PageRenderTime 95ms CodeModel.GetById 3ms app.highlight 79ms RepoModel.GetById 1ms app.codeStats 0ms

/sys/ufs/ffs/ffs_vfsops.c

https://bitbucket.org/gthummalapalle/minix
C | 2144 lines | 1567 code | 184 blank | 393 comment | 401 complexity | 83ac6bcbf6505dbe547ad438974d1365 MD5 | raw file

Large files files are truncated, but you can click here to view the full file

   1/*	$NetBSD: ffs_vfsops.c,v 1.271 2011/11/14 18:35:14 hannken 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 1.271 2011/11/14 18:35:14 hannken 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/malloc.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
 112MODULE(MODULE_CLASS_VFS, ffs, NULL);
 113
 114static int	ffs_vfs_fsync(vnode_t *, int);
 115
 116static struct sysctllog *ffs_sysctl_log;
 117
 118/* how many times ffs_init() was called */
 119int ffs_initcount = 0;
 120
 121extern const struct vnodeopv_desc ffs_vnodeop_opv_desc;
 122extern const struct vnodeopv_desc ffs_specop_opv_desc;
 123extern const struct vnodeopv_desc ffs_fifoop_opv_desc;
 124
 125const struct vnodeopv_desc * const ffs_vnodeopv_descs[] = {
 126	&ffs_vnodeop_opv_desc,
 127	&ffs_specop_opv_desc,
 128	&ffs_fifoop_opv_desc,
 129	NULL,
 130};
 131
 132struct vfsops ffs_vfsops = {
 133	MOUNT_FFS,
 134	sizeof (struct ufs_args),
 135	ffs_mount,
 136	ufs_start,
 137	ffs_unmount,
 138	ufs_root,
 139	ufs_quotactl,
 140	ffs_statvfs,
 141	ffs_sync,
 142	ffs_vget,
 143	ffs_fhtovp,
 144	ffs_vptofh,
 145	ffs_init,
 146	ffs_reinit,
 147	ffs_done,
 148	ffs_mountroot,
 149	ffs_snapshot,
 150	ffs_extattrctl,
 151	ffs_suspendctl,
 152	genfs_renamelock_enter,
 153	genfs_renamelock_exit,
 154	ffs_vfs_fsync,
 155	ffs_vnodeopv_descs,
 156	0,
 157	{ NULL, NULL },
 158};
 159
 160static const struct genfs_ops ffs_genfsops = {
 161	.gop_size = ffs_gop_size,
 162	.gop_alloc = ufs_gop_alloc,
 163	.gop_write = genfs_gop_write,
 164	.gop_markupdate = ufs_gop_markupdate,
 165};
 166
 167static const struct ufs_ops ffs_ufsops = {
 168	.uo_itimes = ffs_itimes,
 169	.uo_update = ffs_update,
 170	.uo_truncate = ffs_truncate,
 171	.uo_valloc = ffs_valloc,
 172	.uo_vfree = ffs_vfree,
 173	.uo_balloc = ffs_balloc,
 174	.uo_unmark_vnode = (void (*)(vnode_t *))nullop,
 175};
 176
 177static int
 178ffs_modcmd(modcmd_t cmd, void *arg)
 179{
 180	int error;
 181
 182#if 0
 183	extern int doasyncfree;
 184#endif
 185#ifdef UFS_EXTATTR
 186	extern int ufs_extattr_autocreate;
 187#endif
 188	extern int ffs_log_changeopt;
 189
 190	switch (cmd) {
 191	case MODULE_CMD_INIT:
 192		error = vfs_attach(&ffs_vfsops);
 193		if (error != 0)
 194			break;
 195
 196		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
 197			       CTLFLAG_PERMANENT,
 198			       CTLTYPE_NODE, "vfs", NULL,
 199			       NULL, 0, NULL, 0,
 200			       CTL_VFS, CTL_EOL);
 201		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
 202			       CTLFLAG_PERMANENT,
 203			       CTLTYPE_NODE, "ffs",
 204			       SYSCTL_DESCR("Berkeley Fast File System"),
 205			       NULL, 0, NULL, 0,
 206			       CTL_VFS, 1, CTL_EOL);
 207		/*
 208		 * @@@ should we even bother with these first three?
 209		 */
 210		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
 211			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
 212			       CTLTYPE_INT, "doclusterread", NULL,
 213			       sysctl_notavail, 0, NULL, 0,
 214			       CTL_VFS, 1, FFS_CLUSTERREAD, CTL_EOL);
 215		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
 216			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
 217			       CTLTYPE_INT, "doclusterwrite", NULL,
 218			       sysctl_notavail, 0, NULL, 0,
 219			       CTL_VFS, 1, FFS_CLUSTERWRITE, CTL_EOL);
 220		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
 221			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
 222			       CTLTYPE_INT, "doreallocblks", NULL,
 223			       sysctl_notavail, 0, NULL, 0,
 224			       CTL_VFS, 1, FFS_REALLOCBLKS, CTL_EOL);
 225#if 0
 226		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
 227			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
 228			       CTLTYPE_INT, "doasyncfree",
 229			       SYSCTL_DESCR("Release dirty blocks asynchronously"),
 230			       NULL, 0, &doasyncfree, 0,
 231			       CTL_VFS, 1, FFS_ASYNCFREE, CTL_EOL);
 232#endif
 233		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
 234			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
 235			       CTLTYPE_INT, "log_changeopt",
 236			       SYSCTL_DESCR("Log changes in optimization strategy"),
 237			       NULL, 0, &ffs_log_changeopt, 0,
 238			       CTL_VFS, 1, FFS_LOG_CHANGEOPT, CTL_EOL);
 239#ifdef UFS_EXTATTR
 240		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
 241			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
 242			       CTLTYPE_INT, "extattr_autocreate",
 243			       SYSCTL_DESCR("Size of attribute for "
 244					    "backing file autocreation"),
 245			       NULL, 0, &ufs_extattr_autocreate, 0,
 246			       CTL_VFS, 1, FFS_EXTATTR_AUTOCREATE, CTL_EOL);
 247		
 248#endif /* UFS_EXTATTR */
 249
 250		break;
 251	case MODULE_CMD_FINI:
 252		error = vfs_detach(&ffs_vfsops);
 253		if (error != 0)
 254			break;
 255		sysctl_teardown(&ffs_sysctl_log);
 256		break;
 257	default:
 258		error = ENOTTY;
 259		break;
 260	}
 261
 262	return (error);
 263}
 264
 265pool_cache_t ffs_inode_cache;
 266pool_cache_t ffs_dinode1_cache;
 267pool_cache_t ffs_dinode2_cache;
 268
 269static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t);
 270static void ffs_oldfscompat_write(struct fs *, struct ufsmount *);
 271
 272/*
 273 * Called by main() when ffs is going to be mounted as root.
