, "doasyncfree", 253 SYSCTL_DESCR("Release dirty blocks asynchronously"), 254 NULL, 0, &doasyncfree, 0, 255 CTL_VFS, 1, FFS_ASYNCFREE, CTL_EOL); 256 #endif 257 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 258 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 259 CTLTYPE_INT, "log_changeopt", 260 SYSCTL_DESCR("Log changes in optimization strategy"), 261 NULL, 0, &ffs_log_changeopt, 0, 262 CTL_VFS, 1, FFS_LOG_CHANGEOPT, CTL_EOL); 263 #ifdef UFS_EXTATTR 264 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 265 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 266 CTLTYPE_INT, "extattr_autocreate", 267 SYSCTL_DESCR("Size of attribute for " 268 "backing file autocreation"), 269 NULL, 0, &ufs_extattr_autocreate, 0, 270 CTL_VFS, 1, FFS_EXTATTR_AUTOCREATE, CTL_EOL); 271 272 #endif /* UFS_EXTATTR */ 273 274 ffs_snapshot_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM, 275 ffs_snapshot_cb, NULL); 276 if (ffs_snapshot_listener == NULL) 277 printf("ffs_modcmd: can't listen on system scope.\n"); 278 279 break; 280 case MODULE_CMD_FINI: 281 error = vfs_detach(&ffs_vfsops); 282 if (error != 0) 283 break; 284 sysctl_teardown(&ffs_sysctl_log); 285 if (ffs_snapshot_listener != NULL) 286 kauth_unlisten_scope(ffs_snapshot_listener); 287 break; 288 default: 289 error = ENOTTY; 290 break; 291 } 292 293 return (error); 294 } 295 296 pool_cache_t ffs_inode_cache; 297 pool_cache_t ffs_dinode1_cache; 298 pool_cache_t ffs_dinode2_cache; 299 300 static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t); 301 static void ffs_oldfscompat_write(struct fs *, struct ufsmount *); 302 303 /* 304 * Called by main() when ffs is going to be mounted as root. 305 */ 306 307 int 308 ffs_mountroot(void) 309 { 310 struct fs *fs; 311 struct mount *mp; 312 struct lwp *l = curlwp; /* XXX */ 313 struct ufsmount *ump; 314 int error; 315 316 if (device_class(root_device) != DV_DISK) 317 return (ENODEV); 318 319 if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp))) { 320 vrele(rootvp); 321 return (error); 322 } 323 324 /* 325 * We always need to be able to mount the root file system. 326 */ 327 mp->mnt_flag |= MNT_FORCE; 328 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) { 329 vfs_unbusy(mp, false, NULL); 330 vfs_destroy(mp); 331 return (error); 332 } 333 mp->mnt_flag &= ~MNT_FORCE; 334 mountlist_append(mp); 335 ump = VFSTOUFS(mp); 336 fs = ump->um_fs; 337 memset(fs->fs_fsmnt, 0, sizeof(fs->fs_fsmnt)); 338 (void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0); 339 (void)ffs_statvfs(mp, &mp->mnt_stat); 340 vfs_unbusy(mp, false, NULL); 341 setrootfstime((time_t)fs->fs_time); 342 return (0); 343 } 344 345 /* 346 * VFS Operations. 347 * 348 * mount system call 349 */ 350 int 351 ffs_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 352 { 353 struct lwp *l = curlwp; 354 struct vnode *devvp = NULL; 355 struct ufs_args *args = data; 356 struct ufsmount *ump = NULL; 357 struct fs *fs; 358 int error = 0, flags, update; 359 mode_t accessmode; 360 361 if (args == NULL) { 362 DPRINTF("NULL args"); 363 return EINVAL; 364 } 365 if (*data_len < sizeof(*args)) { 366 DPRINTF("bad size args %zu != %zu", *data_len, sizeof(*args)); 367 return EINVAL; 368 } 369 370 if (mp->mnt_flag & MNT_GETARGS) { 371 ump = VFSTOUFS(mp); 372 if (ump == NULL) { 373 DPRINTF("no ump"); 374 return EIO; 375 } 376 args->fspec = NULL; 377 *data_len = sizeof *args; 378 return 0; 379 } 380 381 update = mp->mnt_flag & MNT_UPDATE; 382 383 /* Check arguments */ 384 if (args->fspec != NULL) { 385 /* 386 * Look up the name and verify that it's sane. 387 */ 388 error = namei_simple_user(args->fspec, 389 NSM_FOLLOW_NOEMULROOT, &devvp); 390 if (error != 0) { 391 DPRINTF("namei_simple_user returned %d", 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("non block device %d", devvp->v_type); 401 error = ENOTBLK; 402 } else if (bdevsw_lookup(devvp->v_rdev) == NULL) { 403 DPRINTF("can't find block device 0x%jx", 404 devvp->v_rdev); 405 error = ENXIO; 406 } 407 } else { 408 /* 409 * Be sure we're still naming the same device 410 * used for our initial mount 411 */ 412 ump = VFSTOUFS(mp); 413 if (devvp != ump->um_devvp) { 414 if (devvp->v_rdev != ump->um_devvp->v_rdev) { 415 DPRINTF("wrong device 0x%jx != 0x%jx", 416 (uintmax_t)devvp->v_rdev, 417 (uintmax_t)ump->um_devvp->v_rdev); 418 error = EINVAL; 419 } else { 420 vrele(devvp); 421 devvp = ump->um_devvp; 422 vref(devvp); 423 } 424 } 425 } 426 } else { 427 if (!update) { 428 /* New mounts must have a filename for the device */ 429 DPRINTF("no filename for mount"); 430 return EINVAL; 431 } else { 432 /* Use the extant mount */ 433 ump = VFSTOUFS(mp); 434 devvp = ump->um_devvp; 435 vref(devvp); 436 } 437 } 438 439 /* 440 * If mount by non-root, then verify that user has necessary 441 * permissions on the device. 442 * 443 * Permission to update a mount is checked higher, so here we presume 444 * updating the mount is okay (for example, as far as securelevel goes) 445 * which leaves us with the normal check. 446 */ 447 if (error == 0) { 448 accessmode = VREAD; 449 if (update ? 450 (mp->mnt_iflag & IMNT_WANTRDWR) != 0 : 451 (mp->mnt_flag & MNT_RDONLY) == 0) 452 accessmode |= VWRITE; 453 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 454 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 455 KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, 456 KAUTH_ARG(accessmode)); 457 if (error) { 458 DPRINTF("kauth returned %d", error); 459 } 460 VOP_UNLOCK(devvp); 461 } 462 463 if (error) { 464 vrele(devvp); 465 return (error); 466 } 467 468 #ifdef WAPBL 469 /* WAPBL can only be enabled on a r/w mount. */ 470 if ((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) { 471 mp->mnt_flag &= ~MNT_LOG; 472 } 473 #else /* !WAPBL */ 474 mp->mnt_flag &= ~MNT_LOG; 475 #endif /* !WAPBL */ 476 477 if (!update) { 478 int xflags; 479 480 if (mp->mnt_flag & MNT_RDONLY) 481 xflags = FREAD; 482 else 483 xflags = FREAD | FWRITE; 484 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 485 error = VOP_OPEN(devvp, xflags, FSCRED); 486 VOP_UNLOCK(devvp); 487 if (error) { 488 DPRINTF("VOP_OPEN returned %d", error); 489 goto fail; 490 } 491 error = ffs_mountfs(devvp, mp, l); 492 if (error) { 493 DPRINTF("ffs_mountfs returned %d", error); 494 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 495 (void)VOP_CLOSE(devvp, xflags, NOCRED); 496 VOP_UNLOCK(devvp); 497 goto fail; 498 } 499 500 ump = VFSTOUFS(mp); 501 fs = ump->um_fs; 502 } else { 503 /* 504 * Update the mount. 505 */ 506 507 /* 508 * The initial mount got a reference on this 509 * device, so drop the one obtained via 510 * namei(), above. 