ffs_vfsops.c revision 1.324
1 /* $NetBSD: ffs_vfsops.c,v 1.324 2015/03/15 09:21:01 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 1.324 2015/03/15 09:21:01 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 ffs_vfs_fsync(vnode_t *, int); 115 static int ffs_superblock_validate(struct fs *); 116 117 static struct sysctllog *ffs_sysctl_log; 118 119 static kauth_listener_t ffs_snapshot_listener; 120 121 /* how many times ffs_init() was called */ 122 int ffs_initcount = 0; 123 124 #ifdef DEBUG_FFS_MOUNT 125 #define DPRINTF(_fmt, args...) printf("%s: " _fmt "\n", __func__, ##args) 126 #else 127 #define DPRINTF(_fmt, args...) do {} while (/*CONSTCOND*/0) 128 #endif 129 130 extern const struct vnodeopv_desc ffs_vnodeop_opv_desc; 131 extern const struct vnodeopv_desc ffs_specop_opv_desc; 132 extern const struct vnodeopv_desc ffs_fifoop_opv_desc; 133 134 const struct vnodeopv_desc * const ffs_vnodeopv_descs[] = { 135 &ffs_vnodeop_opv_desc, 136 &ffs_specop_opv_desc, 137 &ffs_fifoop_opv_desc, 138 NULL, 139 }; 140 141 struct vfsops ffs_vfsops = { 142 .vfs_name = MOUNT_FFS, 143 .vfs_min_mount_data = sizeof (struct ufs_args), 144 .vfs_mount = ffs_mount, 145 .vfs_start = ufs_start, 146 .vfs_unmount = ffs_unmount, 147 .vfs_root = ufs_root, 148 .vfs_quotactl = ufs_quotactl, 149 .vfs_statvfs = ffs_statvfs, 150 .vfs_sync = ffs_sync, 151 .vfs_vget = ufs_vget, 152 .vfs_loadvnode = ffs_loadvnode, 153 .vfs_fhtovp = ffs_fhtovp, 154 .vfs_vptofh = ffs_vptofh, 155 .vfs_init = ffs_init, 156 .vfs_reinit = ffs_reinit, 157 .vfs_done = ffs_done, 158 .vfs_mountroot = ffs_mountroot, 159 .vfs_snapshot = ffs_snapshot, 160 .vfs_extattrctl = ffs_extattrctl, 161 .vfs_suspendctl = ffs_suspendctl, 162 .vfs_renamelock_enter = genfs_renamelock_enter, 163 .vfs_renamelock_exit = genfs_renamelock_exit, 164 .vfs_fsync = ffs_vfs_fsync, 165 .vfs_opv_descs = ffs_vnodeopv_descs 166 }; 167 168 static const struct genfs_ops ffs_genfsops = { 169 .gop_size = ffs_gop_size, 170 .gop_alloc = ufs_gop_alloc, 171 .gop_write = genfs_gop_write, 172 .gop_markupdate = ufs_gop_markupdate, 173 }; 174 175 static const struct ufs_ops ffs_ufsops = { 176 .uo_itimes = ffs_itimes, 177 .uo_update = ffs_update, 178 .uo_truncate = ffs_truncate, 179 .uo_valloc = ffs_valloc, 180 .uo_vfree = ffs_vfree, 181 .uo_balloc = ffs_balloc, 182 .uo_snapgone = ffs_snapgone, 183 }; 184 185 static int 186 ffs_snapshot_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 187 void *arg0, void *arg1, void *arg2, void *arg3) 188 { 189 vnode_t *vp = arg2; 190 int result = KAUTH_RESULT_DEFER; 191 192 if (action != KAUTH_SYSTEM_FS_SNAPSHOT) 193 return result; 194 195 if (VTOI(vp)->i_uid == kauth_cred_geteuid(cred)) 196 result = KAUTH_RESULT_ALLOW; 197 198 return result; 199 } 200 201 static int 202 ffs_modcmd(modcmd_t cmd, void *arg) 203 { 204 int error; 205 206 #if 0 207 extern int doasyncfree; 208 #endif 209 #ifdef UFS_EXTATTR 210 extern int ufs_extattr_autocreate; 211 #endif 212 extern int ffs_log_changeopt; 213 214 switch (cmd) { 215 case MODULE_CMD_INIT: 216 error = vfs_attach(&ffs_vfsops); 217 if (error != 0) 218 break; 219 220 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 221 CTLFLAG_PERMANENT, 222 CTLTYPE_NODE, "ffs", 223 SYSCTL_DESCR("Berkeley Fast File System"), 224 NULL, 0, NULL, 0, 225 CTL_VFS, 1, CTL_EOL); 226 /* 227 * @@@ should we even bother with these first three? 228 */ 229 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 230 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 231 CTLTYPE_INT, "doclusterread", NULL, 232 sysctl_notavail, 0, NULL, 0, 233 CTL_VFS, 1, FFS_CLUSTERREAD, CTL_EOL); 234 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 235 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 236 CTLTYPE_INT, "doclusterwrite", NULL, 237 sysctl_notavail, 0, NULL, 0, 238 CTL_VFS, 1, FFS_CLUSTERWRITE, CTL_EOL); 239 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 240 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 241 CTLTYPE_INT, "doreallocblks", NULL, 242 sysctl_notavail, 0, NULL, 0, 243 CTL_VFS, 1, FFS_REALLOCBLKS, CTL_EOL); 244 #if 0 245 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 246 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 247 CTLTYPE_INT, "doasyncfree", 248 SYSCTL_DESCR("Release dirty blocks asynchronously"), 249 NULL, 0, &doasyncfree, 0, 250 CTL_VFS, 1, FFS_ASYNCFREE, CTL_EOL); 251 #endif 252 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 253 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 254 CTLTYPE_INT, "log_changeopt", 255 SYSCTL_DESCR("Log changes in optimization strategy"), 256 NULL, 0, &ffs_log_changeopt, 0, 257 CTL_VFS, 1, FFS_LOG_CHANGEOPT, CTL_EOL); 258 #ifdef UFS_EXTATTR 259 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL, 260 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 261 CTLTYPE_INT, "extattr_autocreate", 262 SYSCTL_DESCR("Size of attribute for " 263 "backing file autocreation"), 264 NULL, 0, &ufs_extattr_autocreate, 0, 265 CTL_VFS, 1, FFS_EXTATTR_AUTOCREATE, CTL_EOL); 266 267 #endif /* UFS_EXTATTR */ 268 269 ffs_snapshot_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM, 270 ffs_snapshot_cb, NULL); 271 if (ffs_snapshot_listener == NULL) 272 printf("ffs_modcmd: can't listen on system scope.\n"); 273 274 break; 275 case MODULE_CMD_FINI: 276 error = vfs_detach(&ffs_vfsops); 277 if (error != 0) 278 break; 279 sysctl_teardown(&ffs_sysctl_log); 280 if (ffs_snapshot_listener != NULL) 281 kauth_unlisten_scope(ffs_snapshot_listener); 282 break; 283 default: 284 error = ENOTTY; 285 break; 286 } 287 288 return (error); 289 } 290 291 pool_cache_t ffs_inode_cache; 292 pool_cache_t ffs_dinode1_cache; 293 pool_cache_t ffs_dinode2_cache; 294 295 static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t); 296 static void ffs_oldfscompat_write(struct fs *, struct ufsmount *); 297 298 /* 299 * Called by main() when ffs is going to be mounted as root. 300 */ 301 302 int 303 ffs_mountroot(void) 304 { 305 struct fs *fs; 306 struct mount *mp; 307 struct lwp *l = curlwp; /* XXX */ 308 struct ufsmount *ump; 309 int error; 310 311 if (device_class(root_device) != DV_DISK) 312 return (ENODEV); 313 314 if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp))) { 315 vrele(rootvp); 316 return (error); 317 } 318 319 /* 320 * We always need to be able to mount the root file system. 321 */ 322 mp->mnt_flag |= MNT_FORCE; 323 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) { 324 vfs_unbusy(mp, false, NULL); 325 vfs_destroy(mp); 326 return (error); 327 } 328 mp->mnt_flag &= ~MNT_FORCE; 329 mountlist_append(mp); 330 ump = VFSTOUFS(mp); 331 fs = ump->um_fs; 332 memset(fs->fs_fsmnt, 0, sizeof(fs->fs_fsmnt)); 333 (void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0); 334 (void)ffs_statvfs(mp, &mp->mnt_stat); 335 vfs_unbusy(mp, false, NULL); 336 setrootfstime((time_t)fs->fs_time); 337 return (0); 338 } 339 340 /* 341 * VFS Operations. 