 274 */
 275
 276int
 277ffs_mountroot(void)
 278{
 279	struct fs *fs;
 280	struct mount *mp;
 281	struct lwp *l = curlwp;			/* XXX */
 282	struct ufsmount *ump;
 283	int error;
 284
 285	if (device_class(root_device) != DV_DISK)
 286		return (ENODEV);
 287
 288	if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp))) {
 289		vrele(rootvp);
 290		return (error);
 291	}
 292
 293	/*
 294	 * We always need to be able to mount the root file system.
 295	 */
 296	mp->mnt_flag |= MNT_FORCE;
 297	if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
 298		vfs_unbusy(mp, false, NULL);
 299		vfs_destroy(mp);
 300		return (error);
 301	}
 302	mp->mnt_flag &= ~MNT_FORCE;
 303	mutex_enter(&mountlist_lock);
 304	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
 305	mutex_exit(&mountlist_lock);
 306	ump = VFSTOUFS(mp);
 307	fs = ump->um_fs;
 308	memset(fs->fs_fsmnt, 0, sizeof(fs->fs_fsmnt));
 309	(void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
 310	(void)ffs_statvfs(mp, &mp->mnt_stat);
 311	vfs_unbusy(mp, false, NULL);
 312	setrootfstime((time_t)fs->fs_time);
 313	return (0);
 314}
 315
 316/*
 317 * VFS Operations.
 318 *
 319 * mount system call
 320 */
 321int
 322ffs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
 323{
 324	struct lwp *l = curlwp;
 325	struct vnode *devvp = NULL;
 326	struct ufs_args *args = data;
 327	struct ufsmount *ump = NULL;
 328	struct fs *fs;
 329	int error = 0, flags, update;
 330	mode_t accessmode;
 331
 332	if (*data_len < sizeof *args)
 333		return EINVAL;
 334
 335	if (mp->mnt_flag & MNT_GETARGS) {
 336		ump = VFSTOUFS(mp);
 337		if (ump == NULL)
 338			return EIO;
 339		args->fspec = NULL;
 340		*data_len = sizeof *args;
 341		return 0;
 342	}
 343
 344	update = mp->mnt_flag & MNT_UPDATE;
 345
 346	/* Check arguments */
 347	if (args->fspec != NULL) {
 348		/*
 349		 * Look up the name and verify that it's sane.
 350		 */
 351		error = namei_simple_user(args->fspec,
 352					NSM_FOLLOW_NOEMULROOT, &devvp);
 353		if (error != 0)
 354			return (error);
 355
 356		if (!update) {
 357			/*
 358			 * Be sure this is a valid block device
 359			 */
 360			if (devvp->v_type != VBLK)
 361				error = ENOTBLK;
 362			else if (bdevsw_lookup(devvp->v_rdev) == NULL)
 363				error = ENXIO;
 364		} else {
 365			/*
 366			 * Be sure we're still naming the same device
 367			 * used for our initial mount
 368			 */
 369			ump = VFSTOUFS(mp);
 370			if (devvp != ump->um_devvp) {
 371				if (devvp->v_rdev != ump->um_devvp->v_rdev)
 372					error = EINVAL;
 373				else {
 374					vrele(devvp);
 375					devvp = ump->um_devvp;
 376					vref(devvp);
 377				}
 378			}
 379		}
 380	} else {
 381		if (!update) {
 382			/* New mounts must have a filename for the device */
 383			return (EINVAL);
 384		} else {
 385			/* Use the extant mount */
 386			ump = VFSTOUFS(mp);
 387			devvp = ump->um_devvp;
 388			vref(devvp);
 389		}
 390	}
 391
 392	/*
 393	 * If mount by non-root, then verify that user has necessary
 394	 * permissions on the device.
 395	 *
 396	 * Permission to update a mount is checked higher, so here we presume
 397	 * updating the mount is okay (for example, as far as securelevel goes)
 398	 * which leaves us with the normal check.
 399	 */
 400	if (error == 0) {
 401		accessmode = VREAD;
 402		if (update ?
 403		    (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
 404		    (mp->mnt_flag & MNT_RDONLY) == 0)
 405			accessmode |= VWRITE;
 406		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
 407		error = genfs_can_mount(devvp, accessmode, l->l_cred);
 408		VOP_UNLOCK(devvp);
 409	}
 410
 411	if (error) {
 412		vrele(devvp);
 413		return (error);
 414	}
 415
 416#ifdef WAPBL
 417	/* WAPBL can only be enabled on a r/w mount. */
 418	if ((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) {
 419		mp->mnt_flag &= ~MNT_LOG;
 420	}
 421#else /* !WAPBL */
 422	mp->mnt_flag &= ~MNT_LOG;
 423#endif /* !WAPBL */
 424
 425	if (!update) {
 426		int xflags;
 427
 428		if (mp->mnt_flag & MNT_RDONLY)
 429			xflags = FREAD;
 430		else
 431			xflags = FREAD | FWRITE;
 432		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
 433		error = VOP_OPEN(devvp, xflags, FSCRED);
 434		VOP_UNLOCK(devvp);
 435		if (error)
 436			goto fail;
 437		error = ffs_mountfs(devvp, mp, l);
 438		if (error) {
 439			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
 440			(void)VOP_CLOSE(devvp, xflags, NOCRED);
 441			VOP_UNLOCK(devvp);
 442			goto fail;
 443		}
 444
 445		ump = VFSTOUFS(mp);
 446		fs = ump->um_fs;
 447	} else {
 448		/*
 449		 * Update the mount.
 450		 */
 451
 452		/*
 453		 * The initial mount got a reference on this
 454		 * device, so drop the one obtained via
 455		 * namei(), above.