511 */ 512 vrele(devvp); 513 514 ump = VFSTOUFS(mp); 515 fs = ump->um_fs; 516 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 517 /* 518 * Changing from r/w to r/o 519 */ 520 flags = WRITECLOSE; 521 if (mp->mnt_flag & MNT_FORCE) 522 flags |= FORCECLOSE; 523 error = ffs_flushfiles(mp, flags, l); 524 if (error == 0) 525 error = UFS_WAPBL_BEGIN(mp); 526 if (error == 0 && 527 ffs_cgupdate(ump, MNT_WAIT) == 0 && 528 fs->fs_clean & FS_WASCLEAN) { 529 if (mp->mnt_flag & MNT_SOFTDEP) 530 fs->fs_flags &= ~FS_DOSOFTDEP; 531 fs->fs_clean = FS_ISCLEAN; 532 (void) ffs_sbupdate(ump, MNT_WAIT); 533 } 534 if (error) { 535 DPRINTF("wapbl %d", error); 536 return error; 537 } 538 UFS_WAPBL_END(mp); 539 } 540 541 #ifdef WAPBL 542 if ((mp->mnt_flag & MNT_LOG) == 0) { 543 error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE); 544 if (error) { 545 DPRINTF("ffs_wapbl_stop returned %d", error); 546 return error; 547 } 548 } 549 #endif /* WAPBL */ 550 551 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 552 /* 553 * Finish change from r/w to r/o 554 */ 555 fs->fs_ronly = 1; 556 fs->fs_fmod = 0; 557 } 558 559 if (mp->mnt_flag & MNT_RELOAD) { 560 error = ffs_reload(mp, l->l_cred, l); 561 if (error) { 562 DPRINTF("ffs_reload returned %d", error); 563 return error; 564 } 565 } 566 567 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) { 568 /* 569 * Changing from read-only to read/write 570 */ 571 #ifndef QUOTA2 572 if (fs->fs_flags & FS_DOQUOTA2) { 573 ump->um_flags |= UFS_QUOTA2; 574 uprintf("%s: options QUOTA2 not enabled%s\n", 575 mp->mnt_stat.f_mntonname, 576 (mp->mnt_flag & MNT_FORCE) ? "" : 577 ", not mounting"); 578 DPRINTF("ffs_quota2 %d", EINVAL); 579 return EINVAL; 580 } 581 #endif 582 fs->fs_ronly = 0; 583 fs->fs_clean <<= 1; 584 fs->fs_fmod = 1; 585 #ifdef WAPBL 586 if (fs->fs_flags & FS_DOWAPBL) { 587 const char *nm = mp->mnt_stat.f_mntonname; 588 if (!mp->mnt_wapbl_replay) { 589 printf("%s: log corrupted;" 590 " replay cancelled\n", nm); 591 return EFTYPE; 592 } 593 printf("%s: replaying log to disk\n", nm); 594 error = wapbl_replay_write(mp->mnt_wapbl_replay, 595 devvp); 596 if (error) { 597 DPRINTF("%s: wapbl_replay_write %d", 598 nm, error); 599 return error; 600 } 601 wapbl_replay_stop(mp->mnt_wapbl_replay); 602 fs->fs_clean = FS_WASCLEAN; 603 } 604 #endif /* WAPBL */ 605 if (fs->fs_snapinum[0] != 0) 606 ffs_snapshot_mount(mp); 607 } 608 609 #ifdef WAPBL 610 error = ffs_wapbl_start(mp); 611 if (error) { 612 DPRINTF("ffs_wapbl_start returned %d", error); 613 return error; 614 } 615 #endif /* WAPBL */ 616 617 #ifdef QUOTA2 618 if (!fs->fs_ronly) { 619 error = ffs_quota2_mount(mp); 620 if (error) { 621 DPRINTF("ffs_quota2_mount returned %d", error); 622 return error; 623 } 624 } 625 #endif 626 627 if ((mp->mnt_flag & MNT_DISCARD) && !(ump->um_discarddata)) 628 ump->um_discarddata = ffs_discard_init(devvp, fs); 629 630 if (args->fspec == NULL) 631 return 0; 632 } 633 634 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, 635 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l); 636 if (error == 0) 637 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, 638 sizeof(fs->fs_fsmnt)); 639 else { 640 DPRINTF("set_statvfs_info returned %d", error); 641 } 642 fs->fs_flags &= ~FS_DOSOFTDEP; 643 if (fs->fs_fmod != 0) { /* XXX */ 644 int err; 645 646 fs->fs_fmod = 0; 647 if (fs->fs_clean & FS_WASCLEAN) 648 fs->fs_time = time_second; 649 else { 650 printf("%s: file system not clean (fs_clean=%#x); " 651 "please fsck(8)\n", mp->mnt_stat.f_mntfromname, 652 fs->fs_clean); 653 printf("%s: lost blocks %" PRId64 " files %d\n", 654 mp->mnt_stat.f_mntfromname, fs->fs_pendingblocks, 655 fs->fs_pendinginodes); 656 } 657 err = UFS_WAPBL_BEGIN(mp); 658 if (err == 0) { 659 (void) ffs_cgupdate(ump, MNT_WAIT); 660 UFS_WAPBL_END(mp); 661 } 662 } 663 if ((mp->mnt_flag & MNT_SOFTDEP) != 0) { 664 printf("%s: `-o softdep' is no longer supported, " 665 "consider `-o log'\n", mp->mnt_stat.f_mntfromname); 666 mp->mnt_flag &= ~MNT_SOFTDEP; 667 } 668 669 return (error); 670 671 fail: 672 vrele(devvp); 673 return (error); 674 } 675 676 /* 677 * Reload all incore data for a filesystem (used after running fsck on 678 * the root filesystem and finding things to fix). The filesystem must 679 * be mounted read-only. 680 * 681 * Things to do to update the mount: 682 * 1) invalidate all cached meta-data. 683 * 2) re-read superblock from disk. 684 * 3) re-read summary information from disk. 685 * 4) invalidate all inactive vnodes. 686 * 5) invalidate all cached file data. 687 * 6) re-read inode data for all active vnodes. 688 */ 689 int 690 ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l) 691 { 692 struct vnode *vp, *devvp; 693 struct inode *ip; 694 void *space; 695 struct buf *bp; 696 struct fs *fs, *newfs; 697 int i, bsize, blks, error; 698 int32_t *lp, fs_sbsize; 699 struct ufsmount *ump; 700 daddr_t sblockloc; 701 struct vnode_iterator *marker; 702 703 if ((mp->mnt_flag & MNT_RDONLY) == 0) 704 return (EINVAL); 705 706 ump = VFSTOUFS(mp); 707 708 /* 709 * Step 1: invalidate all cached meta-data. 710 */ 711 devvp = ump->um_devvp; 712 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 713 error = vinvalbuf(devvp, 0, cred, l, 0, 0); 714 VOP_UNLOCK(devvp); 715 if (error) 716 panic("ffs_reload: dirty1"); 717 718 /* 719 * Step 2: re-read superblock from disk. XXX: We don't handle 720 * possibility that superblock moved. Which implies that we don't 721 * want its size to change either. 722 */ 723 fs = ump->um_fs; 724 fs_sbsize = fs->fs_sbsize; 725 error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs_sbsize, 726 0, &bp); 727 if (error) 728 return (error); 729 newfs = kmem_alloc(fs_sbsize, KM_SLEEP); 730 memcpy(newfs, bp->b_data, fs_sbsize); 731 732 #ifdef FFS_EI 733 if (ump->um_flags & UFS_NEEDSWAP) { 734 ffs_sb_swap((struct fs *)bp->b_data, newfs); 735 newfs->fs_flags |= FS_SWAPPED; 736 } else 737 #endif 738 newfs->fs_flags &= ~FS_SWAPPED; 739 740 brelse(bp, 0); 741 742 if ((newfs->fs_magic != FS_UFS1_MAGIC) && 743 (newfs->fs_magic != FS_UFS2_MAGIC)) { 744 kmem_free(newfs, fs_sbsize); 745 return (EIO); /* XXX needs translation */ 746 } 747 if (!ffs_superblock_validate(newfs)) { 748 kmem_free(newfs, fs_sbsize); 749 return (EINVAL); 750 } 751 752 /* 753 * The current implementation doesn't handle the possibility that 754 * these values may have changed. 755 */ 756 if ((newfs->fs_sbsize != fs_sbsize) || 757 (newfs->fs_cssize != fs->fs_cssize) || 758 (newfs->fs_contigsumsize != fs->fs_contigsumsize) || 759 (newfs->fs_ncg != fs->fs_ncg)) { 760 kmem_free(newfs, fs_sbsize); 761 return (EINVAL); 762 } 763 764 /* Store off old fs_sblockloc for fs_oldfscompat_read. */ 765 sblockloc = fs->fs_sblockloc; 766 /* 767 * Copy pointer fields back into superblock before copying in XXX 768 * new superblock. These should really be in the ufsmount. XXX 769 * Note that important parameters (eg fs_ncg) are unchanged. 770 */ 771 newfs->fs_csp = fs->fs_csp; 772 newfs->fs_maxcluster = fs->fs_maxcluster; 773 newfs->fs_contigdirs = fs->fs_contigdirs; 774 newfs->fs_ronly = fs->fs_ronly; 775 newfs->fs_active = fs->fs_active; 776 memcpy(fs, newfs, (u_int)fs_sbsize); 777 kmem_free(newfs, fs_sbsize); 778 779 /* 780 * Recheck for Apple UFS filesystem. 781 */ 782 ump->um_flags &= ~UFS_ISAPPLEUFS; 783 if (ffs_is_appleufs(devvp, fs)) { 784 #ifdef APPLE_UFS 785 ump->um_flags |= UFS_ISAPPLEUFS; 786 #else 787 DPRINTF("AppleUFS not supported"); 788 return (EIO); /* XXX: really? */ 789 #endif 790 } 791 792 if (UFS_MPISAPPLEUFS(ump)) { 793 /* see comment about NeXT below */ 794 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN; 795 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ; 796 mp->mnt_iflag |= IMNT_DTYPE; 797 } else { 798 ump->um_maxsymlinklen = fs->fs_maxsymlinklen; 799 ump->um_dirblksiz = UFS_DIRBLKSIZ; 800 if (ump->um_maxsymlinklen > 0) 801 mp->mnt_iflag |= IMNT_DTYPE; 802 else 803 mp->mnt_iflag &= ~IMNT_DTYPE; 804 } 805 ffs_oldfscompat_read(fs, ump, sblockloc); 806 807 mutex_enter(&ump->um_lock); 808 ump->um_maxfilesize = fs->fs_maxfilesize; 809 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) { 810 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n", 811 mp->mnt_stat.f_mntonname, fs->fs_flags, 812 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 813 if ((mp->mnt_flag & MNT_FORCE) == 0) { 814 mutex_exit(&ump->um_lock); 815 return (EINVAL); 816 } 817 } 818 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 819 fs->fs_pendingblocks = 0; 820 fs->fs_pendinginodes = 0; 821 } 822 mutex_exit(&ump->um_lock); 823 824 ffs_statvfs(mp, &mp->mnt_stat); 825 /* 826 * Step 3: re-read summary information from disk. 827 */ 828 blks = howmany(fs->fs_cssize, fs->fs_fsize); 829 space = fs->fs_csp; 830 for (i = 0; i < blks; i += fs->fs_frag) { 831 bsize = fs->fs_bsize; 832 if (i + fs->fs_frag > blks) 833 bsize = (blks - i) * fs->fs_fsize; 834 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize, 835 0, &bp); 836 if (error) { 837 return (error); 838 } 839 #ifdef FFS_EI 840 if (UFS_FSNEEDSWAP(fs)) 841 ffs_csum_swap((struct csum *)bp->b_data, 842 (struct csum *)space, bsize); 843 else 844 #endif 845 memcpy(space, bp->b_data, (size_t)bsize); 846 space = (char *)space + bsize; 847 brelse(bp, 0); 848 } 849 /* 850 * We no longer know anything about clusters per cylinder group. 