342 * 343 * mount system call 344 */ 345 int 346 ffs_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 347 { 348 struct lwp *l = curlwp; 349 struct vnode *devvp = NULL; 350 struct ufs_args *args = data; 351 struct ufsmount *ump = NULL; 352 struct fs *fs; 353 int error = 0, flags, update; 354 mode_t accessmode; 355 356 if (args == NULL) { 357 DPRINTF("NULL args"); 358 return EINVAL; 359 } 360 if (*data_len < sizeof(*args)) { 361 DPRINTF("bad size args %zu != %zu", *data_len, sizeof(*args)); 362 return EINVAL; 363 } 364 365 if (mp->mnt_flag & MNT_GETARGS) { 366 ump = VFSTOUFS(mp); 367 if (ump == NULL) { 368 DPRINTF("no ump"); 369 return EIO; 370 } 371 args->fspec = NULL; 372 *data_len = sizeof *args; 373 return 0; 374 } 375 376 update = mp->mnt_flag & MNT_UPDATE; 377 378 /* Check arguments */ 379 if (args->fspec != NULL) { 380 /* 381 * Look up the name and verify that it's sane. 382 */ 383 error = namei_simple_user(args->fspec, 384 NSM_FOLLOW_NOEMULROOT, &devvp); 385 if (error != 0) { 386 DPRINTF("namei_simple_user returned %d", error); 387 return error; 388 } 389 390 if (!update) { 391 /* 392 * Be sure this is a valid block device 393 */ 394 if (devvp->v_type != VBLK) { 395 DPRINTF("non block device %d", devvp->v_type); 396 error = ENOTBLK; 397 } else if (bdevsw_lookup(devvp->v_rdev) == NULL) { 398 DPRINTF("can't find block device 0x%jx", 399 devvp->v_rdev); 400 error = ENXIO; 401 } 402 } else { 403 /* 404 * Be sure we're still naming the same device 405 * used for our initial mount 406 */ 407 ump = VFSTOUFS(mp); 408 if (devvp != ump->um_devvp) { 409 if (devvp->v_rdev != ump->um_devvp->v_rdev) { 410 DPRINTF("wrong device 0x%jx != 0x%jx", 411 (uintmax_t)devvp->v_rdev, 412 (uintmax_t)ump->um_devvp->v_rdev); 413 error = EINVAL; 414 } else { 415 vrele(devvp); 416 devvp = ump->um_devvp; 417 vref(devvp); 418 } 419 } 420 } 421 } else { 422 if (!update) { 423 /* New mounts must have a filename for the device */ 424 DPRINTF("no filename for mount"); 425 return EINVAL; 426 } else { 427 /* Use the extant mount */ 428 ump = VFSTOUFS(mp); 429 devvp = ump->um_devvp; 430 vref(devvp); 431 } 432 } 433 434 /* 435 * If mount by non-root, then verify that user has necessary 436 * permissions on the device. 437 * 438 * Permission to update a mount is checked higher, so here we presume 439 * updating the mount is okay (for example, as far as securelevel goes) 440 * which leaves us with the normal check. 441 */ 442 if (error == 0) { 443 accessmode = VREAD; 444 if (update ? 445 (mp->mnt_iflag & IMNT_WANTRDWR) != 0 : 446 (mp->mnt_flag & MNT_RDONLY) == 0) 447 accessmode |= VWRITE; 448 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 449 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 450 KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, 451 KAUTH_ARG(accessmode)); 452 if (error) { 453 DPRINTF("kauth returned %d", error); 454 } 455 VOP_UNLOCK(devvp); 456 } 457 458 if (error) { 459 vrele(devvp); 460 return (error); 461 } 462 463 #ifdef WAPBL 464 /* WAPBL can only be enabled on a r/w mount. */ 465 if ((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) { 466 mp->mnt_flag &= ~MNT_LOG; 467 } 468 #else /* !WAPBL */ 469 mp->mnt_flag &= ~MNT_LOG; 470 #endif /* !WAPBL */ 471 472 if (!update) { 473 int xflags; 474 475 if (mp->mnt_flag & MNT_RDONLY) 476 xflags = FREAD; 477 else 478 xflags = FREAD | FWRITE; 479 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 480 error = VOP_OPEN(devvp, xflags, FSCRED); 481 VOP_UNLOCK(devvp); 482 if (error) { 483 DPRINTF("VOP_OPEN returned %d", error); 484 goto fail; 485 } 486 error = ffs_mountfs(devvp, mp, l); 487 if (error) { 488 DPRINTF("ffs_mountfs returned %d", error); 489 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 490 (void)VOP_CLOSE(devvp, xflags, NOCRED); 491 VOP_UNLOCK(devvp); 492 goto fail; 493 } 494 495 ump = VFSTOUFS(mp); 496 fs = ump->um_fs; 497 } else { 498 /* 499 * Update the mount. 500 */ 501 502 /* 503 * The initial mount got a reference on this 504 * device, so drop the one obtained via 505 * namei(), above. 506 */ 507 vrele(devvp); 508 509 ump = VFSTOUFS(mp); 510 fs = ump->um_fs; 511 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 512 /* 513 * Changing from r/w to r/o 514 */ 515 flags = WRITECLOSE; 516 if (mp->mnt_flag & MNT_FORCE) 517 flags |= FORCECLOSE; 518 error = ffs_flushfiles(mp, flags, l); 519 if (error == 0) 520 error = UFS_WAPBL_BEGIN(mp); 521 if (error == 0 && 522 ffs_cgupdate(ump, MNT_WAIT) == 0 && 523 fs->fs_clean & FS_WASCLEAN) { 524 if (mp->mnt_flag & MNT_SOFTDEP) 525 fs->fs_flags &= ~FS_DOSOFTDEP; 526 fs->fs_clean = FS_ISCLEAN; 527 (void) ffs_sbupdate(ump, MNT_WAIT); 528 } 529 if (error) { 530 DPRINTF("wapbl %d", error); 531 return error; 532 } 533 UFS_WAPBL_END(mp); 534 } 535 536 #ifdef WAPBL 537 if ((mp->mnt_flag & MNT_LOG) == 0) { 538 error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE); 539 if (error) { 540 DPRINTF("ffs_wapbl_stop returned %d", error); 541 return error; 542 } 543 } 544 #endif /* WAPBL */ 545 546 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 547 /* 548 * Finish change from r/w to r/o 549 */ 550 fs->fs_ronly = 1; 551 fs->fs_fmod = 0; 552 } 553 554 if (mp->mnt_flag & MNT_RELOAD) { 555 error = ffs_reload(mp, l->l_cred, l); 556 if (error) { 557 DPRINTF("ffs_reload returned %d", error); 558 return error; 559 } 560 } 561 562 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) { 563 /* 564 * Changing from read-only to read/write 565 */ 566 #ifndef QUOTA2 567 if (fs->fs_flags & FS_DOQUOTA2) { 568 ump->um_flags |= UFS_QUOTA2; 569 uprintf("%s: options QUOTA2 not enabled%s\n", 570 mp->mnt_stat.f_mntonname, 571 (mp->mnt_flag & MNT_FORCE) ? "" : 572 ", not mounting"); 573 DPRINTF("ffs_quota2 %d", EINVAL); 574 return EINVAL; 575 } 576 #endif 577 fs->fs_ronly = 0; 578 fs->fs_clean <<= 1; 579 fs->fs_fmod = 1; 580 #ifdef WAPBL 581 if (fs->fs_flags & FS_DOWAPBL) { 582 const char *nm = mp->mnt_stat.f_mntonname; 583 if (!mp->mnt_wapbl_replay) { 584 printf("%s: log corrupted;" 585 " replay cancelled\n", nm); 586 return EFTYPE; 587 } 588 printf("%s: replaying log to disk\n", nm); 589 error = wapbl_replay_write(mp->mnt_wapbl_replay, 590 devvp); 591 if (error) { 592 DPRINTF("%s: wapbl_replay_write %d", 593 nm, error); 594 return error; 595 } 596 wapbl_replay_stop(mp->mnt_wapbl_replay); 597 fs->fs_clean = FS_WASCLEAN; 598 } 599 #endif /* WAPBL */ 600 if (fs->fs_snapinum[0] != 0) 601 ffs_snapshot_mount(mp); 602 } 603 604 #ifdef WAPBL 605 error = ffs_wapbl_start(mp); 606 if (error) { 607 DPRINTF("ffs_wapbl_start returned %d", error); 608 return error; 609 } 610 #endif /* WAPBL */ 611 612 #ifdef QUOTA2 613 if (!fs->fs_ronly) { 614 error = ffs_quota2_mount(mp); 615 if (error) { 616 DPRINTF("ffs_quota2_mount returned %d", error); 617 return error; 618 } 619 } 620 #endif 621 622 if ((mp->mnt_flag & MNT_DISCARD) && !(ump->um_discarddata)) 623 ump->um_discarddata = ffs_discard_init(devvp, fs); 624 625 if (args->fspec == NULL) 626 return 0; 627 } 628 629 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, 630 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l); 631 if (error == 0) 632 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, 633 sizeof(fs->fs_fsmnt)); 634 else { 635 DPRINTF("set_statvfs_info returned %d", error); 636 } 637 fs->fs_flags &= ~FS_DOSOFTDEP; 638 if (fs->fs_fmod != 0) { /* XXX */ 639 int err; 640 641 fs->fs_fmod = 0; 642 if (fs->fs_clean & FS_WASCLEAN) 643 fs->fs_time = time_second; 644 else { 645 printf("%s: file system not clean (fs_clean=%#x); " 646 "please fsck(8)\n", mp->mnt_stat.f_mntfromname, 647 fs->fs_clean); 648 printf("%s: lost blocks %" PRId64 " files %d\n", 649 mp->mnt_stat.f_mntfromname, fs->fs_pendingblocks, 650 fs->fs_pendinginodes); 651 } 652 err = UFS_WAPBL_BEGIN(mp); 653 if (err == 0) { 654 (void) ffs_cgupdate(ump, MNT_WAIT); 655 UFS_WAPBL_END(mp); 656 } 657 } 658 if ((mp->mnt_flag & MNT_SOFTDEP) != 0) { 659 printf("%s: `-o softdep' is no longer supported, " 660 "consider `-o log'\n", mp->mnt_stat.f_mntfromname); 661 mp->mnt_flag &= ~MNT_SOFTDEP; 662 } 663 664 return (error); 665 666 fail: 667 vrele(devvp); 668 return (error); 669 } 670 671 /* 672 * Reload all incore data for a filesystem (used after running fsck on 673 * the root filesystem and finding things to fix). The filesystem must 674 * be mounted read-only. 675 * 676 * Things to do to update the mount: 677 * 1) invalidate all cached meta-data. 678 * 2) re-read superblock from disk. 679 * 3) re-read summary information from disk. 680 * 4) invalidate all inactive vnodes. 