 456		 */
 457		vrele(devvp);
 458
 459		ump = VFSTOUFS(mp);
 460		fs = ump->um_fs;
 461		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
 462			/*
 463			 * Changing from r/w to r/o
 464			 */
 465			flags = WRITECLOSE;
 466			if (mp->mnt_flag & MNT_FORCE)
 467				flags |= FORCECLOSE;
 468			error = ffs_flushfiles(mp, flags, l);
 469			if (error == 0)
 470				error = UFS_WAPBL_BEGIN(mp);
 471			if (error == 0 &&
 472			    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
 473			    fs->fs_clean & FS_WASCLEAN) {
 474				if (mp->mnt_flag & MNT_SOFTDEP)
 475					fs->fs_flags &= ~FS_DOSOFTDEP;
 476				fs->fs_clean = FS_ISCLEAN;
 477				(void) ffs_sbupdate(ump, MNT_WAIT);
 478			}
 479			if (error == 0)
 480				UFS_WAPBL_END(mp);
 481			if (error)
 482				return (error);
 483		}
 484
 485#ifdef WAPBL
 486		if ((mp->mnt_flag & MNT_LOG) == 0) {
 487			error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE);
 488			if (error)
 489				return error;
 490		}
 491#endif /* WAPBL */
 492
 493		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
 494			/*
 495			 * Finish change from r/w to r/o
 496			 */
 497			fs->fs_ronly = 1;
 498			fs->fs_fmod = 0;
 499		}
 500
 501		if (mp->mnt_flag & MNT_RELOAD) {
 502			error = ffs_reload(mp, l->l_cred, l);
 503			if (error)
 504				return (error);
 505		}
 506
 507		if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
 508			/*
 509			 * Changing from read-only to read/write
 510			 */
 511#ifndef QUOTA2
 512			if (fs->fs_flags & FS_DOQUOTA2) {
 513				ump->um_flags |= UFS_QUOTA2;
 514				uprintf("%s: options QUOTA2 not enabled%s\n",
 515				    mp->mnt_stat.f_mntonname,
 516				    (mp->mnt_flag & MNT_FORCE) ? "" :
 517				    ", not mounting");
 518				return EINVAL;
 519			}
 520#endif
 521			fs->fs_ronly = 0;
 522			fs->fs_clean <<= 1;
 523			fs->fs_fmod = 1;
 524#ifdef WAPBL
 525			if (fs->fs_flags & FS_DOWAPBL) {
 526				printf("%s: replaying log to disk\n",
 527				    fs->fs_fsmnt);
 528				KDASSERT(mp->mnt_wapbl_replay);
 529				error = wapbl_replay_write(mp->mnt_wapbl_replay,
 530							   devvp);
 531				if (error) {
 532					return error;
 533				}
 534				wapbl_replay_stop(mp->mnt_wapbl_replay);
 535				fs->fs_clean = FS_WASCLEAN;
 536			}
 537#endif /* WAPBL */
 538			if (fs->fs_snapinum[0] != 0)
 539				ffs_snapshot_mount(mp);
 540		}
 541
 542#ifdef WAPBL
 543		error = ffs_wapbl_start(mp);
 544		if (error)
 545			return error;
 546#endif /* WAPBL */
 547
 548#ifdef QUOTA2
 549		if (!fs->fs_ronly) {
 550			error = ffs_quota2_mount(mp);
 551			if (error) {
 552				return error;
 553			}
 554		}
 555#endif
 556		if (args->fspec == NULL)
 557			return 0;
 558	}
 559
 560	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
 561	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
 562	if (error == 0)
 563		(void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
 564		    sizeof(fs->fs_fsmnt));
 565	fs->fs_flags &= ~FS_DOSOFTDEP;
 566	if (fs->fs_fmod != 0) {	/* XXX */
 567		int err;
 568
 569		fs->fs_fmod = 0;
 570		if (fs->fs_clean & FS_WASCLEAN)
 571			fs->fs_time = time_second;
 572		else {
 573			printf("%s: file system not clean (fs_clean=%#x); "
 574			    "please fsck(8)\n", mp->mnt_stat.f_mntfromname,
 575			    fs->fs_clean);
 576			printf("%s: lost blocks %" PRId64 " files %d\n",
 577			    mp->mnt_stat.f_mntfromname, fs->fs_pendingblocks,
 578			    fs->fs_pendinginodes);
 579		}
 580		err = UFS_WAPBL_BEGIN(mp);
 581		if (err == 0) {
 582			(void) ffs_cgupdate(ump, MNT_WAIT);
 583			UFS_WAPBL_END(mp);
 584		}
 585	}
 586	if ((mp->mnt_flag & MNT_SOFTDEP) != 0) {
 587		printf("%s: `-o softdep' is no longer supported, "
 588		    "consider `-o log'\n", mp->mnt_stat.f_mntfromname);
 589		mp->mnt_flag &= ~MNT_SOFTDEP;
 590	}
 591
 592	return (error);
 593
 594fail:
 595	vrele(devvp);
 596	return (error);
 597}
 598
 599/*
 600 * Reload all incore data for a filesystem (used after running fsck on
 601 * the root filesystem and finding things to fix). The filesystem must
 602 * be mounted read-only.
 603 *
 604 * Things to do to update the mount:
 605 *	1) invalidate all cached meta-data.
 606 *	2) re-read superblock from disk.
 607 *	3) re-read summary information from disk.
 608 *	4) invalidate all inactive vnodes.
 609 *	5) invalidate all cached file data.
 610 *	6) re-read inode data for all active vnodes.
 611 */
 612int
 613ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
 614{
 615	struct vnode *vp, *mvp, *devvp;
 616	struct inode *ip;
 617	void *space;
 618	struct buf *bp;
 619	struct fs *fs, *newfs;
 620	struct dkwedge_info dkw;
 621	int i, bsize, blks, error;
 622	int32_t *lp;
 623	struct ufsmount *ump;
 624	daddr_t sblockloc;
 625
 626	if ((mp->mnt_flag & MNT_RDONLY) == 0)
 627		return (EINVAL);
 628
 629	ump = VFSTOUFS(mp);
 630	/*
 631	 * Step 1: invalidate all cached meta-data.
 632	 */
 633	devvp = ump->um_devvp;
 634	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
 635	error = vinvalbuf(devvp, 0, cred, l, 0, 0);
 636	VOP_UNLOCK(devvp);
 637	if (error)
 638		panic("ffs_reload: dirty1");
 639	/*
 640	 * Step 2: re-read superblock from disk.