851 */ 852 if (fs->fs_contigsumsize > 0) { 853 lp = fs->fs_maxcluster; 854 for (i = 0; i < fs->fs_ncg; i++) 855 *lp++ = fs->fs_contigsumsize; 856 } 857 858 vfs_vnode_iterator_init(mp, &marker); 859 while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) { 860 /* 861 * Step 4: invalidate all inactive vnodes. 862 */ 863 if (vrecycle(vp)) 864 continue; 865 /* 866 * Step 5: invalidate all cached file data. 867 */ 868 if (vn_lock(vp, LK_EXCLUSIVE)) { 869 vrele(vp); 870 continue; 871 } 872 if (vinvalbuf(vp, 0, cred, l, 0, 0)) 873 panic("ffs_reload: dirty2"); 874 /* 875 * Step 6: re-read inode data for all active vnodes. 876 */ 877 ip = VTOI(vp); 878 error = bread(devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)), 879 (int)fs->fs_bsize, 0, &bp); 880 if (error) { 881 vput(vp); 882 break; 883 } 884 ffs_load_inode(bp, ip, fs, ip->i_number); 885 brelse(bp, 0); 886 vput(vp); 887 } 888 vfs_vnode_iterator_destroy(marker); 889 return (error); 890 } 891 892 /* 893 * Possible superblock locations ordered from most to least likely. 894 */ 895 static const int sblock_try[] = SBLOCKSEARCH; 896 897 898 static int 899 ffs_superblock_validate(struct fs *fs) 900 { 901 int32_t i, fs_bshift = 0, fs_fshift = 0, fs_fragshift = 0, fs_frag; 902 int32_t fs_inopb, fs_cgsize; 903 904 /* Check the superblock size */ 905 if (fs->fs_sbsize > SBLOCKSIZE || fs->fs_sbsize < sizeof(struct fs)) 906 return 0; 907 908 /* Check the file system blocksize */ 909 if (fs->fs_bsize > MAXBSIZE || fs->fs_bsize < MINBSIZE) 910 return 0; 911 if (!powerof2(fs->fs_bsize)) 912 return 0; 913 914 /* Check the size of frag blocks */ 915 if (!powerof2(fs->fs_fsize)) 916 return 0; 917 if (fs->fs_fsize == 0) 918 return 0; 919 920 /* 921 * XXX: these values are just zero-checked to prevent obvious 922 * bugs. We need more strict checks. 923 */ 924 if (fs->fs_size == 0) 925 return 0; 926 if (fs->fs_cssize == 0) 927 return 0; 928 if (fs->fs_ipg == 0) 929 return 0; 930 if (fs->fs_fpg == 0) 931 return 0; 932 if (fs->fs_ncg == 0) 933 return 0; 934 if (fs->fs_maxbpg == 0) 935 return 0; 936 937 /* Check the number of inodes per block */ 938 if (fs->fs_magic == FS_UFS1_MAGIC) 939 fs_inopb = fs->fs_bsize / sizeof(struct ufs1_dinode); 940 else /* fs->fs_magic == FS_UFS2_MAGIC */ 941 fs_inopb = fs->fs_bsize / sizeof(struct ufs2_dinode); 942 if (fs->fs_inopb != fs_inopb) 943 return 0; 944 945 /* Block size cannot be smaller than fragment size */ 946 if (fs->fs_bsize < fs->fs_fsize) 947 return 0; 948 949 /* Compute fs_bshift and ensure it is consistent */ 950 for (i = fs->fs_bsize; i > 1; i >>= 1) 951 fs_bshift++; 952 if (fs->fs_bshift != fs_bshift) 953 return 0; 954 955 /* Compute fs_fshift and ensure it is consistent */ 956 for (i = fs->fs_fsize; i > 1; i >>= 1) 957 fs_fshift++; 958 if (fs->fs_fshift != fs_fshift) 959 return 0; 960 961 /* Compute fs_fragshift and ensure it is consistent */ 962 for (i = fs->fs_frag; i > 1; i >>= 1) 963 fs_fragshift++; 964 if (fs->fs_fragshift != fs_fragshift) 965 return 0; 966 967 /* Check the masks */ 968 if (fs->fs_bmask != ~(fs->fs_bsize - 1)) 969 return 0; 970 if (fs->fs_fmask != ~(fs->fs_fsize - 1)) 971 return 0; 972 973 /* 974 * Now that the shifts and masks are sanitized, we can use the ffs_ API. 975 */ 976 977 /* Check the number of frag blocks */ 978 if ((fs_frag = ffs_numfrags(fs, fs->fs_bsize)) > MAXFRAG) 979 return 0; 980 if (fs->fs_frag != fs_frag) 981 return 0; 982 983 /* Check the size of cylinder groups */ 984 fs_cgsize = ffs_fragroundup(fs, CGSIZE(fs)); 985 if (fs->fs_cgsize != fs_cgsize) { 986 if (fs->fs_cgsize+1 == CGSIZE(fs)) { 987 printf("CGSIZE(fs) miscalculated by one - this file " 988 "system may have been created by\n" 989 " an old (buggy) userland, see\n" 990 " http://www.netbsd.org/" 991 "docs/ffsv1badsuperblock.html\n"); 992 } else { 993 printf("ERROR: cylinder group size mismatch: " 994 "fs_cgsize = 0x%x, " 995 "fs->fs_cgsize = 0x%x, CGSIZE(fs) = 0x%lx\n", 996 fs_cgsize, fs->fs_cgsize, CGSIZE(fs)); 997 return 0; 998 } 999 } 1000 1001 return 1; 1002 } 1003 1004 static int 1005 ffs_is_appleufs(struct vnode *devvp, struct fs *fs) 1006 { 1007 struct dkwedge_info dkw; 1008 int ret = 0; 1009 1010 /* 1011 * First check to see if this is tagged as an Apple UFS filesystem 1012 * in the disklabel. 1013 */ 1014 if (getdiskinfo(devvp, &dkw) == 0 && 1015 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0) 1016 ret = 1; 1017 #ifdef APPLE_UFS 1018 else { 1019 struct appleufslabel *applefs; 1020 struct buf *bp; 1021 daddr_t blkno = APPLEUFS_LABEL_OFFSET / DEV_BSIZE; 1022 int error; 1023 1024 /* 1025 * Manually look for an Apple UFS label, and if a valid one 1026 * is found, then treat it like an Apple UFS filesystem anyway. 1027 */ 1028 error = bread(devvp, blkno, APPLEUFS_LABEL_SIZE, 0, &bp); 1029 if (error) { 1030 DPRINTF("bread@0x%jx returned %d", (intmax_t)blkno, error); 1031 return 0; 1032 } 1033 applefs = (struct appleufslabel *)bp->b_data; 1034 error = ffs_appleufs_validate(fs->fs_fsmnt, applefs, NULL); 1035 if (error == 0) 1036 ret = 1; 1037 brelse(bp, 0); 1038 } 1039 #endif 1040 1041 return ret; 1042 } 1043 1044 /* 1045 * Common code for mount and mountroot 1046 */ 1047 int 1048 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l) 1049 { 1050 struct ufsmount *ump = NULL; 1051 struct buf *bp = NULL; 1052 struct fs *fs = NULL; 1053 dev_t dev; 1054 void *space; 1055 daddr_t sblockloc = 0; 1056 int blks, fstype = 0; 1057 int error, i, bsize, ronly, bset = 0; 1058 #ifdef FFS_EI 1059 int needswap = 0; /* keep gcc happy */ 1060 #endif 1061 int32_t *lp; 1062 kauth_cred_t cred; 1063 u_int32_t allocsbsize, fs_sbsize = 0; 1064 1065 dev = devvp->v_rdev; 1066 cred = l ? l->l_cred : NOCRED; 1067 1068 /* Flush out any old buffers remaining from a previous use. */ 1069 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 1070 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0); 1071 VOP_UNLOCK(devvp); 1072 if (error) { 1073 DPRINTF("vinvalbuf returned %d", error); 1074 return error; 1075 } 1076 1077 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 1078 1079 error = fstrans_mount(mp); 1080 if (error) { 1081 DPRINTF("fstrans_mount returned %d", error); 1082 return error; 1083 } 1084 1085 ump = kmem_zalloc(sizeof(*ump), KM_SLEEP); 1086 mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE); 1087 error = ffs_snapshot_init(ump); 1088 if (error) { 1089 DPRINTF("ffs_snapshot_init returned %d", error); 1090 goto out; 1091 } 1092 ump->um_ops = &ffs_ufsops; 1093 1094 #ifdef WAPBL 1095 sbagain: 1096 #endif 1097 /* 1098 * Try reading the superblock in each of its possible locations. 