681 * 5) invalidate all cached file data. 682 * 6) re-read inode data for all active vnodes. 683 */ 684 int 685 ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l) 686 { 687 struct vnode *vp, *devvp; 688 struct inode *ip; 689 void *space; 690 struct buf *bp; 691 struct fs *fs, *newfs; 692 struct dkwedge_info dkw; 693 int i, bsize, blks, error; 694 int32_t *lp, fs_sbsize; 695 struct ufsmount *ump; 696 daddr_t sblockloc; 697 struct vnode_iterator *marker; 698 699 if ((mp->mnt_flag & MNT_RDONLY) == 0) 700 return (EINVAL); 701 702 ump = VFSTOUFS(mp); 703 704 /* 705 * Step 1: invalidate all cached meta-data. 706 */ 707 devvp = ump->um_devvp; 708 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 709 error = vinvalbuf(devvp, 0, cred, l, 0, 0); 710 VOP_UNLOCK(devvp); 711 if (error) 712 panic("ffs_reload: dirty1"); 713 714 /* 715 * Step 2: re-read superblock from disk. XXX: We don't handle 716 * possibility that superblock moved. Which implies that we don't 717 * want its size to change either. 718 */ 719 fs = ump->um_fs; 720 fs_sbsize = fs->fs_sbsize; 721 error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs_sbsize, 722 NOCRED, 0, &bp); 723 if (error) 724 return (error); 725 newfs = kmem_alloc(fs_sbsize, KM_SLEEP); 726 memcpy(newfs, bp->b_data, fs_sbsize); 727 728 #ifdef FFS_EI 729 if (ump->um_flags & UFS_NEEDSWAP) { 730 ffs_sb_swap((struct fs *)bp->b_data, newfs); 731 newfs->fs_flags |= FS_SWAPPED; 732 } else 733 #endif 734 newfs->fs_flags &= ~FS_SWAPPED; 735 736 brelse(bp, 0); 737 738 if ((newfs->fs_magic != FS_UFS1_MAGIC) && 739 (newfs->fs_magic != FS_UFS2_MAGIC)) { 740 kmem_free(newfs, fs_sbsize); 741 return (EIO); /* XXX needs translation */ 742 } 743 if (!ffs_superblock_validate(newfs)) { 744 kmem_free(newfs, fs_sbsize); 745 return (EINVAL); 746 } 747 748 /* 749 * The current implementation doesn't handle the possibility that 750 * these values may have changed. 751 */ 752 if ((newfs->fs_sbsize != fs_sbsize) || 753 (newfs->fs_cssize != fs->fs_cssize) || 754 (newfs->fs_contigsumsize != fs->fs_contigsumsize) || 755 (newfs->fs_ncg != fs->fs_ncg)) { 756 kmem_free(newfs, fs_sbsize); 757 return (EINVAL); 758 } 759 760 /* Store off old fs_sblockloc for fs_oldfscompat_read. */ 761 sblockloc = fs->fs_sblockloc; 762 /* 763 * Copy pointer fields back into superblock before copying in XXX 764 * new superblock. These should really be in the ufsmount. XXX 765 * Note that important parameters (eg fs_ncg) are unchanged. 766 */ 767 newfs->fs_csp = fs->fs_csp; 768 newfs->fs_maxcluster = fs->fs_maxcluster; 769 newfs->fs_contigdirs = fs->fs_contigdirs; 770 newfs->fs_ronly = fs->fs_ronly; 771 newfs->fs_active = fs->fs_active; 772 memcpy(fs, newfs, (u_int)fs_sbsize); 773 kmem_free(newfs, fs_sbsize); 774 775 /* Recheck for apple UFS filesystem */ 776 ump->um_flags &= ~UFS_ISAPPLEUFS; 777 /* First check to see if this is tagged as an Apple UFS filesystem 778 * in the disklabel 779 */ 780 if (getdiskinfo(devvp, &dkw) == 0 && 781 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0) 782 ump->um_flags |= UFS_ISAPPLEUFS; 783 #ifdef APPLE_UFS 784 else { 785 /* Manually look for an apple ufs label, and if a valid one 786 * is found, then treat it like an Apple UFS filesystem anyway 787 * 788 * EINVAL is most probably a blocksize or alignment problem, 789 * it is unlikely that this is an Apple UFS filesystem then. 790 */ 791 error = bread(devvp, 792 (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE), 793 APPLEUFS_LABEL_SIZE, cred, 0, &bp); 794 if (error && error != EINVAL) { 795 return error; 796 } 797 if (error == 0) { 798 error = ffs_appleufs_validate(fs->fs_fsmnt, 799 (struct appleufslabel *)bp->b_data, NULL); 800 if (error == 0) 801 ump->um_flags |= UFS_ISAPPLEUFS; 802 brelse(bp, 0); 803 } 804 bp = NULL; 805 } 806 #else 807 if (ump->um_flags & UFS_ISAPPLEUFS) 808 return (EIO); 809 #endif 810 811 if (UFS_MPISAPPLEUFS(ump)) { 812 /* see comment about NeXT below */ 813 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN; 814 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ; 815 mp->mnt_iflag |= IMNT_DTYPE; 816 } else { 817 ump->um_maxsymlinklen = fs->fs_maxsymlinklen; 818 ump->um_dirblksiz = UFS_DIRBLKSIZ; 819 if (ump->um_maxsymlinklen > 0) 820 mp->mnt_iflag |= IMNT_DTYPE; 821 else 822 mp->mnt_iflag &= ~IMNT_DTYPE; 823 } 824 ffs_oldfscompat_read(fs, ump, sblockloc); 825 826 mutex_enter(&ump->um_lock); 827 ump->um_maxfilesize = fs->fs_maxfilesize; 828 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) { 829 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n", 830 mp->mnt_stat.f_mntonname, fs->fs_flags, 831 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 832 if ((mp->mnt_flag & MNT_FORCE) == 0) { 833 mutex_exit(&ump->um_lock); 834 return (EINVAL); 835 } 836 } 837 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 838 fs->fs_pendingblocks = 0; 839 fs->fs_pendinginodes = 0; 840 } 841 mutex_exit(&ump->um_lock); 842 843 ffs_statvfs(mp, &mp->mnt_stat); 844 /* 845 * Step 3: re-read summary information from disk. 846 */ 847 blks = howmany(fs->fs_cssize, fs->fs_fsize); 848 space = fs->fs_csp; 849 for (i = 0; i < blks; i += fs->fs_frag) { 850 bsize = fs->fs_bsize; 851 if (i + fs->fs_frag > blks) 852 bsize = (blks - i) * fs->fs_fsize; 853 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize, 854 NOCRED, 0, &bp); 855 if (error) { 856 return (error); 857 } 858 #ifdef FFS_EI 859 if (UFS_FSNEEDSWAP(fs)) 860 ffs_csum_swap((struct csum *)bp->b_data, 861 (struct csum *)space, bsize); 862 else 863 #endif 864 memcpy(space, bp->b_data, (size_t)bsize); 865 space = (char *)space + bsize; 866 brelse(bp, 0); 867 } 868 /* 869 * We no longer know anything about clusters per cylinder group. 870 */ 871 if (fs->fs_contigsumsize > 0) { 872 lp = fs->fs_maxcluster; 873 for (i = 0; i < fs->fs_ncg; i++) 874 *lp++ = fs->fs_contigsumsize; 875 } 876 877 vfs_vnode_iterator_init(mp, &marker); 878 while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) { 879 /* 880 * Step 4: invalidate all inactive vnodes. 881 */ 882 if (vrecycle(vp)) 883 continue; 884 /* 885 * Step 5: invalidate all cached file data. 886 */ 887 if (vn_lock(vp, LK_EXCLUSIVE)) { 888 vrele(vp); 889 continue; 890 } 891 if (vinvalbuf(vp, 0, cred, l, 0, 0)) 892 panic("ffs_reload: dirty2"); 893 /* 894 * Step 6: re-read inode data for all active vnodes. 895 */ 896 ip = VTOI(vp); 897 error = bread(devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)), 898 (int)fs->fs_bsize, NOCRED, 0, &bp); 899 if (error) { 900 vput(vp); 901 break; 902 } 903 ffs_load_inode(bp, ip, fs, ip->i_number); 904 brelse(bp, 0); 905 vput(vp); 906 } 907 vfs_vnode_iterator_destroy(marker); 908 return (error); 909 } 910 911 /* 912 * Possible superblock locations ordered from most to least likely. 913 */ 914 static const int sblock_try[] = SBLOCKSEARCH; 915 916 917 static int 918 ffs_superblock_validate(struct fs *fs) 919 { 920 int32_t i, fs_bshift = 0, fs_fshift = 0, fs_fragshift = 0, fs_frag; 921 int32_t fs_inopb; 922 923 /* Check the superblock size */ 924 if (fs->fs_sbsize > SBLOCKSIZE || fs->fs_sbsize < sizeof(struct fs)) 925 return 0; 926 927 /* Check the file system blocksize */ 928 if (fs->fs_bsize > MAXBSIZE || fs->fs_bsize < MINBSIZE) 929 return 0; 930 if (!powerof2(fs->fs_bsize)) 931 return 0; 932 933 /* Check the size of frag blocks */ 934 if (!powerof2(fs->fs_fsize)) 935 return 0; 936 if (fs->fs_fsize == 0) 937 return 0; 938 939 /* 940 * XXX: these values are just zero-checked to prevent obvious 941 * bugs. We need more strict checks. 