 641	 */
 642	fs = ump->um_fs;
 643
 644	/* XXX we don't handle possibility that superblock moved. */
 645	error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs->fs_sbsize,
 646		      NOCRED, 0, &bp);
 647	if (error) {
 648		brelse(bp, 0);
 649		return (error);
 650	}
 651	newfs = malloc(fs->fs_sbsize, M_UFSMNT, M_WAITOK);
 652	memcpy(newfs, bp->b_data, fs->fs_sbsize);
 653#ifdef FFS_EI
 654	if (ump->um_flags & UFS_NEEDSWAP) {
 655		ffs_sb_swap((struct fs*)bp->b_data, newfs);
 656		fs->fs_flags |= FS_SWAPPED;
 657	} else
 658#endif
 659		fs->fs_flags &= ~FS_SWAPPED;
 660	if ((newfs->fs_magic != FS_UFS1_MAGIC &&
 661	     newfs->fs_magic != FS_UFS2_MAGIC)||
 662	     newfs->fs_bsize > MAXBSIZE ||
 663	     newfs->fs_bsize < sizeof(struct fs)) {
 664		brelse(bp, 0);
 665		free(newfs, M_UFSMNT);
 666		return (EIO);		/* XXX needs translation */
 667	}
 668	/* Store off old fs_sblockloc for fs_oldfscompat_read. */
 669	sblockloc = fs->fs_sblockloc;
 670	/*
 671	 * Copy pointer fields back into superblock before copying in	XXX
 672	 * new superblock. These should really be in the ufsmount.	XXX
 673	 * Note that important parameters (eg fs_ncg) are unchanged.
 674	 */
 675	newfs->fs_csp = fs->fs_csp;
 676	newfs->fs_maxcluster = fs->fs_maxcluster;
 677	newfs->fs_contigdirs = fs->fs_contigdirs;
 678	newfs->fs_ronly = fs->fs_ronly;
 679	newfs->fs_active = fs->fs_active;
 680	memcpy(fs, newfs, (u_int)fs->fs_sbsize);
 681	brelse(bp, 0);
 682	free(newfs, M_UFSMNT);
 683
 684	/* Recheck for apple UFS filesystem */
 685	ump->um_flags &= ~UFS_ISAPPLEUFS;
 686	/* First check to see if this is tagged as an Apple UFS filesystem
 687	 * in the disklabel
 688	 */
 689	if (getdiskinfo(devvp, &dkw) == 0 &&
 690	    strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0)
 691		ump->um_flags |= UFS_ISAPPLEUFS;
 692#ifdef APPLE_UFS
 693	else {
 694		/* Manually look for an apple ufs label, and if a valid one
 695		 * is found, then treat it like an Apple UFS filesystem anyway
 696		 *
 697		 * EINVAL is most probably a blocksize or alignment problem,
 698		 * it is unlikely that this is an Apple UFS filesystem then.
 699		 */
 700		error = bread(devvp, (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE),
 701			APPLEUFS_LABEL_SIZE, cred, 0, &bp);
 702		if (error && error != EINVAL) {
 703			brelse(bp, 0);
 704			return (error);
 705		}
 706		if (error == 0) {
 707			error = ffs_appleufs_validate(fs->fs_fsmnt,
 708				(struct appleufslabel *)bp->b_data, NULL);
 709			if (error == 0)
 710				ump->um_flags |= UFS_ISAPPLEUFS;
 711		}
 712		brelse(bp, 0);
 713		bp = NULL;
 714	}
 715#else
 716	if (ump->um_flags & UFS_ISAPPLEUFS)
 717		return (EIO);
 718#endif
 719
 720	if (UFS_MPISAPPLEUFS(ump)) {
 721		/* see comment about NeXT below */
 722		ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
 723		ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
 724		mp->mnt_iflag |= IMNT_DTYPE;
 725	} else {
 726		ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
 727		ump->um_dirblksiz = DIRBLKSIZ;
 728		if (ump->um_maxsymlinklen > 0)
 729			mp->mnt_iflag |= IMNT_DTYPE;
 730		else
 731			mp->mnt_iflag &= ~IMNT_DTYPE;
 732	}
 733	ffs_oldfscompat_read(fs, ump, sblockloc);
 734
 735	mutex_enter(&ump->um_lock);
 736	ump->um_maxfilesize = fs->fs_maxfilesize;
 737	if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
 738		uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
 739		    mp->mnt_stat.f_mntonname, fs->fs_flags,
 740		    (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
 741		if ((mp->mnt_flag & MNT_FORCE) == 0) {
 742			mutex_exit(&ump->um_lock);
 743			return (EINVAL);
 744		}
 745	}
 746	if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
 747		fs->fs_pendingblocks = 0;
 748		fs->fs_pendinginodes = 0;
 749	}
 750	mutex_exit(&ump->um_lock);
 751
 752	ffs_statvfs(mp, &mp->mnt_stat);
 753	/*
 754	 * Step 3: re-read summary information from disk.
 755	 */
 756	blks = howmany(fs->fs_cssize, fs->fs_fsize);
 757	space = fs->fs_csp;
 758	for (i = 0; i < blks; i += fs->fs_frag) {
 759		bsize = fs->fs_bsize;
 760		if (i + fs->fs_frag > blks)
 761			bsize = (blks - i) * fs->fs_fsize;
 762		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), bsize,
 763			      NOCRED, 0, &bp);
 764		if (error) {
 765			brelse(bp, 0);
 766			return (error);
 767		}
 768#ifdef FFS_EI
 769		if (UFS_FSNEEDSWAP(fs))
 770			ffs_csum_swap((struct csum *)bp->b_data,
 771			    (struct csum *)space, bsize);
 772		else
 773#endif
 774			memcpy(space, bp->b_data, (size_t)bsize);
 775		space = (char *)space + bsize;
 776		brelse(bp, 0);
 777	}
 778	if (fs->fs_snapinum[0] != 0)
 779		ffs_snapshot_mount(mp);
 780	/*
 781	 * We no longer know anything about clusters per cylinder group.
 782	 */
 783	if (fs->fs_contigsumsize > 0) {
 784		lp = fs->fs_maxcluster;
 785		for (i = 0; i < fs->fs_ncg; i++)
 786			*lp++ = fs->fs_contigsumsize;
 787	}
 788
 789	/* Allocate a marker vnode. */
 790	mvp = vnalloc(mp);
 791	/*
 792	 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
 793	 * and vclean() can be called indirectly
 794	 */
 795	mutex_enter(&mntvnode_lock);
 796 loop:
 797	for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
 798		vmark(mvp, vp);
 799		if (vp->v_mount != mp || vismarker(vp))
 800			continue;
 801		/*
 802		 * Step 4: invalidate all inactive vnodes.