1099 */ 1100 for (i = 0; ; i++) { 1101 daddr_t fs_sblockloc; 1102 1103 if (bp != NULL) { 1104 brelse(bp, BC_NOCACHE); 1105 bp = NULL; 1106 } 1107 if (sblock_try[i] == -1) { 1108 DPRINTF("no superblock found"); 1109 error = EINVAL; 1110 fs = NULL; 1111 goto out; 1112 } 1113 1114 error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, 1115 0, &bp); 1116 if (error) { 1117 DPRINTF("bread@0x%x returned %d", 1118 sblock_try[i] / DEV_BSIZE, error); 1119 fs = NULL; 1120 goto out; 1121 } 1122 fs = (struct fs *)bp->b_data; 1123 1124 sblockloc = sblock_try[i]; 1125 DPRINTF("fs_magic 0x%x", fs->fs_magic); 1126 1127 /* 1128 * Swap: here, we swap fs->fs_sbsize in order to get the correct 1129 * size to read the superblock. Once read, we swap the whole 1130 * superblock structure. 1131 */ 1132 if (fs->fs_magic == FS_UFS1_MAGIC) { 1133 fs_sbsize = fs->fs_sbsize; 1134 fstype = UFS1; 1135 #ifdef FFS_EI 1136 needswap = 0; 1137 } else if (fs->fs_magic == FS_UFS1_MAGIC_SWAPPED) { 1138 fs_sbsize = bswap32(fs->fs_sbsize); 1139 fstype = UFS1; 1140 needswap = 1; 1141 #endif 1142 } else if (fs->fs_magic == FS_UFS2_MAGIC) { 1143 fs_sbsize = fs->fs_sbsize; 1144 fstype = UFS2; 1145 #ifdef FFS_EI 1146 needswap = 0; 1147 } else if (fs->fs_magic == FS_UFS2_MAGIC_SWAPPED) { 1148 fs_sbsize = bswap32(fs->fs_sbsize); 1149 fstype = UFS2; 1150 needswap = 1; 1151 #endif 1152 } else 1153 continue; 1154 1155 /* fs->fs_sblockloc isn't defined for old filesystems */ 1156 if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) { 1157 if (sblockloc == SBLOCK_UFS2) 1158 /* 1159 * This is likely to be the first alternate 1160 * in a filesystem with 64k blocks. 1161 * Don't use it. 1162 */ 1163 continue; 1164 fs_sblockloc = sblockloc; 1165 } else { 1166 fs_sblockloc = fs->fs_sblockloc; 1167 #ifdef FFS_EI 1168 if (needswap) 1169 fs_sblockloc = bswap64(fs_sblockloc); 1170 #endif 1171 } 1172 1173 /* Check we haven't found an alternate superblock */ 1174 if (fs_sblockloc != sblockloc) 1175 continue; 1176 1177 /* Check the superblock size */ 1178 if (fs_sbsize > SBLOCKSIZE || fs_sbsize < sizeof(struct fs)) 1179 continue; 1180 fs = kmem_alloc((u_long)fs_sbsize, KM_SLEEP); 1181 memcpy(fs, bp->b_data, fs_sbsize); 1182 1183 /* Swap the whole superblock structure, if necessary. */ 1184 #ifdef FFS_EI 1185 if (needswap) { 1186 ffs_sb_swap((struct fs*)bp->b_data, fs); 1187 fs->fs_flags |= FS_SWAPPED; 1188 } else 1189 #endif 1190 fs->fs_flags &= ~FS_SWAPPED; 1191 1192 /* 1193 * Now that everything is swapped, the superblock is ready to 1194 * be sanitized. 1195 */ 1196 if (!ffs_superblock_validate(fs)) { 1197 kmem_free(fs, fs_sbsize); 1198 continue; 1199 } 1200 1201 /* Ok seems to be a good superblock */ 1202 break; 1203 } 1204 1205 ump->um_fs = fs; 1206 1207 #ifdef WAPBL 1208 if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) { 1209 error = ffs_wapbl_replay_start(mp, fs, devvp); 1210 if (error && (mp->mnt_flag & MNT_FORCE) == 0) { 1211 DPRINTF("ffs_wapbl_replay_start returned %d", error); 1212 goto out; 1213 } 1214 if (!error) { 1215 if (!ronly) { 1216 /* XXX fsmnt may be stale. */ 1217 printf("%s: replaying log to disk\n", 1218 fs->fs_fsmnt); 1219 error = wapbl_replay_write(mp->mnt_wapbl_replay, 1220 devvp); 1221 if (error) { 1222 DPRINTF("wapbl_replay_write returned %d", 1223 error); 1224 goto out; 1225 } 1226 wapbl_replay_stop(mp->mnt_wapbl_replay); 1227 fs->fs_clean = FS_WASCLEAN; 1228 } else { 1229 /* XXX fsmnt may be stale */ 1230 printf("%s: replaying log to memory\n", 1231 fs->fs_fsmnt); 1232 } 1233 1234 /* Force a re-read of the superblock */ 1235 brelse(bp, BC_INVAL); 1236 bp = NULL; 1237 kmem_free(fs, fs_sbsize); 1238 fs = NULL; 1239 goto sbagain; 1240 } 1241 } 1242 #else /* !WAPBL */ 1243 if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) { 1244 error = EPERM; 1245 DPRINTF("no force %d", error); 1246 goto out; 1247 } 1248 #endif /* !WAPBL */ 1249 1250 ffs_oldfscompat_read(fs, ump, sblockloc); 1251 ump->um_maxfilesize = fs->fs_maxfilesize; 1252 1253 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) { 1254 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n", 1255 mp->mnt_stat.f_mntonname, fs->fs_flags, 1256 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 1257 if ((mp->mnt_flag & MNT_FORCE) == 0) { 1258 error = EINVAL; 1259 DPRINTF("no force %d", error); 1260 goto out; 1261 } 1262 } 1263 1264 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 1265 fs->fs_pendingblocks = 0; 1266 fs->fs_pendinginodes = 0; 1267 } 1268 1269 ump->um_fstype = fstype; 1270 if (fs->fs_sbsize < SBLOCKSIZE) 1271 brelse(bp, BC_INVAL); 1272 else 1273 brelse(bp, 0); 1274 bp = NULL; 1275 1276 if (ffs_is_appleufs(devvp, fs)) { 1277 #ifdef APPLE_UFS 1278 ump->um_flags |= UFS_ISAPPLEUFS; 1279 #else 1280 DPRINTF("AppleUFS not supported"); 1281 error = EINVAL; 1282 goto out; 1283 #endif 1284 } 1285 1286 #if 0 1287 /* 1288 * XXX This code changes the behaviour of mounting dirty filesystems, to 1289 * XXX require "mount -f ..." to mount them. This doesn't match what 1290 * XXX mount(8) describes and is disabled for now. 1291 */ 1292 /* 1293 * If the file system is not clean, don't allow it to be mounted 1294 * unless MNT_FORCE is specified. (Note: MNT_FORCE is always set 1295 * for the root file system.) 1296 */ 1297 if (fs->fs_flags & FS_DOWAPBL) { 1298 /* 1299 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL 1300 * bit is set, although there's a window in unmount where it 1301 * could be FS_ISCLEAN 1302 */ 1303 if ((mp->mnt_flag & MNT_FORCE) == 0 && 1304 (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) { 1305 error = EPERM; 1306 goto out; 1307 } 1308 } else 1309 if ((fs->fs_clean & FS_ISCLEAN) == 0 && 1310 (mp->mnt_flag & MNT_FORCE) == 0) { 1311 error = EPERM; 1312 goto out; 1313 } 1314 #endif 1315 1316 /* 1317 * Verify that we can access the last block in the fs 1318 * if we're mounting read/write. 1319 */ 1320 if (!ronly) { 1321 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_size - 1), 1322 fs->fs_fsize, 0, &bp); 1323 if (error) { 1324 DPRINTF("bread@0x%jx returned %d", 1325 (intmax_t)FFS_FSBTODB(fs, fs->fs_size - 1), 1326 error); 1327 bset = BC_INVAL; 1328 goto out; 1329 } 1330 if (bp->b_bcount != fs->fs_fsize) { 1331 DPRINTF("bcount %x != fsize %x", bp->b_bcount, 1332 fs->fs_fsize); 1333 error = EINVAL; 1334 } 1335 brelse(bp, BC_INVAL); 1336 bp = NULL; 1337 } 1338 1339 fs->fs_ronly = ronly; 1340 /* Don't bump fs_clean if we're replaying journal */ 1341 if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN))) { 1342 if (ronly == 0) { 1343 fs->fs_clean <<= 1; 1344 fs->fs_fmod = 1; 1345 } 1346 } 1347 1348 bsize = fs->fs_cssize; 1349 blks = howmany(bsize, fs->fs_fsize); 1350 if (fs->fs_contigsumsize > 0) 1351 bsize += fs->fs_ncg * sizeof(int32_t); 1352 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1353 allocsbsize = bsize; 1354 space = kmem_alloc((u_long)allocsbsize, KM_SLEEP); 1355 fs->fs_csp = space; 1356 1357 for (i = 0; i < blks; i += fs->fs_frag) { 1358 bsize = fs->fs_bsize; 1359 if (i + fs->fs_frag > blks) 1360 bsize = (blks - i) * fs->fs_fsize; 1361 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize, 1362 0, &bp); 1363 if (error) { 1364 DPRINTF("bread@0x%jx %d", 1365 (intmax_t)FFS_FSBTODB(fs, fs->fs_csaddr + i), 1366 error); 1367 goto out1; 1368 } 1369 #ifdef FFS_EI 1370 if (needswap) 1371 ffs_csum_swap((struct csum *)bp->b_data, 