942 */ 943 if (fs->fs_size == 0) 944 return 0; 945 if (fs->fs_cssize == 0) 946 return 0; 947 if (fs->fs_ipg == 0) 948 return 0; 949 if (fs->fs_fpg == 0) 950 return 0; 951 952 /* Check the number of inodes per block */ 953 if (fs->fs_magic == FS_UFS1_MAGIC) 954 fs_inopb = fs->fs_bsize / sizeof(struct ufs1_dinode); 955 else /* fs->fs_magic == FS_UFS2_MAGIC */ 956 fs_inopb = fs->fs_bsize / sizeof(struct ufs2_dinode); 957 if (fs->fs_inopb != fs_inopb) 958 return 0; 959 960 /* Block size cannot be smaller than fragment size */ 961 if (fs->fs_bsize < fs->fs_fsize) 962 return 0; 963 964 /* Compute fs_bshift and ensure it is consistent */ 965 for (i = fs->fs_bsize; i > 1; i >>= 1) 966 fs_bshift++; 967 if (fs->fs_bshift != fs_bshift) 968 return 0; 969 970 /* Compute fs_fshift and ensure it is consistent */ 971 for (i = fs->fs_fsize; i > 1; i >>= 1) 972 fs_fshift++; 973 if (fs->fs_fshift != fs_fshift) 974 return 0; 975 976 /* Compute fs_fragshift and ensure it is consistent */ 977 for (i = fs->fs_frag; i > 1; i >>= 1) 978 fs_fragshift++; 979 if (fs->fs_fragshift != fs_fragshift) 980 return 0; 981 982 /* Check the masks */ 983 if (fs->fs_bmask != ~(fs->fs_bsize - 1)) 984 return 0; 985 if (fs->fs_fmask != ~(fs->fs_fsize - 1)) 986 return 0; 987 988 /* 989 * Now that the shifts and masks are sanitized, we can use the ffs_ API. 990 */ 991 992 /* Check the number of frag blocks */ 993 if ((fs_frag = ffs_numfrags(fs, fs->fs_bsize)) > MAXFRAG) 994 return 0; 995 if (fs->fs_frag != fs_frag) 996 return 0; 997 998 return 1; 999 } 1000 1001 /* 1002 * Common code for mount and mountroot 1003 */ 1004 int 1005 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l) 1006 { 1007 struct ufsmount *ump = NULL; 1008 struct buf *bp = NULL; 1009 struct fs *fs = NULL; 1010 dev_t dev; 1011 struct dkwedge_info dkw; 1012 void *space; 1013 daddr_t sblockloc = 0; 1014 int blks, fstype = 0; 1015 int error, i, bsize, ronly, bset = 0; 1016 #ifdef FFS_EI 1017 int needswap = 0; /* keep gcc happy */ 1018 #endif 1019 int32_t *lp; 1020 kauth_cred_t cred; 1021 u_int32_t allocsbsize, fs_sbsize = 0; 1022 1023 dev = devvp->v_rdev; 1024 cred = l ? l->l_cred : NOCRED; 1025 1026 /* Flush out any old buffers remaining from a previous use. */ 1027 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); 1028 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0); 1029 VOP_UNLOCK(devvp); 1030 if (error) { 1031 DPRINTF("vinvalbuf returned %d", error); 1032 return error; 1033 } 1034 1035 ronly = (mp->mnt_flag & MNT_RDONLY) != 0; 1036 1037 error = fstrans_mount(mp); 1038 if (error) { 1039 DPRINTF("fstrans_mount returned %d", error); 1040 return error; 1041 } 1042 1043 ump = kmem_zalloc(sizeof(*ump), KM_SLEEP); 1044 mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE); 1045 error = ffs_snapshot_init(ump); 1046 if (error) { 1047 DPRINTF("ffs_snapshot_init returned %d", error); 1048 goto out; 1049 } 1050 ump->um_ops = &ffs_ufsops; 1051 1052 #ifdef WAPBL 1053 sbagain: 1054 #endif 1055 /* 1056 * Try reading the superblock in each of its possible locations. 1057 */ 1058 for (i = 0; ; i++) { 1059 daddr_t fs_sblockloc; 1060 1061 if (bp != NULL) { 1062 brelse(bp, BC_NOCACHE); 1063 bp = NULL; 1064 } 1065 if (sblock_try[i] == -1) { 1066 DPRINTF("no superblock found"); 1067 error = EINVAL; 1068 fs = NULL; 1069 goto out; 1070 } 1071 1072 error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, 1073 cred, 0, &bp); 1074 if (error) { 1075 DPRINTF("bread@0x%x returned %d", 1076 sblock_try[i] / DEV_BSIZE, error); 1077 fs = NULL; 1078 goto out; 1079 } 1080 fs = (struct fs *)bp->b_data; 1081 1082 sblockloc = sblock_try[i]; 1083 DPRINTF("fs_magic 0x%x", fs->fs_magic); 1084 1085 /* 1086 * Swap: here, we swap fs->fs_sbsize in order to get the correct 1087 * size to read the superblock. Once read, we swap the whole 1088 * superblock structure. 1089 */ 1090 if (fs->fs_magic == FS_UFS1_MAGIC) { 1091 fs_sbsize = fs->fs_sbsize; 1092 fstype = UFS1; 1093 #ifdef FFS_EI 1094 needswap = 0; 1095 } else if (fs->fs_magic == FS_UFS1_MAGIC_SWAPPED) { 1096 fs_sbsize = bswap32(fs->fs_sbsize); 1097 fstype = UFS1; 1098 needswap = 1; 1099 #endif 1100 } else if (fs->fs_magic == FS_UFS2_MAGIC) { 1101 fs_sbsize = fs->fs_sbsize; 1102 fstype = UFS2; 1103 #ifdef FFS_EI 1104 needswap = 0; 1105 } else if (fs->fs_magic == FS_UFS2_MAGIC_SWAPPED) { 1106 fs_sbsize = bswap32(fs->fs_sbsize); 1107 fstype = UFS2; 1108 needswap = 1; 1109 #endif 1110 } else 1111 continue; 1112 1113 /* fs->fs_sblockloc isn't defined for old filesystems */ 1114 if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) { 1115 if (sblockloc == SBLOCK_UFS2) 1116 /* 1117 * This is likely to be the first alternate 1118 * in a filesystem with 64k blocks. 1119 * Don't use it. 1120 */ 1121 continue; 1122 fs_sblockloc = sblockloc; 1123 } else { 1124 fs_sblockloc = fs->fs_sblockloc; 1125 #ifdef FFS_EI 1126 if (needswap) 1127 fs_sblockloc = bswap64(fs_sblockloc); 1128 #endif 1129 } 1130 1131 /* Check we haven't found an alternate superblock */ 1132 if (fs_sblockloc != sblockloc) 1133 continue; 1134 1135 /* Check the superblock size */ 1136 if (fs_sbsize > SBLOCKSIZE || fs_sbsize < sizeof(struct fs)) 1137 continue; 1138 fs = kmem_alloc((u_long)fs_sbsize, KM_SLEEP); 1139 memcpy(fs, bp->b_data, fs_sbsize); 1140 1141 /* Swap the whole superblock structure, if necessary. */ 1142 #ifdef FFS_EI 1143 if (needswap) { 1144 ffs_sb_swap((struct fs*)bp->b_data, fs); 1145 fs->fs_flags |= FS_SWAPPED; 1146 } else 1147 #endif 1148 fs->fs_flags &= ~FS_SWAPPED; 1149 1150 /* 1151 * Now that everything is swapped, the superblock is ready to 1152 * be sanitized. 1153 */ 1154 if (!ffs_superblock_validate(fs)) { 1155 kmem_free(fs, fs_sbsize); 1156 continue; 1157 } 1158 1159 /* Ok seems to be a good superblock */ 1160 break; 1161 } 1162 1163 ump->um_fs = fs; 1164 1165 #ifdef WAPBL 1166 if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) { 1167 error = ffs_wapbl_replay_start(mp, fs, devvp); 1168 if (error && (mp->mnt_flag & MNT_FORCE) == 0) { 1169 DPRINTF("ffs_wapbl_replay_start returned %d", error); 1170 goto out; 1171 } 1172 if (!error) { 1173 if (!ronly) { 1174 /* XXX fsmnt may be stale. */ 1175 printf("%s: replaying log to disk\n", 1176 fs->fs_fsmnt); 1177 error = wapbl_replay_write(mp->mnt_wapbl_replay, 1178 devvp); 1179 if (error) { 1180 DPRINTF("wapbl_replay_write returned %d", 1181 error); 1182 goto out; 1183 } 1184 wapbl_replay_stop(mp->mnt_wapbl_replay); 1185 fs->fs_clean = FS_WASCLEAN; 1186 } else { 1187 /* XXX fsmnt may be stale */ 1188 printf("%s: replaying log to memory\n", 1189 fs->fs_fsmnt); 1190 } 1191 1192 /* Force a re-read of the superblock */ 1193 brelse(bp, BC_INVAL); 1194 bp = NULL; 1195 kmem_free(fs, fs_sbsize); 1196 fs = NULL; 1197 goto sbagain; 1198 } 1199 } 1200 #else /* !WAPBL */ 1201 if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) { 1202 error = EPERM; 1203 DPRINTF("no force %d", error); 1204 goto out; 1205 } 1206 #endif /* !WAPBL */ 1207 1208 ffs_oldfscompat_read(fs, ump, sblockloc); 1209 ump->um_maxfilesize = fs->fs_maxfilesize; 1210 1211 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) { 1212 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n", 1213 