 803		 */
 804		if (vrecycle(vp, &mntvnode_lock, l)) {
 805			mutex_enter(&mntvnode_lock);
 806			(void)vunmark(mvp);
 807			goto loop;
 808		}
 809		/*
 810		 * Step 5: invalidate all cached file data.
 811		 */
 812		mutex_enter(vp->v_interlock);
 813		mutex_exit(&mntvnode_lock);
 814		if (vget(vp, LK_EXCLUSIVE)) {
 815			(void)vunmark(mvp);
 816			goto loop;
 817		}
 818		if (vinvalbuf(vp, 0, cred, l, 0, 0))
 819			panic("ffs_reload: dirty2");
 820		/*
 821		 * Step 6: re-read inode data for all active vnodes.
 822		 */
 823		ip = VTOI(vp);
 824		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
 825			      (int)fs->fs_bsize, NOCRED, 0, &bp);
 826		if (error) {
 827			brelse(bp, 0);
 828			vput(vp);
 829			(void)vunmark(mvp);
 830			break;
 831		}
 832		ffs_load_inode(bp, ip, fs, ip->i_number);
 833		brelse(bp, 0);
 834		vput(vp);
 835		mutex_enter(&mntvnode_lock);
 836	}
 837	mutex_exit(&mntvnode_lock);
 838	vnfree(mvp);
 839	return (error);
 840}
 841
 842/*
 843 * Possible superblock locations ordered from most to least likely.
 844 */
 845static const int sblock_try[] = SBLOCKSEARCH;
 846
 847/*
 848 * Common code for mount and mountroot
 849 */
 850int
 851ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l)
 852{
 853	struct ufsmount *ump;
 854	struct buf *bp;
 855	struct fs *fs;
 856	dev_t dev;
 857	struct dkwedge_info dkw;
 858	void *space;
 859	daddr_t sblockloc, fsblockloc;
 860	int blks, fstype;
 861	int error, i, bsize, ronly, bset = 0;
 862#ifdef FFS_EI
 863	int needswap = 0;		/* keep gcc happy */
 864#endif
 865	int32_t *lp;
 866	kauth_cred_t cred;
 867	u_int32_t sbsize = 8192;	/* keep gcc happy*/
 868	int32_t fsbsize;
 869
 870	dev = devvp->v_rdev;
 871	cred = l ? l->l_cred : NOCRED;
 872
 873	/* Flush out any old buffers remaining from a previous use. */
 874	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
 875	error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
 876	VOP_UNLOCK(devvp);
 877	if (error)
 878		return (error);
 879
 880	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
 881
 882	bp = NULL;
 883	ump = NULL;
 884	fs = NULL;
 885	sblockloc = 0;
 886	fstype = 0;
 887
 888	error = fstrans_mount(mp);
 889	if (error)
 890		return error;
 891
 892	ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
 893	memset(ump, 0, sizeof *ump);
 894	mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE);
 895	error = ffs_snapshot_init(ump);
 896	if (error)
 897		goto out;
 898	ump->um_ops = &ffs_ufsops;
 899
 900#ifdef WAPBL
 901 sbagain:
 902#endif
 903	/*
 904	 * Try reading the superblock in each of its possible locations.
 905	 */
 906	for (i = 0; ; i++) {
 907		if (bp != NULL) {
 908			brelse(bp, BC_NOCACHE);
 909			bp = NULL;
 910		}
 911		if (sblock_try[i] == -1) {
 912			error = EINVAL;
 913			fs = NULL;
 914			goto out;
 915		}
 916		error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, cred,
 917			      0, &bp);
 918		if (error) {
 919			fs = NULL;
 920			goto out;
 921		}
 922		fs = (struct fs*)bp->b_data;
 923		fsblockloc = sblockloc = sblock_try[i];
 924		if (fs->fs_magic == FS_UFS1_MAGIC) {
 925			sbsize = fs->fs_sbsize;
 926			fstype = UFS1;
 927			fsbsize = fs->fs_bsize;
 928#ifdef FFS_EI
 929			needswap = 0;
 930		} else if (fs->fs_magic == bswap32(FS_UFS1_MAGIC)) {
 931			sbsize = bswap32(fs->fs_sbsize);
 932			fstype = UFS1;
 933			fsbsize = bswap32(fs->fs_bsize);
 934			needswap = 1;
 935#endif
 936		} else if (fs->fs_magic == FS_UFS2_MAGIC) {
 937			sbsize = fs->fs_sbsize;
 938			fstype = UFS2;
 939			fsbsize = fs->fs_bsize;
 940#ifdef FFS_EI
 941			needswap = 0;
 942		} else if (fs->fs_magic == bswap32(FS_UFS2_MAGIC)) {
 943			sbsize = bswap32(fs->fs_sbsize);
 944			fstype = UFS2;
 945			fsbsize = bswap32(fs->fs_bsize);
 946			needswap = 1;
 947#endif
 948		} else
 949			continue;
 950
 951
 952		/* fs->fs_sblockloc isn't defined for old filesystems */
 953		if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) {
 954			if (sblockloc == SBLOCK_UFS2)
 955				/*
 956				 * This is likely to be the first alternate
 957				 * in a filesystem with 64k blocks.
 958				 * Don't use it.