1372 (struct csum *)space, bsize); 1373 else 1374 #endif 1375 memcpy(space, bp->b_data, (u_int)bsize); 1376 1377 space = (char *)space + bsize; 1378 brelse(bp, 0); 1379 bp = NULL; 1380 } 1381 if (fs->fs_contigsumsize > 0) { 1382 fs->fs_maxcluster = lp = space; 1383 for (i = 0; i < fs->fs_ncg; i++) 1384 *lp++ = fs->fs_contigsumsize; 1385 space = lp; 1386 } 1387 bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1388 fs->fs_contigdirs = space; 1389 space = (char *)space + bsize; 1390 memset(fs->fs_contigdirs, 0, bsize); 1391 1392 /* Compatibility for old filesystems - XXX */ 1393 if (fs->fs_avgfilesize <= 0) 1394 fs->fs_avgfilesize = AVFILESIZ; 1395 if (fs->fs_avgfpdir <= 0) 1396 fs->fs_avgfpdir = AFPDIR; 1397 fs->fs_active = NULL; 1398 1399 mp->mnt_data = ump; 1400 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev; 1401 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS); 1402 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0]; 1403 mp->mnt_stat.f_namemax = FFS_MAXNAMLEN; 1404 if (UFS_MPISAPPLEUFS(ump)) { 1405 /* NeXT used to keep short symlinks in the inode even 1406 * when using FS_42INODEFMT. In that case fs->fs_maxsymlinklen 1407 * is probably -1, but we still need to be able to identify 1408 * short symlinks. 1409 */ 1410 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN; 1411 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ; 1412 mp->mnt_iflag |= IMNT_DTYPE; 1413 } else { 1414 ump->um_maxsymlinklen = fs->fs_maxsymlinklen; 1415 ump->um_dirblksiz = UFS_DIRBLKSIZ; 1416 if (ump->um_maxsymlinklen > 0) 1417 mp->mnt_iflag |= IMNT_DTYPE; 1418 else 1419 mp->mnt_iflag &= ~IMNT_DTYPE; 1420 } 1421 mp->mnt_fs_bshift = fs->fs_bshift; 1422 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */ 1423 mp->mnt_flag |= MNT_LOCAL; 1424 mp->mnt_iflag |= IMNT_MPSAFE; 1425 #ifdef FFS_EI 1426 if (needswap) 1427 ump->um_flags |= UFS_NEEDSWAP; 1428 #endif 1429 ump->um_mountp = mp; 1430 ump->um_dev = dev; 1431 ump->um_devvp = devvp; 1432 ump->um_nindir = fs->fs_nindir; 1433 ump->um_lognindir = ffs(fs->fs_nindir) - 1; 1434 ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT; 1435 ump->um_seqinc = fs->fs_frag; 1436 for (i = 0; i < MAXQUOTAS; i++) 1437 ump->um_quotas[i] = NULLVP; 1438 spec_node_setmountedfs(devvp, mp); 1439 if (ronly == 0 && fs->fs_snapinum[0] != 0) 1440 ffs_snapshot_mount(mp); 1441 #ifdef WAPBL 1442 if (!ronly) { 1443 KDASSERT(fs->fs_ronly == 0); 1444 /* 1445 * ffs_wapbl_start() needs mp->mnt_stat initialised if it 1446 * needs to create a new log file in-filesystem. 1447 */ 1448 error = ffs_statvfs(mp, &mp->mnt_stat); 1449 if (error) { 1450 DPRINTF("ffs_statvfs returned %d", error); 1451 goto out1; 1452 } 1453 1454 error = ffs_wapbl_start(mp); 1455 if (error) { 1456 DPRINTF("ffs_wapbl_start returned %d", error); 1457 goto out1; 1458 } 1459 } 1460 #endif /* WAPBL */ 1461 if (ronly == 0) { 1462 #ifdef QUOTA2 1463 error = ffs_quota2_mount(mp); 1464 if (error) { 1465 DPRINTF("ffs_quota2_mount returned %d", error); 1466 goto out1; 1467 } 1468 #else 1469 if (fs->fs_flags & FS_DOQUOTA2) { 1470 ump->um_flags |= UFS_QUOTA2; 1471 uprintf("%s: options QUOTA2 not enabled%s\n", 1472 mp->mnt_stat.f_mntonname, 1473 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 1474 if ((mp->mnt_flag & MNT_FORCE) == 0) { 1475 error = EINVAL; 1476 DPRINTF("quota disabled %d", error); 1477 goto out1; 1478 } 1479 } 1480 #endif 1481 } 1482 1483 if (mp->mnt_flag & MNT_DISCARD) 1484 ump->um_discarddata = ffs_discard_init(devvp, fs); 1485 1486 return (0); 1487 out1: 1488 kmem_free(fs->fs_csp, allocsbsize); 1489 out: 1490 #ifdef WAPBL 1491 if (mp->mnt_wapbl_replay) { 1492 wapbl_replay_stop(mp->mnt_wapbl_replay); 1493 wapbl_replay_free(mp->mnt_wapbl_replay); 1494 mp->mnt_wapbl_replay = 0; 1495 } 1496 #endif 1497 1498 fstrans_unmount(mp); 1499 if (fs) 1500 kmem_free(fs, fs->fs_sbsize); 1501 spec_node_setmountedfs(devvp, NULL); 1502 if (bp) 1503 brelse(bp, bset); 1504 if (ump) { 1505 if (ump->um_oldfscompat) 1506 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t)); 1507 mutex_destroy(&ump->um_lock); 1508 kmem_free(ump, sizeof(*ump)); 1509 mp->mnt_data = NULL; 1510 } 1511 return (error); 1512 } 1513 1514 /* 1515 * Sanity checks for loading old filesystem superblocks. 1516 * See ffs_oldfscompat_write below for unwound actions. 1517 * 1518 * XXX - Parts get retired eventually. 1519 * Unfortunately new bits get added. 1520 */ 1521 static void 1522 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc) 1523 { 1524 off_t maxfilesize; 1525 int32_t *extrasave; 1526 1527 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1528 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1529 return; 1530 1531 if (!ump->um_oldfscompat) 1532 ump->um_oldfscompat = kmem_alloc(512 + 3*sizeof(int32_t), 1533 KM_SLEEP); 1534 1535 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512); 1536 extrasave = ump->um_oldfscompat; 1537 extrasave += 512/sizeof(int32_t); 1538 extrasave[0] = fs->fs_old_npsect; 1539 extrasave[1] = fs->fs_old_interleave; 1540 extrasave[2] = fs->fs_old_trackskew; 1541 1542 /* These fields will be overwritten by their 1543 * original values in fs_oldfscompat_write, so it is harmless 1544 * to modify them here. 1545 */ 1546 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir; 1547 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree; 1548 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree; 1549 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree; 1550 1551 fs->fs_maxbsize = fs->fs_bsize; 1552 fs->fs_time = fs->fs_old_time; 1553 fs->fs_size = fs->fs_old_size; 1554 fs->fs_dsize = fs->fs_old_dsize; 1555 fs->fs_csaddr = fs->fs_old_csaddr; 1556 fs->fs_sblockloc = sblockloc; 1557 1558 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL); 1559 1560 if (fs->fs_old_postblformat == FS_42POSTBLFMT) { 1561 fs->fs_old_nrpos = 8; 1562 fs->fs_old_npsect = fs->fs_old_nsect; 1563 fs->fs_old_interleave = 1; 1564 fs->fs_old_trackskew = 0; 1565 } 1566 1567 if (fs->fs_old_inodefmt < FS_44INODEFMT) { 1568 fs->fs_maxfilesize = (u_quad_t) 1LL << 39; 1569 fs->fs_qbmask = ~fs->fs_bmask; 1570 fs->fs_qfmask = ~fs->fs_fmask; 1571 } 1572 1573 maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1; 1574 if (fs->fs_maxfilesize > maxfilesize) 1575 fs->fs_maxfilesize = maxfilesize; 1576 1577 /* Compatibility for old filesystems */ 1578 if (fs->fs_avgfilesize <= 0) 1579 fs->fs_avgfilesize = AVFILESIZ; 1580 if (fs->fs_avgfpdir <= 0) 1581 fs->fs_avgfpdir = AFPDIR; 1582 1583 #if 0 1584 if (bigcgs) { 1585 fs->fs_save_cgsize = fs->fs_cgsize; 1586 fs->fs_cgsize = fs->fs_bsize; 1587 } 1588 #endif 1589 } 1590 1591 /* 1592 * Unwinding superblock updates for old filesystems. 1593 * See ffs_oldfscompat_read above for details. 1594 * 1595 * XXX - Parts get retired eventually. 1596 * Unfortunately new bits get added. 