mp->mnt_stat.f_mntonname, fs->fs_flags, 1214 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 1215 if ((mp->mnt_flag & MNT_FORCE) == 0) { 1216 error = EINVAL; 1217 DPRINTF("no force %d", error); 1218 goto out; 1219 } 1220 } 1221 1222 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) { 1223 fs->fs_pendingblocks = 0; 1224 fs->fs_pendinginodes = 0; 1225 } 1226 1227 ump->um_fstype = fstype; 1228 if (fs->fs_sbsize < SBLOCKSIZE) 1229 brelse(bp, BC_INVAL); 1230 else 1231 brelse(bp, 0); 1232 bp = NULL; 1233 1234 /* 1235 * First check to see if this is tagged as an Apple UFS filesystem 1236 * in the disklabel 1237 */ 1238 if (getdiskinfo(devvp, &dkw) == 0 && 1239 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0) 1240 ump->um_flags |= UFS_ISAPPLEUFS; 1241 #ifdef APPLE_UFS 1242 else { 1243 /* 1244 * Manually look for an apple ufs label, and if a valid one 1245 * is found, then treat it like an Apple UFS filesystem anyway 1246 */ 1247 error = bread(devvp, 1248 (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE), 1249 APPLEUFS_LABEL_SIZE, cred, 0, &bp); 1250 if (error) { 1251 DPRINTF("apple bread@0x%jx returned %d", 1252 (intmax_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE), 1253 error); 1254 goto out; 1255 } 1256 error = ffs_appleufs_validate(fs->fs_fsmnt, 1257 (struct appleufslabel *)bp->b_data, NULL); 1258 if (error == 0) 1259 ump->um_flags |= UFS_ISAPPLEUFS; 1260 brelse(bp, 0); 1261 bp = NULL; 1262 } 1263 #else 1264 if (ump->um_flags & UFS_ISAPPLEUFS) { 1265 DPRINTF("AppleUFS not supported"); 1266 error = EINVAL; 1267 goto out; 1268 } 1269 #endif 1270 1271 #if 0 1272 /* 1273 * XXX This code changes the behaviour of mounting dirty filesystems, to 1274 * XXX require "mount -f ..." to mount them. This doesn't match what 1275 * XXX mount(8) describes and is disabled for now. 1276 */ 1277 /* 1278 * If the file system is not clean, don't allow it to be mounted 1279 * unless MNT_FORCE is specified. (Note: MNT_FORCE is always set 1280 * for the root file system.) 1281 */ 1282 if (fs->fs_flags & FS_DOWAPBL) { 1283 /* 1284 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL 1285 * bit is set, although there's a window in unmount where it 1286 * could be FS_ISCLEAN 1287 */ 1288 if ((mp->mnt_flag & MNT_FORCE) == 0 && 1289 (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) { 1290 error = EPERM; 1291 goto out; 1292 } 1293 } else 1294 if ((fs->fs_clean & FS_ISCLEAN) == 0 && 1295 (mp->mnt_flag & MNT_FORCE) == 0) { 1296 error = EPERM; 1297 goto out; 1298 } 1299 #endif 1300 1301 /* 1302 * Verify that we can access the last block in the fs 1303 * if we're mounting read/write. 1304 */ 1305 if (!ronly) { 1306 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_size - 1), 1307 fs->fs_fsize, cred, 0, &bp); 1308 if (error) { 1309 DPRINTF("bread@0x%jx returned %d", 1310 (intmax_t)FFS_FSBTODB(fs, fs->fs_size - 1), 1311 error); 1312 bset = BC_INVAL; 1313 goto out; 1314 } 1315 if (bp->b_bcount != fs->fs_fsize) { 1316 DPRINTF("bcount %x != fsize %x", bp->b_bcount, 1317 fs->fs_fsize); 1318 error = EINVAL; 1319 } 1320 brelse(bp, BC_INVAL); 1321 bp = NULL; 1322 } 1323 1324 fs->fs_ronly = ronly; 1325 /* Don't bump fs_clean if we're replaying journal */ 1326 if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN))) { 1327 if (ronly == 0) { 1328 fs->fs_clean <<= 1; 1329 fs->fs_fmod = 1; 1330 } 1331 } 1332 1333 bsize = fs->fs_cssize; 1334 blks = howmany(bsize, fs->fs_fsize); 1335 if (fs->fs_contigsumsize > 0) 1336 bsize += fs->fs_ncg * sizeof(int32_t); 1337 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1338 allocsbsize = bsize; 1339 space = kmem_alloc((u_long)allocsbsize, KM_SLEEP); 1340 fs->fs_csp = space; 1341 1342 for (i = 0; i < blks; i += fs->fs_frag) { 1343 bsize = fs->fs_bsize; 1344 if (i + fs->fs_frag > blks) 1345 bsize = (blks - i) * fs->fs_fsize; 1346 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize, 1347 cred, 0, &bp); 1348 if (error) { 1349 DPRINTF("bread@0x%jx %d", 1350 (intmax_t)FFS_FSBTODB(fs, fs->fs_csaddr + i), 1351 error); 1352 goto out1; 1353 } 1354 #ifdef FFS_EI 1355 if (needswap) 1356 ffs_csum_swap((struct csum *)bp->b_data, 1357 (struct csum *)space, bsize); 1358 else 1359 #endif 1360 memcpy(space, bp->b_data, (u_int)bsize); 1361 1362 space = (char *)space + bsize; 1363 brelse(bp, 0); 1364 bp = NULL; 1365 } 1366 if (fs->fs_contigsumsize > 0) { 1367 fs->fs_maxcluster = lp = space; 1368 for (i = 0; i < fs->fs_ncg; i++) 1369 *lp++ = fs->fs_contigsumsize; 1370 space = lp; 1371 } 1372 bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1373 fs->fs_contigdirs = space; 1374 space = (char *)space + bsize; 1375 memset(fs->fs_contigdirs, 0, bsize); 1376 1377 /* Compatibility for old filesystems - XXX */ 1378 if (fs->fs_avgfilesize <= 0) 1379 fs->fs_avgfilesize = AVFILESIZ; 1380 if (fs->fs_avgfpdir <= 0) 1381 fs->fs_avgfpdir = AFPDIR; 1382 fs->fs_active = NULL; 1383 1384 mp->mnt_data = ump; 1385 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev; 1386 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS); 1387 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0]; 1388 mp->mnt_stat.f_namemax = FFS_MAXNAMLEN; 1389 if (UFS_MPISAPPLEUFS(ump)) { 1390 /* NeXT used to keep short symlinks in the inode even 1391 * when using FS_42INODEFMT. In that case fs->fs_maxsymlinklen 1392 * is probably -1, but we still need to be able to identify 1393 * short symlinks. 1394 */ 1395 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN; 1396 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ; 1397 mp->mnt_iflag |= IMNT_DTYPE; 1398 } else { 1399 ump->um_maxsymlinklen = fs->fs_maxsymlinklen; 1400 ump->um_dirblksiz = UFS_DIRBLKSIZ; 1401 if (ump->um_maxsymlinklen > 0) 1402 mp->mnt_iflag |= IMNT_DTYPE; 1403 else 1404 mp->mnt_iflag &= ~IMNT_DTYPE; 1405 } 1406 mp->mnt_fs_bshift = fs->fs_bshift; 1407 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */ 1408 mp->mnt_flag |= MNT_LOCAL; 1409 mp->mnt_iflag |= IMNT_MPSAFE; 1410 #ifdef FFS_EI 1411 if (needswap) 1412 ump->um_flags |= UFS_NEEDSWAP; 1413 #endif 1414 ump->um_mountp = mp; 1415 ump->um_dev = dev; 1416 ump->um_devvp = devvp; 1417 ump->um_nindir = fs->fs_nindir; 1418 ump->um_lognindir = ffs(fs->fs_nindir) - 1; 1419 ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT; 1420 ump->um_seqinc = fs->fs_frag; 1421 for (i = 0; i < MAXQUOTAS; i++) 1422 ump->um_quotas[i] = NULLVP; 1423 spec_node_setmountedfs(devvp, mp); 1424 if (ronly == 0 && fs->fs_snapinum[0] != 0) 1425 ffs_snapshot_mount(mp); 1426 #ifdef WAPBL 1427 if (!ronly) { 1428 KDASSERT(fs->fs_ronly == 0); 1429 /* 1430 * ffs_wapbl_start() needs mp->mnt_stat initialised if it 1431 * needs to create a new log file in-filesystem. 1432 */ 1433 error = ffs_statvfs(mp, &mp->mnt_stat); 1434 if (error) { 1435 DPRINTF("ffs_statvfs returned %d", error); 1436 goto out1; 1437 } 1438 1439 error = ffs_wapbl_start(mp); 1440 if (error) { 1441 DPRINTF("ffs_wapbl_start returned %d", error); 1442 goto out1; 1443 } 1444 } 1445 #endif /* WAPBL */ 1446 if (ronly == 0) { 1447 #ifdef QUOTA2 1448 error = ffs_quota2_mount(mp); 1449 if (error) { 1450 DPRINTF("ffs_quota2_mount returned %d", error); 1451 goto out1; 1452 } 1453 #else 1454 if (fs->fs_flags & FS_DOQUOTA2) { 1455 ump->um_flags |= UFS_QUOTA2; 1456 uprintf("%s: options QUOTA2 not enabled%s\n", 1457 mp->mnt_stat.f_mntonname, 1458 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting"); 1459 if ((mp->mnt_flag & MNT_FORCE) == 0) { 1460 error = EINVAL; 1461 DPRINTF("quota disabled %d", error); 1462 goto out1; 1463 } 1464 } 1465 #endif 1466 } 1467 1468 if (mp->mnt_flag & MNT_DISCARD) 1469 ump->um_discarddata = ffs_discard_init(devvp, fs); 1470 1471 return (0); 1472 out1: 1473 kmem_free(fs->fs_csp, allocsbsize); 1474 out: 1475 #ifdef WAPBL 1476 if (mp->mnt_wapbl_replay) { 1477 wapbl_replay_stop(mp->mnt_wapbl_replay); 1478 wapbl_replay_free(mp->mnt_wapbl_replay); 1479 mp->mnt_wapbl_replay = 0; 1480 } 1481 #endif 1482 1483 fstrans_unmount(mp); 1484 if (fs) 1485 kmem_free(fs, fs->fs_sbsize); 1486 spec_node_setmountedfs(devvp, NULL); 1487 if (bp) 1488 brelse(bp, bset); 1489 if (ump) { 1490 if (ump->um_oldfscompat) 1491 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t)); 1492 mutex_destroy(&ump->um_lock); 1493 kmem_free(ump, sizeof(*ump)); 1494 mp->mnt_data = NULL; 1495 } 1496 return (error); 1497 } 1498 1499 /* 1500 * Sanity checks for loading old filesystem superblocks. 