 959				 */
 960				continue;
 961			fsblockloc = sblockloc;
 962		} else {
 963			fsblockloc = fs->fs_sblockloc;
 964#ifdef FFS_EI
 965			if (needswap)
 966				fsblockloc = bswap64(fsblockloc);
 967#endif
 968		}
 969
 970		/* Check we haven't found an alternate superblock */
 971		if (fsblockloc != sblockloc)
 972			continue;
 973
 974		/* Validate size of superblock */
 975		if (sbsize > MAXBSIZE || sbsize < sizeof(struct fs))
 976			continue;
 977
 978		/* Check that we can handle the file system blocksize */
 979		if (fsbsize > MAXBSIZE) {
 980			printf("ffs_mountfs: block size (%d) > MAXBSIZE (%d)\n",
 981			    fsbsize, MAXBSIZE);
 982			continue;
 983		}
 984
 985		/* Ok seems to be a good superblock */
 986		break;
 987	}
 988
 989	fs = malloc((u_long)sbsize, M_UFSMNT, M_WAITOK);
 990	memcpy(fs, bp->b_data, sbsize);
 991	ump->um_fs = fs;
 992
 993#ifdef FFS_EI
 994	if (needswap) {
 995		ffs_sb_swap((struct fs*)bp->b_data, fs);
 996		fs->fs_flags |= FS_SWAPPED;
 997	} else
 998#endif
 999		fs->fs_flags &= ~FS_SWAPPED;
1000
1001#ifdef WAPBL
1002	if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) {
1003		error = ffs_wapbl_replay_start(mp, fs, devvp);
1004		if (error && (mp->mnt_flag & MNT_FORCE) == 0)
1005			goto out;
1006		if (!error) {
1007			if (!ronly) {
1008				/* XXX fsmnt may be stale. */
1009				printf("%s: replaying log to disk\n",
1010				    fs->fs_fsmnt);
1011				error = wapbl_replay_write(mp->mnt_wapbl_replay,
1012				    devvp);
1013				if (error)
1014					goto out;
1015				wapbl_replay_stop(mp->mnt_wapbl_replay);
1016				fs->fs_clean = FS_WASCLEAN;
1017			} else {
1018				/* XXX fsmnt may be stale */
1019				printf("%s: replaying log to memory\n",
1020				    fs->fs_fsmnt);
1021			}
1022
1023			/* Force a re-read of the superblock */
1024			brelse(bp, BC_INVAL);
1025			bp = NULL;
1026			free(fs, M_UFSMNT);
1027			fs = NULL;
1028			goto sbagain;
1029		}
1030	}
1031#else /* !WAPBL */
1032	if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) {
1033		error = EPERM;
1034		goto out;
1035	}
1036#endif /* !WAPBL */
1037
1038	ffs_oldfscompat_read(fs, ump, sblockloc);
1039	ump->um_maxfilesize = fs->fs_maxfilesize;
1040
1041	if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
1042		uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
1043		    mp->mnt_stat.f_mntonname, fs->fs_flags,
1044		    (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1045		if ((mp->mnt_flag & MNT_FORCE) == 0) {
1046			error = EINVAL;
1047			goto out;
1048		}
1049	}
1050
1051	if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
1052		fs->fs_pendingblocks = 0;
1053		fs->fs_pendinginodes = 0;
1054	}
1055
1056	ump->um_fstype = fstype;
1057	if (fs->fs_sbsize < SBLOCKSIZE)
1058		brelse(bp, BC_INVAL);
1059	else
1060		brelse(bp, 0);
1061	bp = NULL;
1062
1063	/* First check to see if this is tagged as an Apple UFS filesystem
1064	 * in the disklabel
1065	 */
1066	if (getdiskinfo(devvp, &dkw) == 0 &&
1067	    strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0)
1068		ump->um_flags |= UFS_ISAPPLEUFS;
1069#ifdef APPLE_UFS
1070	else {
1071		/* Manually look for an apple ufs label, and if a valid one
1072		 * is found, then treat it like an Apple UFS filesystem anyway
1073		 */
1074		error = bread(devvp, (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE),
1075			APPLEUFS_LABEL_SIZE, cred, 0, &bp);
1076		if (error)
1077			goto out;
1078		error = ffs_appleufs_validate(fs->fs_fsmnt,
1079			(struct appleufslabel *)bp->b_data, NULL);
1080		if (error == 0) {
1081			ump->um_flags |= UFS_ISAPPLEUFS;
1082		}
1083		brelse(bp, 0);
1084		bp = NULL;
1085	}
1086#else
1087	if (ump->um_flags & UFS_ISAPPLEUFS) {
1088		error = EINVAL;
1089		goto out;
1090	}
1091#endif
1092
1093#if 0
1094/*
1095 * XXX This code changes the behaviour of mounting dirty filesystems, to
1096 * XXX require "mount -f ..." to mount them.  This doesn't match what
1097 * XXX mount(8) describes and is disabled for now.
1098 */
1099	/*
1100	 * If the file system is not clean, don't allow it to be mounted
1101	 * unless MNT_FORCE is specified.  (Note: MNT_FORCE is always set
1102	 * for the root file system.)
1103	 */
1104	if (fs->fs_flags & FS_DOWAPBL) {
1105		/*
1106		 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL
1107		 * bit is set, although there's a window in unmount where it
1108		 * could be FS_ISCLEAN
1109		 */
1110		if ((mp->mnt_flag & MNT_FORCE) == 0 &&
1111		    (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) {
1112			error = EPERM;
1113			goto out;
1114		}
1115	} else
1116		if ((fs->fs_clean & FS_ISCLEAN) == 0 &&
1117		    (mp->mnt_flag & MNT_FORCE) == 0) {
1118			error = EPERM;
1119			goto out;
1120		}
1121#endif
1122
1123	/*
1124	 * verify that we can access the last block in the fs
1125	 * if we're mounting read/write.