1597 */ 1598 static void 1599 ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump) 1600 { 1601 int32_t *extrasave; 1602 1603 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1604 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1605 return; 1606 1607 fs->fs_old_time = fs->fs_time; 1608 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir; 1609 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree; 1610 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree; 1611 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree; 1612 fs->fs_old_flags = fs->fs_flags; 1613 1614 #if 0 1615 if (bigcgs) { 1616 fs->fs_cgsize = fs->fs_save_cgsize; 1617 } 1618 #endif 1619 1620 memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512); 1621 extrasave = ump->um_oldfscompat; 1622 extrasave += 512/sizeof(int32_t); 1623 fs->fs_old_npsect = extrasave[0]; 1624 fs->fs_old_interleave = extrasave[1]; 1625 fs->fs_old_trackskew = extrasave[2]; 1626 1627 } 1628 1629 /* 1630 * unmount vfs operation 1631 */ 1632 int 1633 ffs_unmount(struct mount *mp, int mntflags) 1634 { 1635 struct lwp *l = curlwp; 1636 struct ufsmount *ump = VFSTOUFS(mp); 1637 struct fs *fs = ump->um_fs; 1638 int error, flags; 1639 u_int32_t bsize; 1640 #ifdef WAPBL 1641 extern int doforce; 1642 #endif 1643 1644 if (ump->um_discarddata) { 1645 ffs_discard_finish(ump->um_discarddata, mntflags); 1646 ump->um_discarddata = NULL; 1647 } 1648 1649 flags = 0; 1650 if (mntflags & MNT_FORCE) 1651 flags |= FORCECLOSE; 1652 if ((error = ffs_flushfiles(mp, flags, l)) != 0) 1653 return (error); 1654 error = UFS_WAPBL_BEGIN(mp); 1655 if (error == 0) 1656 if (fs->fs_ronly == 0 && 1657 ffs_cgupdate(ump, MNT_WAIT) == 0 && 1658 fs->fs_clean & FS_WASCLEAN) { 1659 fs->fs_clean = FS_ISCLEAN; 1660 fs->fs_fmod = 0; 1661 (void) ffs_sbupdate(ump, MNT_WAIT); 1662 } 1663 if (error == 0) 1664 UFS_WAPBL_END(mp); 1665 #ifdef WAPBL 1666 KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl)); 1667 if (mp->mnt_wapbl_replay) { 1668 KDASSERT(fs->fs_ronly); 1669 wapbl_replay_stop(mp->mnt_wapbl_replay); 1670 wapbl_replay_free(mp->mnt_wapbl_replay); 1671 mp->mnt_wapbl_replay = 0; 1672 } 1673 error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE)); 1674 if (error) { 1675 return error; 1676 } 1677 #endif /* WAPBL */ 1678 1679 if (ump->um_devvp->v_type != VBAD) 1680 spec_node_setmountedfs(ump->um_devvp, NULL); 1681 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1682 (void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE, 1683 NOCRED); 1684 vput(ump->um_devvp); 1685 1686 bsize = fs->fs_cssize; 1687 if (fs->fs_contigsumsize > 0) 1688 bsize += fs->fs_ncg * sizeof(int32_t); 1689 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1690 kmem_free(fs->fs_csp, bsize); 1691 1692 kmem_free(fs, fs->fs_sbsize); 1693 if (ump->um_oldfscompat != NULL) 1694 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t)); 1695 mutex_destroy(&ump->um_lock); 1696 ffs_snapshot_fini(ump); 1697 kmem_free(ump, sizeof(*ump)); 1698 mp->mnt_data = NULL; 1699 mp->mnt_flag &= ~MNT_LOCAL; 1700 fstrans_unmount(mp); 1701 return (0); 1702 } 1703 1704 /* 1705 * Flush out all the files in a filesystem. 1706 */ 1707 int 1708 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l) 1709 { 1710 extern int doforce; 1711 struct ufsmount *ump; 1712 int error; 1713 1714 if (!doforce) 1715 flags &= ~FORCECLOSE; 1716 ump = VFSTOUFS(mp); 1717 #ifdef QUOTA 1718 if ((error = quota1_umount(mp, flags)) != 0) 1719 return (error); 1720 #endif 1721 #ifdef QUOTA2 1722 if ((error = quota2_umount(mp, flags)) != 0) 1723 return (error); 1724 #endif 1725 #ifdef UFS_EXTATTR 1726 if (ump->um_fstype == UFS1) { 1727 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_STARTED) 1728 ufs_extattr_stop(mp, l); 1729 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_INITIALIZED) 1730 ufs_extattr_uepm_destroy(&ump->um_extattr); 1731 mp->mnt_flag &= ~MNT_EXTATTR; 1732 } 1733 #endif 1734 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0) 1735 return (error); 1736 ffs_snapshot_unmount(mp); 1737 /* 1738 * Flush all the files. 1739 */ 1740 error = vflush(mp, NULLVP, flags); 1741 if (error) 1742 return (error); 1743 /* 1744 * Flush filesystem metadata. 1745 */ 1746 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1747 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0); 1748 VOP_UNLOCK(ump->um_devvp); 1749 if (flags & FORCECLOSE) /* XXXDBJ */ 1750 error = 0; 1751 1752 #ifdef WAPBL 1753 if (error) 1754 return error; 1755 if (mp->mnt_wapbl) { 1756 error = wapbl_flush(mp->mnt_wapbl, 1); 1757 if (flags & FORCECLOSE) 1758 error = 0; 1759 } 1760 #endif 1761 1762 return (error); 1763 } 1764 1765 /* 1766 * Get file system statistics. 1767 */ 1768 int 1769 ffs_statvfs(struct mount *mp, struct statvfs *sbp) 1770 { 1771 struct ufsmount *ump; 1772 struct fs *fs; 1773 1774 ump = VFSTOUFS(mp); 1775 fs = ump->um_fs; 1776 mutex_enter(&ump->um_lock); 1777 sbp->f_bsize = fs->fs_bsize; 1778 sbp->f_frsize = fs->fs_fsize; 1779 sbp->f_iosize = fs->fs_bsize; 1780 sbp->f_blocks = fs->fs_dsize; 1781 sbp->f_bfree = ffs_blkstofrags(fs, fs->fs_cstotal.cs_nbfree) + 1782 fs->fs_cstotal.cs_nffree + FFS_DBTOFSB(fs, fs->fs_pendingblocks); 1783 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t) 1784 fs->fs_minfree) / (u_int64_t) 100; 1785 if (sbp->f_bfree > sbp->f_bresvd) 1786 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd; 1787 else 1788 sbp->f_bavail = 0; 1789 sbp->f_files = fs->fs_ncg * fs->fs_ipg - UFS_ROOTINO; 1790 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes; 1791 sbp->f_favail = sbp->f_ffree; 1792 sbp->f_fresvd = 0; 1793 mutex_exit(&ump->um_lock); 1794 copy_statvfs_info(sbp, mp); 1795 1796 return (0); 1797 } 1798 1799 struct ffs_sync_ctx { 1800 int waitfor; 1801 bool is_suspending; 1802 }; 1803 1804 static bool 1805 ffs_sync_selector(void *cl, struct vnode *vp) 1806 { 1807 struct ffs_sync_ctx *c = cl; 1808 struct inode *ip; 1809 1810 ip = VTOI(vp); 1811 /* 1812 * Skip the vnode/inode if inaccessible. 1813 */ 1814 if (ip == NULL || vp->v_type == VNON) 1815 return false; 1816 1817 /* 1818 * We deliberately update inode times here. This will 1819 * prevent a massive queue of updates accumulating, only 1820 * to be handled by a call to unmount. 1821 * 1822 * XXX It would be better to have the syncer trickle these 1823 * out. Adjustment needed to allow registering vnodes for 1824 * sync when the vnode is clean, but the inode dirty. Or 1825 * have ufs itself trickle out inode updates. 1826 * 1827 * If doing a lazy sync, we don't care about metadata or 1828 * data updates, because they are handled by each vnode's 1829 * synclist entry. In this case we are only interested in 1830 * writing back modified inodes. 1831 */ 1832 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | 1833 IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 && 1834 (c->waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) && 1835 UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) 1836 return false; 1837 1838 if (vp->v_type == VBLK && c->is_suspending) 1839 return false; 1840 1841 return true; 1842 } 1843 1844 /* 1845 * Go through the disk queues to initiate sandbagged IO; 1846 * go through the inodes to write those that have been modified; 1847 * initiate the writing of the super block if it has been modified. 1848 * 1849 * Note: we are always called with the filesystem marked `MPBUSY'. 1850 */ 1851 int 1852 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred) 1853 { 1854 struct vnode *vp; 1855 struct ufsmount *ump = VFSTOUFS(mp); 1856 struct fs *fs; 1857 struct vnode_iterator *marker; 1858 int error, allerror = 0; 1859 bool is_suspending; 1860 struct ffs_sync_ctx ctx; 1861 1862 fs = ump->um_fs; 1863 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */ 1864 printf("fs = %s\n", fs->fs_fsmnt); 1865 panic("update: rofs mod"); 1866 } 1867 1868 fstrans_start(mp, FSTRANS_SHARED); 1869 is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING); 1870 /* 1871 * Write back each (modified) inode. 