1501 * See ffs_oldfscompat_write below for unwound actions. 1502 * 1503 * XXX - Parts get retired eventually. 1504 * Unfortunately new bits get added. 1505 */ 1506 static void 1507 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc) 1508 { 1509 off_t maxfilesize; 1510 int32_t *extrasave; 1511 1512 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1513 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1514 return; 1515 1516 if (!ump->um_oldfscompat) 1517 ump->um_oldfscompat = kmem_alloc(512 + 3*sizeof(int32_t), 1518 KM_SLEEP); 1519 1520 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512); 1521 extrasave = ump->um_oldfscompat; 1522 extrasave += 512/sizeof(int32_t); 1523 extrasave[0] = fs->fs_old_npsect; 1524 extrasave[1] = fs->fs_old_interleave; 1525 extrasave[2] = fs->fs_old_trackskew; 1526 1527 /* These fields will be overwritten by their 1528 * original values in fs_oldfscompat_write, so it is harmless 1529 * to modify them here. 1530 */ 1531 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir; 1532 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree; 1533 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree; 1534 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree; 1535 1536 fs->fs_maxbsize = fs->fs_bsize; 1537 fs->fs_time = fs->fs_old_time; 1538 fs->fs_size = fs->fs_old_size; 1539 fs->fs_dsize = fs->fs_old_dsize; 1540 fs->fs_csaddr = fs->fs_old_csaddr; 1541 fs->fs_sblockloc = sblockloc; 1542 1543 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL); 1544 1545 if (fs->fs_old_postblformat == FS_42POSTBLFMT) { 1546 fs->fs_old_nrpos = 8; 1547 fs->fs_old_npsect = fs->fs_old_nsect; 1548 fs->fs_old_interleave = 1; 1549 fs->fs_old_trackskew = 0; 1550 } 1551 1552 if (fs->fs_old_inodefmt < FS_44INODEFMT) { 1553 fs->fs_maxfilesize = (u_quad_t) 1LL << 39; 1554 fs->fs_qbmask = ~fs->fs_bmask; 1555 fs->fs_qfmask = ~fs->fs_fmask; 1556 } 1557 1558 maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1; 1559 if (fs->fs_maxfilesize > maxfilesize) 1560 fs->fs_maxfilesize = maxfilesize; 1561 1562 /* Compatibility for old filesystems */ 1563 if (fs->fs_avgfilesize <= 0) 1564 fs->fs_avgfilesize = AVFILESIZ; 1565 if (fs->fs_avgfpdir <= 0) 1566 fs->fs_avgfpdir = AFPDIR; 1567 1568 #if 0 1569 if (bigcgs) { 1570 fs->fs_save_cgsize = fs->fs_cgsize; 1571 fs->fs_cgsize = fs->fs_bsize; 1572 } 1573 #endif 1574 } 1575 1576 /* 1577 * Unwinding superblock updates for old filesystems. 1578 * See ffs_oldfscompat_read above for details. 1579 * 1580 * XXX - Parts get retired eventually. 1581 * Unfortunately new bits get added. 1582 */ 1583 static void 1584 ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump) 1585 { 1586 int32_t *extrasave; 1587 1588 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1589 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1590 return; 1591 1592 fs->fs_old_time = fs->fs_time; 1593 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir; 1594 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree; 1595 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree; 1596 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree; 1597 fs->fs_old_flags = fs->fs_flags; 1598 1599 #if 0 1600 if (bigcgs) { 1601 fs->fs_cgsize = fs->fs_save_cgsize; 1602 } 1603 #endif 1604 1605 memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512); 1606 extrasave = ump->um_oldfscompat; 1607 extrasave += 512/sizeof(int32_t); 1608 fs->fs_old_npsect = extrasave[0]; 1609 fs->fs_old_interleave = extrasave[1]; 1610 fs->fs_old_trackskew = extrasave[2]; 1611 1612 } 1613 1614 /* 1615 * unmount vfs operation 1616 */ 1617 int 1618 ffs_unmount(struct mount *mp, int mntflags) 1619 { 1620 struct lwp *l = curlwp; 1621 struct ufsmount *ump = VFSTOUFS(mp); 1622 struct fs *fs = ump->um_fs; 1623 int error, flags; 1624 u_int32_t bsize; 1625 #ifdef WAPBL 1626 extern int doforce; 1627 #endif 1628 1629 if (ump->um_discarddata) { 1630 ffs_discard_finish(ump->um_discarddata, mntflags); 1631 ump->um_discarddata = NULL; 1632 } 1633 1634 flags = 0; 1635 if (mntflags & MNT_FORCE) 1636 flags |= FORCECLOSE; 1637 if ((error = ffs_flushfiles(mp, flags, l)) != 0) 1638 return (error); 1639 error = UFS_WAPBL_BEGIN(mp); 1640 if (error == 0) 1641 if (fs->fs_ronly == 0 && 1642 ffs_cgupdate(ump, MNT_WAIT) == 0 && 1643 fs->fs_clean & FS_WASCLEAN) { 1644 fs->fs_clean = FS_ISCLEAN; 1645 fs->fs_fmod = 0; 1646 (void) ffs_sbupdate(ump, MNT_WAIT); 1647 } 1648 if (error == 0) 1649 UFS_WAPBL_END(mp); 1650 #ifdef WAPBL 1651 KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl)); 1652 if (mp->mnt_wapbl_replay) { 1653 KDASSERT(fs->fs_ronly); 1654 wapbl_replay_stop(mp->mnt_wapbl_replay); 1655 wapbl_replay_free(mp->mnt_wapbl_replay); 1656 mp->mnt_wapbl_replay = 0; 1657 } 1658 error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE)); 1659 if (error) { 1660 return error; 1661 } 1662 #endif /* WAPBL */ 1663 1664 if (ump->um_devvp->v_type != VBAD) 1665 spec_node_setmountedfs(ump->um_devvp, NULL); 1666 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1667 (void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE, 1668 NOCRED); 1669 vput(ump->um_devvp); 1670 1671 bsize = fs->fs_cssize; 1672 if (fs->fs_contigsumsize > 0) 1673 bsize += fs->fs_ncg * sizeof(int32_t); 1674 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs); 1675 kmem_free(fs->fs_csp, bsize); 1676 1677 kmem_free(fs, fs->fs_sbsize); 1678 if (ump->um_oldfscompat != NULL) 1679 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t)); 1680 mutex_destroy(&ump->um_lock); 1681 ffs_snapshot_fini(ump); 1682 kmem_free(ump, sizeof(*ump)); 1683 mp->mnt_data = NULL; 1684 mp->mnt_flag &= ~MNT_LOCAL; 1685 fstrans_unmount(mp); 1686 return (0); 1687 } 1688 1689 /* 1690 * Flush out all the files in a filesystem. 1691 */ 1692 int 1693 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l) 1694 { 1695 extern int doforce; 1696 struct ufsmount *ump; 1697 int error; 1698 1699 if (!doforce) 1700 flags &= ~FORCECLOSE; 1701 ump = VFSTOUFS(mp); 1702 #ifdef QUOTA 1703 if ((error = quota1_umount(mp, flags)) != 0) 1704 return (error); 1705 #endif 1706 #ifdef QUOTA2 1707 if ((error = quota2_umount(mp, flags)) != 0) 1708 return (error); 1709 #endif 1710 #ifdef UFS_EXTATTR 1711 if (ump->um_fstype == UFS1) { 1712 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_STARTED) 1713 ufs_extattr_stop(mp, l); 1714 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_INITIALIZED) 1715 ufs_extattr_uepm_destroy(&ump->um_extattr); 1716 mp->mnt_flag &= ~MNT_EXTATTR; 1717 } 1718 #endif 1719 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0) 1720 return (error); 1721 ffs_snapshot_unmount(mp); 1722 /* 1723 * Flush all the files. 