1126	 */
1127
1128	if (!ronly) {
1129		error = bread(devvp, fsbtodb(fs, fs->fs_size - 1), fs->fs_fsize,
1130		    cred, 0, &bp);
1131		if (bp->b_bcount != fs->fs_fsize)
1132			error = EINVAL;
1133		if (error) {
1134			bset = BC_INVAL;
1135			goto out;
1136		}
1137		brelse(bp, BC_INVAL);
1138		bp = NULL;
1139	}
1140
1141	fs->fs_ronly = ronly;
1142	/* Don't bump fs_clean if we're replaying journal */
1143	if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN)))
1144		if (ronly == 0) {
1145			fs->fs_clean <<= 1;
1146			fs->fs_fmod = 1;
1147		}
1148	bsize = fs->fs_cssize;
1149	blks = howmany(bsize, fs->fs_fsize);
1150	if (fs->fs_contigsumsize > 0)
1151		bsize += fs->fs_ncg * sizeof(int32_t);
1152	bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1153	space = malloc((u_long)bsize, M_UFSMNT, M_WAITOK);
1154	fs->fs_csp = space;
1155	for (i = 0; i < blks; i += fs->fs_frag) {
1156		bsize = fs->fs_bsize;
1157		if (i + fs->fs_frag > blks)
1158			bsize = (blks - i) * fs->fs_fsize;
1159		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), bsize,
1160			      cred, 0, &bp);
1161		if (error) {
1162			free(fs->fs_csp, M_UFSMNT);
1163			goto out;
1164		}
1165#ifdef FFS_EI
1166		if (needswap)
1167			ffs_csum_swap((struct csum *)bp->b_data,
1168				(struct csum *)space, bsize);
1169		else
1170#endif
1171			memcpy(space, bp->b_data, (u_int)bsize);
1172
1173		space = (char *)space + bsize;
1174		brelse(bp, 0);
1175		bp = NULL;
1176	}
1177	if (fs->fs_contigsumsize > 0) {
1178		fs->fs_maxcluster = lp = space;
1179		for (i = 0; i < fs->fs_ncg; i++)
1180			*lp++ = fs->fs_contigsumsize;
1181		space = lp;
1182	}
1183	bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1184	fs->fs_contigdirs = space;
1185	space = (char *)space + bsize;
1186	memset(fs->fs_contigdirs, 0, bsize);
1187		/* Compatibility for old filesystems - XXX */
1188	if (fs->fs_avgfilesize <= 0)
1189		fs->fs_avgfilesize = AVFILESIZ;
1190	if (fs->fs_avgfpdir <= 0)
1191		fs->fs_avgfpdir = AFPDIR;
1192	fs->fs_active = NULL;
1193	mp->mnt_data = ump;
1194	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
1195	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS);
1196	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
1197	mp->mnt_stat.f_namemax = FFS_MAXNAMLEN;
1198	if (UFS_MPISAPPLEUFS(ump)) {
1199		/* NeXT used to keep short symlinks in the inode even
1200		 * when using FS_42INODEFMT.  In that case fs->fs_maxsymlinklen
1201		 * is probably -1, but we still need to be able to identify
1202		 * short symlinks.
1203		 */
1204		ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
1205		ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
1206		mp->mnt_iflag |= IMNT_DTYPE;
1207	} else {
1208		ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
1209		ump->um_dirblksiz = DIRBLKSIZ;
1210		if (ump->um_maxsymlinklen > 0)
1211			mp->mnt_iflag |= IMNT_DTYPE;
1212		else
1213			mp->mnt_iflag &= ~IMNT_DTYPE;
1214	}
1215	mp->mnt_fs_bshift = fs->fs_bshift;
1216	mp->mnt_dev_bshift = DEV_BSHIFT;	/* XXX */
1217	mp->mnt_flag |= MNT_LOCAL;
1218	mp->mnt_iflag |= IMNT_MPSAFE;
1219#ifdef FFS_EI
1220	if (needswap)
1221		ump->um_flags |= UFS_NEEDSWAP;
1222#endif
1223	ump->um_mountp = mp;
1224	ump->um_dev = dev;
1225	ump->um_devvp = devvp;
1226	ump->um_nindir = fs->fs_nindir;
1227	ump->um_lognindir = ffs(fs->fs_nindir) - 1;
1228	ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT;
1229	ump->um_seqinc = fs->fs_frag;
1230	for (i = 0; i < MAXQUOTAS; i++)
1231		ump->um_quotas[i] = NULLVP;
1232	devvp->v_specmountpoint = mp;
1233	if (ronly == 0 && fs->fs_snapinum[0] != 0)
1234		ffs_snapshot_mount(mp);
1235#ifdef WAPBL
1236	if (!ronly) {
1237		KDASSERT(fs->fs_ronly == 0);
1238		/*
1239		 * ffs_wapbl_start() needs mp->mnt_stat initialised if it
1240		 * needs to create a new log file in-filesystem.
1241		 */
1242		ffs_statvfs(mp, &mp->mnt_stat);
1243
1244		error = ffs_wapbl_start(mp);
1245		if (error) {
1246			free(fs->fs_csp, M_UFSMNT);
1247			goto out;
1248		}
1249	}
1250#endif /* WAPBL */
1251	if (ronly == 0) {
1252#ifdef QUOTA2
1253		error = ffs_quota2_mount(mp);
1254		if (error) {
1255			free(fs->fs_csp, M_UFSMNT);
1256			goto out;
1257		}
1258#else
1259		if (fs->fs_flags & FS_DOQUOTA2) {
1260			ump->um_flags |= UFS_QUOTA2;
1261			uprintf("%s: options QUOTA2 not enabled%s\n",
1262			    mp->mnt_stat.f_mntonname,
1263			    (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1264			if ((mp->mnt_flag & MNT_FORCE) == 0) {
1265				error = EINVAL;
1266				free(fs->fs_csp, M_UFSMNT);
1267				goto out;
1268			}
1269		}
1270#endif
1271	 }
1272#ifdef UFS_EXTATTR
1273	/*
1274	 * Initialize file-backed extended attributes on UFS1 file
1275	 * systems.
1276	 */
1277	if (ump->um_fstype == UFS1)
1278		ufs_extattr_uepm_init(&ump->um_extattr);	
1279#endif /* UFS_EXTATTR */
1280
1281	return (0);
1282out:
1283#ifdef WAPBL
1284	if (mp->mnt_wapbl_replay) {
1285		wapbl_replay_stop(mp->mnt_wapbl_replay);
1286		wapbl_replay_free(mp->mnt_wapbl_replay);
1287		mp->mnt_wapbl_replay = 0;
1288	}
1289#endif
1290
1291	fstrans_unmount(mp);
1292	if (fs)
1293		free(fs, M_UFSMNT);
1294	devvp->v_specmountpoint = NULL;
1295	if (bp)
1296		brelse(bp, bset);
1297	if (ump) {
1298		if (ump->um_oldfscompat)
1299			free(ump->um_oldfscompat, M_UFSMNT);
1300		mutex_destroy(&ump->um_lock);
1301		free(ump, M_UFSMNT);
1302		mp->mnt_data = NULL;
1303	}
1304	return (error);
1305}
1306
1307/*
1308 * Sanity checks for loading old filesystem superblocks.
1309 * See ffs_oldfscompat_write below for unwound actions.
1310 *
1311 * XXX - Parts get retired eventually.
1312 * Unfortunately new bits get added.