1872 */ 1873 vfs_vnode_iterator_init(mp, &marker); 1874 1875 ctx.waitfor = waitfor; 1876 ctx.is_suspending = is_suspending; 1877 while ((vp = vfs_vnode_iterator_next(marker, ffs_sync_selector, &ctx))) 1878 { 1879 error = vn_lock(vp, LK_EXCLUSIVE); 1880 if (error) { 1881 vrele(vp); 1882 continue; 1883 } 1884 if (waitfor == MNT_LAZY) { 1885 error = UFS_WAPBL_BEGIN(vp->v_mount); 1886 if (!error) { 1887 error = ffs_update(vp, NULL, NULL, 1888 UPDATE_CLOSE); 1889 UFS_WAPBL_END(vp->v_mount); 1890 } 1891 } else { 1892 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG | 1893 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0); 1894 } 1895 if (error) 1896 allerror = error; 1897 vput(vp); 1898 } 1899 vfs_vnode_iterator_destroy(marker); 1900 1901 /* 1902 * Force stale file system control information to be flushed. 1903 */ 1904 if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 || 1905 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) { 1906 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1907 if ((error = VOP_FSYNC(ump->um_devvp, cred, 1908 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG, 1909 0, 0)) != 0) 1910 allerror = error; 1911 VOP_UNLOCK(ump->um_devvp); 1912 } 1913 #if defined(QUOTA) || defined(QUOTA2) 1914 qsync(mp); 1915 #endif 1916 /* 1917 * Write back modified superblock. 1918 */ 1919 if (fs->fs_fmod != 0) { 1920 fs->fs_fmod = 0; 1921 fs->fs_time = time_second; 1922 error = UFS_WAPBL_BEGIN(mp); 1923 if (error) 1924 allerror = error; 1925 else { 1926 if ((error = ffs_cgupdate(ump, waitfor))) 1927 allerror = error; 1928 UFS_WAPBL_END(mp); 1929 } 1930 } 1931 1932 #ifdef WAPBL 1933 if (mp->mnt_wapbl) { 1934 error = wapbl_flush(mp->mnt_wapbl, 0); 1935 if (error) 1936 allerror = error; 1937 } 1938 #endif 1939 1940 fstrans_done(mp); 1941 return (allerror); 1942 } 1943 1944 /* 1945 * Load inode from disk and initialize vnode. 1946 */ 1947 static int 1948 ffs_init_vnode(struct ufsmount *ump, struct vnode *vp, ino_t ino) 1949 { 1950 struct fs *fs; 1951 struct inode *ip; 1952 struct buf *bp; 1953 int error; 1954 1955 fs = ump->um_fs; 1956 1957 /* Read in the disk contents for the inode. */ 1958 error = bread(ump->um_devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ino)), 1959 (int)fs->fs_bsize, 0, &bp); 1960 if (error) 1961 return error; 1962 1963 /* Allocate and initialize inode. */ 1964 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK); 1965 memset(ip, 0, sizeof(struct inode)); 1966 ip->i_ump = ump; 1967 ip->i_fs = fs; 1968 ip->i_dev = ump->um_dev; 1969 ip->i_number = ino; 1970 if (ump->um_fstype == UFS1) 1971 ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache, 1972 PR_WAITOK); 1973 else 1974 ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache, 1975 PR_WAITOK); 1976 ffs_load_inode(bp, ip, fs, ino); 1977 brelse(bp, 0); 1978 ip->i_vnode = vp; 1979 #if defined(QUOTA) || defined(QUOTA2) 1980 ufsquota_init(ip); 1981 #endif 1982 1983 /* Initialise vnode with this inode. */ 1984 vp->v_tag = VT_UFS; 1985 vp->v_op = ffs_vnodeop_p; 1986 vp->v_vflag |= VV_LOCKSWORK; 1987 vp->v_data = ip; 1988 1989 /* Initialize genfs node. */ 1990 genfs_node_init(vp, &ffs_genfsops); 1991 1992 return 0; 1993 } 1994 1995 /* 1996 * Undo ffs_init_vnode(). 1997 */ 1998 static void 1999 ffs_deinit_vnode(struct ufsmount *ump, struct vnode *vp) 2000 { 2001 struct inode *ip = VTOI(vp); 2002 2003 if (ump->um_fstype == UFS1) 2004 pool_cache_put(ffs_dinode1_cache, ip->i_din.ffs1_din); 2005 else 2006 pool_cache_put(ffs_dinode2_cache, ip->i_din.ffs2_din); 2007 pool_cache_put(ffs_inode_cache, ip); 2008 2009 genfs_node_destroy(vp); 2010 vp->v_data = NULL; 2011 } 2012 2013 /* 2014 * Read an inode from disk and initialize this vnode / inode pair. 2015 * Caller assures no other thread will try to load this inode. 2016 */ 2017 int 2018 ffs_loadvnode(struct mount *mp, struct vnode *vp, 2019 const void *key, size_t key_len, const void **new_key) 2020 { 2021 ino_t ino; 2022 struct fs *fs; 2023 struct inode *ip; 2024 struct ufsmount *ump; 2025 int error; 2026 2027 KASSERT(key_len == sizeof(ino)); 2028 memcpy(&ino, key, key_len); 2029 ump = VFSTOUFS(mp); 2030 fs = ump->um_fs; 2031 2032 error = ffs_init_vnode(ump, vp, ino); 2033 if (error) 2034 return error; 2035 2036 ip = VTOI(vp); 2037 if (ip->i_mode == 0) { 2038 ffs_deinit_vnode(ump, vp); 2039 2040 return ENOENT; 2041 } 2042 2043 /* Initialize the vnode from the inode. */ 2044 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp); 2045 2046 /* Finish inode initialization. */ 2047 ip->i_devvp = ump->um_devvp; 2048 vref(ip->i_devvp); 2049 2050 /* 2051 * Ensure that uid and gid are correct. This is a temporary 2052 * fix until fsck has been changed to do the update. 2053 */ 2054 2055 if (fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */ 2056 ip->i_uid = ip->i_ffs1_ouid; /* XXX */ 2057 ip->i_gid = ip->i_ffs1_ogid; /* XXX */ 2058 } /* XXX */ 2059 uvm_vnp_setsize(vp, ip->i_size); 2060 *new_key = &ip->i_number; 2061 return 0; 2062 } 2063 2064 /* 2065 * Create a new inode on disk and initialize this vnode / inode pair. 2066 */ 2067 int 2068 ffs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp, 2069 struct vattr *vap, kauth_cred_t cred, 2070 size_t *key_len, const void **new_key) 2071 { 2072 ino_t ino; 2073 struct fs *fs; 2074 struct inode *ip; 2075 struct timespec ts; 2076 struct ufsmount *ump; 2077 int error, mode; 2078 2079 KASSERT(dvp->v_mount == mp); 2080 KASSERT(vap->va_type != VNON); 2081 2082 *key_len = sizeof(ino); 2083 ump = VFSTOUFS(mp); 2084 fs = ump->um_fs; 2085 mode = MAKEIMODE(vap->va_type, vap->va_mode); 2086 2087 /* Allocate fresh inode. */ 2088 error = ffs_valloc(dvp, mode, cred, &ino); 2089 if (error) 2090 return error; 2091 2092 /* Attach inode to vnode. */ 2093 error = ffs_init_vnode(ump, vp, ino); 2094 if (error) { 2095 if (UFS_WAPBL_BEGIN(mp) == 0) { 2096 ffs_vfree(dvp, ino, mode); 2097 UFS_WAPBL_END(mp); 2098 } 2099 return error; 2100 } 2101 2102 ip = VTOI(vp); 2103 if (ip->i_mode || DIP(ip, size) || DIP(ip, blocks)) { 2104 printf("free ino %" PRId64 " on %s:\n", ino, fs->fs_fsmnt); 2105 printf("dmode %x mode %x dgen %x gen %x\n", 2106 DIP(ip, mode), ip->i_mode, 2107 DIP(ip, gen), ip->i_gen); 2108 printf("size %" PRIx64 " blocks %" PRIx64 "\n", 2109 DIP(ip, size), DIP(ip, blocks)); 2110 panic("ffs_init_vnode: dup alloc"); 2111 } 2112 2113 /* Set uid / gid. */ 2114 if (cred == NOCRED || cred == FSCRED) { 2115 ip->i_gid = 0; 2116 ip->i_uid = 0; 2117 } else { 2118 ip->i_gid = VTOI(dvp)->i_gid; 2119 ip->i_uid = kauth_cred_geteuid(cred); 2120 } 2121 DIP_ASSIGN(ip, gid, ip->i_gid); 2122 DIP_ASSIGN(ip, uid, ip->i_uid); 2123 2124 #if defined(QUOTA) || defined(QUOTA2) 2125 error = UFS_WAPBL_BEGIN(mp); 2126 if (error) { 2127 ffs_deinit_vnode(ump, vp); 2128 2129 return error; 2130 } 2131 error = chkiq(ip, 1, cred, 0); 2132 if (error) { 2133 ffs_vfree(dvp, ino, mode); 2134 UFS_WAPBL_END(mp); 2135 ffs_deinit_vnode(ump, vp); 2136 2137 return error; 2138 } 2139 UFS_WAPBL_END(mp); 2140 #endif 2141 2142 /* Set type and finalize. */ 2143 ip->i_flags = 0; 2144 DIP_ASSIGN(ip, flags, 0); 2145 ip->i_mode = mode; 2146 DIP_ASSIGN(ip, mode, mode); 2147 if (vap->va_rdev != VNOVAL) { 2148 /* 2149 * Want to be able to use this to make badblock 2150 * inodes, so don't truncate the dev number. 