1724 */ 1725 error = vflush(mp, NULLVP, flags); 1726 if (error) 1727 return (error); 1728 /* 1729 * Flush filesystem metadata. 1730 */ 1731 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1732 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0); 1733 VOP_UNLOCK(ump->um_devvp); 1734 if (flags & FORCECLOSE) /* XXXDBJ */ 1735 error = 0; 1736 1737 #ifdef WAPBL 1738 if (error) 1739 return error; 1740 if (mp->mnt_wapbl) { 1741 error = wapbl_flush(mp->mnt_wapbl, 1); 1742 if (flags & FORCECLOSE) 1743 error = 0; 1744 } 1745 #endif 1746 1747 return (error); 1748 } 1749 1750 /* 1751 * Get file system statistics. 1752 */ 1753 int 1754 ffs_statvfs(struct mount *mp, struct statvfs *sbp) 1755 { 1756 struct ufsmount *ump; 1757 struct fs *fs; 1758 1759 ump = VFSTOUFS(mp); 1760 fs = ump->um_fs; 1761 mutex_enter(&ump->um_lock); 1762 sbp->f_bsize = fs->fs_bsize; 1763 sbp->f_frsize = fs->fs_fsize; 1764 sbp->f_iosize = fs->fs_bsize; 1765 sbp->f_blocks = fs->fs_dsize; 1766 sbp->f_bfree = ffs_blkstofrags(fs, fs->fs_cstotal.cs_nbfree) + 1767 fs->fs_cstotal.cs_nffree + FFS_DBTOFSB(fs, fs->fs_pendingblocks); 1768 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t) 1769 fs->fs_minfree) / (u_int64_t) 100; 1770 if (sbp->f_bfree > sbp->f_bresvd) 1771 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd; 1772 else 1773 sbp->f_bavail = 0; 1774 sbp->f_files = fs->fs_ncg * fs->fs_ipg - UFS_ROOTINO; 1775 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes; 1776 sbp->f_favail = sbp->f_ffree; 1777 sbp->f_fresvd = 0; 1778 mutex_exit(&ump->um_lock); 1779 copy_statvfs_info(sbp, mp); 1780 1781 return (0); 1782 } 1783 1784 struct ffs_sync_ctx { 1785 int waitfor; 1786 bool is_suspending; 1787 }; 1788 1789 static bool 1790 ffs_sync_selector(void *cl, struct vnode *vp) 1791 { 1792 struct ffs_sync_ctx *c = cl; 1793 struct inode *ip; 1794 1795 ip = VTOI(vp); 1796 /* 1797 * Skip the vnode/inode if inaccessible. 1798 */ 1799 if (ip == NULL || vp->v_type == VNON) 1800 return false; 1801 1802 /* 1803 * We deliberately update inode times here. This will 1804 * prevent a massive queue of updates accumulating, only 1805 * to be handled by a call to unmount. 1806 * 1807 * XXX It would be better to have the syncer trickle these 1808 * out. Adjustment needed to allow registering vnodes for 1809 * sync when the vnode is clean, but the inode dirty. Or 1810 * have ufs itself trickle out inode updates. 1811 * 1812 * If doing a lazy sync, we don't care about metadata or 1813 * data updates, because they are handled by each vnode's 1814 * synclist entry. In this case we are only interested in 1815 * writing back modified inodes. 1816 */ 1817 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE | 1818 IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 && 1819 (c->waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) && 1820 UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) 1821 return false; 1822 1823 if (vp->v_type == VBLK && c->is_suspending) 1824 return false; 1825 1826 return true; 1827 } 1828 1829 /* 1830 * Go through the disk queues to initiate sandbagged IO; 1831 * go through the inodes to write those that have been modified; 1832 * initiate the writing of the super block if it has been modified. 1833 * 1834 * Note: we are always called with the filesystem marked `MPBUSY'. 1835 */ 1836 int 1837 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred) 1838 { 1839 struct vnode *vp; 1840 struct ufsmount *ump = VFSTOUFS(mp); 1841 struct fs *fs; 1842 struct vnode_iterator *marker; 1843 int error, allerror = 0; 1844 bool is_suspending; 1845 struct ffs_sync_ctx ctx; 1846 1847 fs = ump->um_fs; 1848 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */ 1849 printf("fs = %s\n", fs->fs_fsmnt); 1850 panic("update: rofs mod"); 1851 } 1852 1853 fstrans_start(mp, FSTRANS_SHARED); 1854 is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING); 1855 /* 1856 * Write back each (modified) inode. 1857 */ 1858 vfs_vnode_iterator_init(mp, &marker); 1859 1860 ctx.waitfor = waitfor; 1861 ctx.is_suspending = is_suspending; 1862 while ((vp = vfs_vnode_iterator_next(marker, ffs_sync_selector, &ctx))) 1863 { 1864 error = vn_lock(vp, LK_EXCLUSIVE); 1865 if (error) { 1866 vrele(vp); 1867 continue; 1868 } 1869 if (waitfor == MNT_LAZY) { 1870 error = UFS_WAPBL_BEGIN(vp->v_mount); 1871 if (!error) { 1872 error = ffs_update(vp, NULL, NULL, 1873 UPDATE_CLOSE); 1874 UFS_WAPBL_END(vp->v_mount); 1875 } 1876 } else { 1877 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG | 1878 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0); 1879 } 1880 if (error) 1881 allerror = error; 1882 vput(vp); 1883 } 1884 vfs_vnode_iterator_destroy(marker); 1885 1886 /* 1887 * Force stale file system control information to be flushed. 1888 */ 1889 if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 || 1890 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) { 1891 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY); 1892 if ((error = VOP_FSYNC(ump->um_devvp, cred, 1893 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG, 1894 0, 0)) != 0) 1895 allerror = error; 1896 VOP_UNLOCK(ump->um_devvp); 1897 } 1898 #if defined(QUOTA) || defined(QUOTA2) 1899 qsync(mp); 1900 #endif 1901 /* 1902 * Write back modified superblock. 1903 */ 1904 if (fs->fs_fmod != 0) { 1905 fs->fs_fmod = 0; 1906 fs->fs_time = time_second; 1907 error = UFS_WAPBL_BEGIN(mp); 1908 if (error) 1909 allerror = error; 1910 else { 1911 if ((error = ffs_cgupdate(ump, waitfor))) 1912 allerror = error; 1913 UFS_WAPBL_END(mp); 1914 } 1915 } 1916 1917 #ifdef WAPBL 1918 if (mp->mnt_wapbl) { 1919 error = wapbl_flush(mp->mnt_wapbl, 0); 1920 if (error) 1921 allerror = error; 1922 } 1923 #endif 1924 1925 fstrans_done(mp); 1926 return (allerror); 1927 } 1928 1929 /* 1930 * Read an inode from disk and initialize this vnode / inode pair. 1931 * Caller assures no other thread will try to load this inode. 1932 */ 1933 int 1934 ffs_loadvnode(struct mount *mp, struct vnode *vp, 1935 const void *key, size_t key_len, const void **new_key) 1936 { 1937 ino_t ino; 1938 struct fs *fs; 1939 struct inode *ip; 1940 struct ufsmount *ump; 1941 struct buf *bp; 1942 dev_t dev; 1943 int error; 1944 1945 KASSERT(key_len == sizeof(ino)); 1946 memcpy(&ino, key, key_len); 1947 ump = VFSTOUFS(mp); 1948 dev = ump->um_dev; 1949 fs = ump->um_fs; 1950 1951 /* Read in the disk contents for the inode. */ 1952 error = bread(ump->um_devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ino)), 1953 (int)fs->fs_bsize, NOCRED, 0, &bp); 1954 if (error) 1955 return error; 1956 1957 /* Allocate and initialize inode. */ 1958 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK); 1959 memset(ip, 0, sizeof(struct inode)); 1960 vp->v_tag = VT_UFS; 1961 vp->v_op = ffs_vnodeop_p; 1962 vp->v_vflag |= VV_LOCKSWORK; 1963 vp->v_data = ip; 1964 ip->i_vnode = vp; 1965 ip->i_ump = ump; 1966 ip->i_fs = fs; 1967 ip->i_dev = dev; 1968 ip->i_number = ino; 1969 #if defined(QUOTA) || defined(QUOTA2) 1970 ufsquota_init(ip); 1971 #endif 1972 1973 /* Initialize genfs node. */ 1974 genfs_node_init(vp, &ffs_genfsops); 1975 1976 if (ip->i_ump->um_fstype == UFS1) 1977 ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache, 1978 PR_WAITOK); 1979 else 1980 ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache, 1981 PR_WAITOK); 1982 ffs_load_inode(bp, ip, fs, ino); 1983 brelse(bp, 0); 1984 1985 /* Initialize the vnode from the inode. */ 1986 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp); 1987 1988 /* Finish inode initialization. */ 1989 ip->i_devvp = ump->um_devvp; 1990 vref(ip->i_devvp); 1991 1992 /* 1993 * Ensure that uid and gid are correct. This is a temporary 1994 * fix until fsck has been changed to do the update. 1995 */ 1996 1997 if (fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */ 1998 ip->i_uid = ip->i_ffs1_ouid; /* XXX */ 1999 ip->i_gid = ip->i_ffs1_ogid; /* XXX */ 2000 } /* XXX */ 2001 uvm_vnp_setsize(vp, ip->i_size); 2002 *new_key = &ip->i_number; 2003 return 0; 2004 } 2005 2006 /* 2007 * File handle to vnode 2008 * 2009 * Have to be really careful about stale file handles: 2010 * - check that the inode number is valid 2011 * - call ffs_vget() to get the locked inode 2012 * - check for an unallocated inode (i_mode == 0) 2013 * - check that the given client host has export rights and return 2014 * those rights via. exflagsp and credanonp 2015 */ 2016 int 2017 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp) 2018 { 2019 struct ufid ufh; 2020 struct fs *fs; 2021 2022 if (fhp->fid_len != sizeof(struct ufid)) 2023 return EINVAL; 2024 2025 memcpy(&ufh, fhp, sizeof(ufh)); 2026 fs = VFSTOUFS(mp)->um_fs; 2027 if (ufh.ufid_ino < UFS_ROOTINO || 2028 ufh.ufid_ino >= fs->fs_ncg * fs->fs_ipg) 2029 return (ESTALE); 2030 return (ufs_fhtovp(mp, &ufh, vpp)); 2031 } 2032 2033 /* 2034 * Vnode pointer to File handle 2035 */ 2036 /* ARGSUSED */ 2037 int 2038 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size) 2039 { 2040 struct inode *ip; 2041 struct ufid ufh; 2042 2043 if (*fh_size < sizeof(struct ufid)) { 2044 *fh_size = sizeof(struct ufid); 2045 return E2BIG; 2046 } 2047 ip = VTOI(vp); 2048 *fh_size = sizeof(struct ufid); 2049 memset(&ufh, 0, sizeof(ufh)); 2050 ufh.ufid_len = sizeof(struct ufid); 2051 ufh.ufid_ino = ip->i_number; 2052 ufh.ufid_gen = ip->i_gen; 2053 memcpy(fhp, &ufh, sizeof(ufh)); 2054 return (0); 2055 } 2056 2057 void 2058 ffs_init(void) 2059 { 2060 if (ffs_initcount++ > 0) 2061 return; 2062 2063 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0, 2064 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL); 2065 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0, 2066 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL); 2067 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0, 2068 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL); 2069 ufs_init(); 2070 } 2071 2072 void 2073 ffs_reinit(void) 2074 { 2075 ufs_reinit(); 2076 } 2077 2078 void 2079 ffs_done(void) 2080 { 2081 if (--ffs_initcount > 0) 2082 return; 2083 2084 ufs_done(); 2085 pool_cache_destroy(ffs_dinode2_cache); 2086 pool_cache_destroy(ffs_dinode1_cache); 2087 pool_cache_destroy(ffs_inode_cache); 2088 } 2089 2090 /* 2091 * Write a superblock and associated information back to disk. 2092 */ 2093 int 2094 ffs_sbupdate(struct ufsmount *mp, int waitfor) 2095 { 2096 struct fs *fs = mp->um_fs; 2097 struct buf *bp; 2098 int error = 0; 2099 u_int32_t saveflag; 2100 2101 error = ffs_getblk(mp->um_devvp, 2102 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK, 2103 fs->fs_sbsize, false, &bp); 2104 if (error) 2105 return error; 2106 saveflag = fs->fs_flags & FS_INTERNAL; 2107 fs->fs_flags &= ~FS_INTERNAL; 2108 2109 memcpy(bp->b_data, fs, fs->fs_sbsize); 2110 2111 ffs_oldfscompat_write((struct fs *)bp->b_data, mp); 2112 #ifdef FFS_EI 2113 if (mp->um_flags & UFS_NEEDSWAP) 2114 ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data); 2115 #endif 2116 fs->fs_flags |= saveflag; 2117 2118 if (waitfor == MNT_WAIT) 2119 error = bwrite(bp); 2120 else 2121 bawrite(bp); 2122 return (error); 2123 } 2124 2125 int 2126 ffs_cgupdate(struct ufsmount *mp, int waitfor) 2127 { 2128 struct fs *fs = mp->um_fs; 2129 struct buf *bp; 2130 int blks; 2131 void *space; 2132 int i, size, error = 0, allerror = 0; 2133 2134 allerror = ffs_sbupdate(mp, waitfor); 2135 blks = howmany(fs->fs_cssize, fs->fs_fsize); 2136 space = fs->fs_csp; 2137 for (i = 0; i < blks; i += fs->fs_frag) { 2138 size = fs->fs_bsize; 2139 if (i + fs->fs_frag > blks) 2140 size = (blks - i) * fs->fs_fsize; 2141 error = ffs_getblk(mp->um_devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), 2142 FFS_NOBLK, size, false, &bp); 2143 if (error) 2144 break; 2145 #ifdef FFS_EI 2146 if (mp->um_flags & UFS_NEEDSWAP) 2147 ffs_csum_swap((struct csum*)space, 2148 (struct csum*)bp->b_data, size); 2149 else 2150 #endif 2151 memcpy(bp->b_data, space, (u_int)size); 2152 space = (char *)space + size; 2153 if (waitfor == MNT_WAIT) 2154 error = bwrite(bp); 2155 else 2156 bawrite(bp); 2157 } 2158 if (!allerror && error) 2159 allerror = error; 2160 return (allerror); 2161 } 2162 2163 int 2164 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp, 2165 int attrnamespace, const char *attrname) 2166 { 2167 #ifdef UFS_EXTATTR 2168 /* 2169 * File-backed extended attributes are only supported on UFS1. 2170 * UFS2 has native extended attributes. 2171 */ 2172 if (VFSTOUFS(mp)->um_fstype == UFS1) 2173 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname)); 2174 #endif 2175 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname)); 2176 } 2177 2178 int 2179 ffs_suspendctl(struct mount *mp, int cmd) 2180 { 2181 int error; 2182 struct lwp *l = curlwp; 2183 2184 switch (cmd) { 2185 case SUSPEND_SUSPEND: 2186 if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0) 2187 return error; 2188 error = ffs_sync(mp, MNT_WAIT, l->l_proc->p_cred); 2189 if (error == 0) 2190 error = fstrans_setstate(mp, FSTRANS_SUSPENDED); 2191 #ifdef WAPBL 2192 if (error == 0 && mp->mnt_wapbl) 2193 error = wapbl_flush(mp->mnt_wapbl, 1); 2194 #endif 2195 if (error != 0) { 2196 (void) fstrans_setstate(mp, FSTRANS_NORMAL); 2197 return error; 2198 } 2199 return 0; 2200 2201 case SUSPEND_RESUME: 2202 return fstrans_setstate(mp, FSTRANS_NORMAL); 2203 2204 default: 2205 return EINVAL; 2206 } 2207 } 2208 2209 /* 2210 * Synch vnode for a mounted file system. 2211 */ 2212 static int 2213 ffs_vfs_fsync(vnode_t *vp, int flags) 2214 { 2215 int error, i, pflags; 2216 #ifdef WAPBL 2217 struct mount *mp; 2218 #endif 2219 2220 KASSERT(vp->v_type == VBLK); 2221 KASSERT(spec_node_getmountedfs(vp) != NULL); 2222 2223 /* 2224 * Flush all dirty data associated with the vnode. 2225 */ 2226 pflags = PGO_ALLPAGES | PGO_CLEANIT; 2227 if ((flags & FSYNC_WAIT) != 0) 2228 pflags |= PGO_SYNCIO; 2229 mutex_enter(vp->v_interlock); 2230 error = VOP_PUTPAGES(vp, 0, 0, pflags); 2231 if (error) 2232 return error; 2233 2234 #ifdef WAPBL 2235 mp = spec_node_getmountedfs(vp); 2236 if (mp && mp->mnt_wapbl) { 2237 /* 2238 * Don't bother writing out metadata if the syncer is 2239 * making the request. We will let the sync vnode 2240 * write it out in a single burst through a call to 2241 * VFS_SYNC(). 2242 */ 2243 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0) 2244 return 0; 2245 2246 /* 2247 * Don't flush the log if the vnode being flushed 2248 * contains no dirty buffers that could be in the log. 2249 */ 2250 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) { 2251 error = wapbl_flush(mp->mnt_wapbl, 0); 2252 if (error) 2253 return error; 2254 } 2255 2256 if ((flags & FSYNC_WAIT) != 0) { 2257 mutex_enter(vp->v_interlock); 2258 while (vp->v_numoutput) 2259 cv_wait(&vp->v_cv, vp->v_interlock); 2260 mutex_exit(vp->v_interlock); 2261 } 2262 2263 return 0; 2264 } 2265 #endif /* WAPBL */ 2266 2267 error = vflushbuf(vp, flags); 2268 if (error == 0 && (flags & FSYNC_CACHE) != 0) { 2269 i = 1; 2270 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE, 2271 kauth_cred_get()); 2272 } 2273 2274 return error; 2275 } 2276
Indexes created Wed Oct 01 19:09:53 GMT 2025