2151 */ 2152 if (ump->um_fstype == UFS1) 2153 ip->i_ffs1_rdev = ufs_rw32(vap->va_rdev, 2154 UFS_MPNEEDSWAP(ump)); 2155 else 2156 ip->i_ffs2_rdev = ufs_rw64(vap->va_rdev, 2157 UFS_MPNEEDSWAP(ump)); 2158 } 2159 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp); 2160 ip->i_devvp = ump->um_devvp; 2161 vref(ip->i_devvp); 2162 2163 /* Set up a new generation number for this inode. */ 2164 ip->i_gen++; 2165 DIP_ASSIGN(ip, gen, ip->i_gen); 2166 if (fs->fs_magic == FS_UFS2_MAGIC) { 2167 vfs_timestamp(&ts); 2168 ip->i_ffs2_birthtime = ts.tv_sec; 2169 ip->i_ffs2_birthnsec = ts.tv_nsec; 2170 } 2171 2172 uvm_vnp_setsize(vp, ip->i_size); 2173 *new_key = &ip->i_number; 2174 return 0; 2175 } 2176 2177 /* 2178 * File handle to vnode 2179 * 2180 * Have to be really careful about stale file handles: 2181 * - check that the inode number is valid 2182 * - call ffs_vget() to get the locked inode 2183 * - check for an unallocated inode (i_mode == 0) 2184 * - check that the given client host has export rights and return 2185 * those rights via. exflagsp and credanonp 2186 */ 2187 int 2188 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp) 2189 { 2190 struct ufid ufh; 2191 struct fs *fs; 2192 2193 if (fhp->fid_len != sizeof(struct ufid)) 2194 return EINVAL; 2195 2196 memcpy(&ufh, fhp, sizeof(ufh)); 2197 fs = VFSTOUFS(mp)->um_fs; 2198 if (ufh.ufid_ino < UFS_ROOTINO || 2199 ufh.ufid_ino >= fs->fs_ncg * fs->fs_ipg) 2200 return (ESTALE); 2201 return (ufs_fhtovp(mp, &ufh, vpp)); 2202 } 2203 2204 /* 2205 * Vnode pointer to File handle 2206 */ 2207 /* ARGSUSED */ 2208 int 2209 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size) 2210 { 2211 struct inode *ip; 2212 struct ufid ufh; 2213 2214 if (*fh_size < sizeof(struct ufid)) { 2215 *fh_size = sizeof(struct ufid); 2216 return E2BIG; 2217 } 2218 ip = VTOI(vp); 2219 *fh_size = sizeof(struct ufid); 2220 memset(&ufh, 0, sizeof(ufh)); 2221 ufh.ufid_len = sizeof(struct ufid); 2222 ufh.ufid_ino = ip->i_number; 2223 ufh.ufid_gen = ip->i_gen; 2224 memcpy(fhp, &ufh, sizeof(ufh)); 2225 return (0); 2226 } 2227 2228 void 2229 ffs_init(void) 2230 { 2231 if (ffs_initcount++ > 0) 2232 return; 2233 2234 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0, 2235 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL); 2236 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0, 2237 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL); 2238 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0, 2239 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL); 2240 ufs_init(); 2241 } 2242 2243 void 2244 ffs_reinit(void) 2245 { 2246 ufs_reinit(); 2247 } 2248 2249 void 2250 ffs_done(void) 2251 { 2252 if (--ffs_initcount > 0) 2253 return; 2254 2255 ufs_done(); 2256 pool_cache_destroy(ffs_dinode2_cache); 2257 pool_cache_destroy(ffs_dinode1_cache); 2258 pool_cache_destroy(ffs_inode_cache); 2259 } 2260 2261 /* 2262 * Write a superblock and associated information back to disk. 2263 */ 2264 int 2265 ffs_sbupdate(struct ufsmount *mp, int waitfor) 2266 { 2267 struct fs *fs = mp->um_fs; 2268 struct buf *bp; 2269 int error = 0; 2270 u_int32_t saveflag; 2271 2272 error = ffs_getblk(mp->um_devvp, 2273 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK, 2274 fs->fs_sbsize, false, &bp); 2275 if (error) 2276 return error; 2277 saveflag = fs->fs_flags & FS_INTERNAL; 2278 fs->fs_flags &= ~FS_INTERNAL; 2279 2280 memcpy(bp->b_data, fs, fs->fs_sbsize); 2281 2282 ffs_oldfscompat_write((struct fs *)bp->b_data, mp); 2283 #ifdef FFS_EI 2284 if (mp->um_flags & UFS_NEEDSWAP) 2285 ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data); 2286 #endif 2287 fs->fs_flags |= saveflag; 2288 2289 if (waitfor == MNT_WAIT) 2290 error = bwrite(bp); 2291 else 2292 bawrite(bp); 2293 return (error); 2294 } 2295 2296 int 2297 ffs_cgupdate(struct ufsmount *mp, int waitfor) 2298 { 2299 struct fs *fs = mp->um_fs; 2300 struct buf *bp; 2301 int blks; 2302 void *space; 2303 int i, size, error = 0, allerror = 0; 2304 2305 allerror = ffs_sbupdate(mp, waitfor); 2306 blks = howmany(fs->fs_cssize, fs->fs_fsize); 2307 space = fs->fs_csp; 2308 for (i = 0; i < blks; i += fs->fs_frag) { 2309 size = fs->fs_bsize; 2310 if (i + fs->fs_frag > blks) 2311 size = (blks - i) * fs->fs_fsize; 2312 error = ffs_getblk(mp->um_devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), 2313 FFS_NOBLK, size, false, &bp); 2314 if (error) 2315 break; 2316 #ifdef FFS_EI 2317 if (mp->um_flags & UFS_NEEDSWAP) 2318 ffs_csum_swap((struct csum*)space, 2319 (struct csum*)bp->b_data, size); 2320 else 2321 #endif 2322 memcpy(bp->b_data, space, (u_int)size); 2323 space = (char *)space + size; 2324 if (waitfor == MNT_WAIT) 2325 error = bwrite(bp); 2326 else 2327 bawrite(bp); 2328 } 2329 if (!allerror && error) 2330 allerror = error; 2331 return (allerror); 2332 } 2333 2334 int 2335 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp, 2336 int attrnamespace, const char *attrname) 2337 { 2338 #ifdef UFS_EXTATTR 2339 /* 2340 * File-backed extended attributes are only supported on UFS1. 2341 * UFS2 has native extended attributes. 2342 */ 2343 if (VFSTOUFS(mp)->um_fstype == UFS1) 2344 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname)); 2345 #endif 2346 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname)); 2347 } 2348 2349 int 2350 ffs_suspendctl(struct mount *mp, int cmd) 2351 { 2352 int error; 2353 struct lwp *l = curlwp; 2354 2355 switch (cmd) { 2356 case SUSPEND_SUSPEND: 2357 if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0) 2358 return error; 2359 error = ffs_sync(mp, MNT_WAIT, l->l_proc->p_cred); 2360 if (error == 0) 2361 error = fstrans_setstate(mp, FSTRANS_SUSPENDED); 2362 #ifdef WAPBL 2363 if (error == 0 && mp->mnt_wapbl) 2364 error = wapbl_flush(mp->mnt_wapbl, 1); 2365 #endif 2366 if (error != 0) { 2367 (void) fstrans_setstate(mp, FSTRANS_NORMAL); 2368 return error; 2369 } 2370 return 0; 2371 2372 case SUSPEND_RESUME: 2373 return fstrans_setstate(mp, FSTRANS_NORMAL); 2374 2375 default: 2376 return EINVAL; 2377 } 2378 } 2379 2380 /* 2381 * Synch vnode for a mounted file system. 2382 */ 2383 static int 2384 ffs_vfs_fsync(vnode_t *vp, int flags) 2385 { 2386 int error, i, pflags; 2387 #ifdef WAPBL 2388 struct mount *mp; 2389 #endif 2390 2391 KASSERT(vp->v_type == VBLK); 2392 KASSERT(spec_node_getmountedfs(vp) != NULL); 2393 2394 /* 2395 * Flush all dirty data associated with the vnode. 2396 */ 2397 pflags = PGO_ALLPAGES | PGO_CLEANIT; 2398 if ((flags & FSYNC_WAIT) != 0) 2399 pflags |= PGO_SYNCIO; 2400 mutex_enter(vp->v_interlock); 2401 error = VOP_PUTPAGES(vp, 0, 0, pflags); 2402 if (error) 2403 return error; 2404 2405 #ifdef WAPBL 2406 mp = spec_node_getmountedfs(vp); 2407 if (mp && mp->mnt_wapbl) { 2408 /* 2409 * Don't bother writing out metadata if the syncer is 2410 * making the request. We will let the sync vnode 2411 * write it out in a single burst through a call to 2412 * VFS_SYNC(). 2413 */ 2414 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0) 2415 return 0; 2416 2417 /* 2418 * Don't flush the log if the vnode being flushed 2419 * contains no dirty buffers that could be in the log. 2420 */ 2421 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 2422 error = wapbl_flush(mp->mnt_wapbl, 0); 2423 if (error) 2424 return error; 2425 } 2426 2427 if ((flags & FSYNC_WAIT) != 0) { 2428 mutex_enter(vp->v_interlock); 2429 while (vp->v_numoutput) 2430 cv_wait(&vp->v_cv, vp->v_interlock); 2431 mutex_exit(vp->v_interlock); 2432 } 2433 2434 return 0; 2435 } 2436 #endif /* WAPBL */ 2437 2438 error = vflushbuf(vp, flags); 2439 if (error == 0 && (flags & FSYNC_CACHE) != 0) { 2440 i = 1; 2441 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE, 2442 kauth_cred_get()); 2443 } 2444 2445 return error; 2446 } 2447
Indexes created Thu Oct 02 14:10:14 GMT 2025