1313 */
1314static void
1315ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc)
1316{
1317	off_t maxfilesize;
1318	int32_t *extrasave;
1319
1320	if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1321	    (fs->fs_old_flags & FS_FLAGS_UPDATED))
1322		return;
1323
1324	if (!ump->um_oldfscompat)
1325		ump->um_oldfscompat = malloc(512 + 3*sizeof(int32_t),
1326		    M_UFSMNT, M_WAITOK);
1327
1328	memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512);
1329	extrasave = ump->um_oldfscompat;
1330	extrasave += 512/sizeof(int32_t);
1331	extrasave[0] = fs->fs_old_npsect;
1332	extrasave[1] = fs->fs_old_interleave;
1333	extrasave[2] = fs->fs_old_trackskew;
1334
1335	/* These fields will be overwritten by their
1336	 * original values in fs_oldfscompat_write, so it is harmless
1337	 * to modify them here.
1338	 */
1339	fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir;
1340	fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree;
1341	fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree;
1342	fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree;
1343
1344	fs->fs_maxbsize = fs->fs_bsize;
1345	fs->fs_time = fs->fs_old_time;
1346	fs->fs_size = fs->fs_old_size;
1347	fs->fs_dsize = fs->fs_old_dsize;
1348	fs->fs_csaddr = fs->fs_old_csaddr;
1349	fs->fs_sblockloc = sblockloc;
1350
1351	fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL);
1352
1353	if (fs->fs_old_postblformat == FS_42POSTBLFMT) {
1354		fs->fs_old_nrpos = 8;
1355		fs->fs_old_npsect = fs->fs_old_nsect;
1356		fs->fs_old_interleave = 1;
1357		fs->fs_old_trackskew = 0;
1358	}
1359
1360	if (fs->fs_old_inodefmt < FS_44INODEFMT) {
1361		fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
1362		fs->fs_qbmask = ~fs->fs_bmask;
1363		fs->fs_qfmask = ~fs->fs_fmask;
1364	}
1365
1366	maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;
1367	if (fs->fs_maxfilesize > maxfilesize)
1368		fs->fs_maxfilesize = maxfilesize;
1369
1370	/* Compatibility for old filesystems */
1371	if (fs->fs_avgfilesize <= 0)
1372		fs->fs_avgfilesize = AVFILESIZ;
1373	if (fs->fs_avgfpdir <= 0)
1374		fs->fs_avgfpdir = AFPDIR;
1375
1376#if 0
1377	if (bigcgs) {
1378		fs->fs_save_cgsize = fs->fs_cgsize;
1379		fs->fs_cgsize = fs->fs_bsize;
1380	}
1381#endif
1382}
1383
1384/*
1385 * Unwinding superblock updates for old filesystems.
1386 * See ffs_oldfscompat_read above for details.
1387 *
1388 * XXX - Parts get retired eventually.
1389 * Unfortunately new bits get added.
1390 */
1391static void
1392ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump)
1393{
1394	int32_t *extrasave;
1395
1396	if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1397	    (fs->fs_old_flags & FS_FLAGS_UPDATED))
1398		return;
1399
1400	fs->fs_old_time = fs->fs_time;
1401	fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir;
1402	fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree;
1403	fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree;
1404	fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree;
1405	fs->fs_old_flags = fs->fs_flags;
1406
1407#if 0
1408	if (bigcgs) {
1409		fs->fs_cgsize = fs->fs_save_cgsize;
1410	}
1411#endif
1412
1413	memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512);
1414	extrasave = ump->um_oldfscompat;
1415	extrasave += 512/sizeof(int32_t);
1416	fs->fs_old_npsect = extrasave[0];
1417	fs->fs_old_interleave = extrasave[1];
1418	fs->fs_old_trackskew = extrasave[2];
1419
1420}
1421
1422/*
1423 * unmount vfs operation
1424 */
1425int
1426ffs_unmount(struct mount *mp, int mntflags)
1427{
1428	struct lwp *l = curlwp;
1429	struct ufsmount *ump = VFSTOUFS(mp);
1430	struct fs *fs = ump->um_fs;
1431	int error, flags;
1432#ifdef WAPBL
1433	extern int doforce;
1434#endif
1435
1436	flags = 0;
1437	if (mntflags & MNT_FORCE)
1438		flags |= FORCECLOSE;
1439	if ((error = ffs_flushfiles(mp, flags, l)) != 0)
1440		return (error);
1441	error = UFS_WAPBL_BEGIN(mp);
1442	if (error == 0)
1443		if (fs->fs_ronly == 0 &&
1444		    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
1445		    fs->fs_clean & FS_WASCLEAN) {
1446			fs->fs_clean = FS_ISCLEAN;
1447			fs->fs_fmod = 0;
1448			(void) ffs_sbupdate(ump, MNT_WAIT);
1449		}
1450	if (error == 0)
1451		UFS_WAPBL_END(mp);
1452#ifdef WAPBL
1453	KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl));
1454	if (mp->mnt_wapbl_replay) {
1455		KDASSERT(fs->fs_ronly);
1456		wapbl_replay_stop(mp->mnt_wapbl_replay);
1457		wapbl_replay_free(mp->mnt_wapbl_replay);
1458		mp->mnt_wapbl_replay = 0;
1459	}
1460	error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE));
1461	if (error) {
1462		return error;
1463	}
1464#endif /* WAPBL */
1465#ifdef UFS_EXTATTR
1466	if (ump->um_fstype == UFS1) {
1467		ufs_extattr_stop(mp, l);
1468		ufs_extattr_uepm_destroy(&ump->um_extattr);
1469	}
1470#endif /* UFS_EXTATTR */
1471
1472	if (ump->um_devvp->v_type != VBAD)
1473		ump->um_devvp->v_specmountpoint = NULL;
1474	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1475	(void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE,
1476		NOCRED);
1477	vput(ump->um_devvp);
1478	free(fs->fs_csp, M_UFSMNT);
1479	free(fs, M_UFSMNT);
1480	if (ump->um_oldfscompat != NULL)
1481		free(ump->um_oldfscompat, M_UFSMNT);
1482	mutex_destroy(&ump->um_lock);
1483	ffs_snapshot_fini(ump);
1484	free(ump, M_UFSMNT);
1485	mp->mnt_data = NULL;
1486	mp->mnt_flag &= ~MNT_LOCAL;
1487	fstrans_unmount(mp);
1488	return (0);
1489}
1490
1491/*
1492 * Flush out all the files in a filesystem.
1493 */
1494int
1495ffs_flushfiles(struct mount *mp, int flags, struct lwp *l)
1496{
1497	extern int doforce;
1498	struct ufsmount *ump;
1499	int error;
1500
1501	if (!doforce)
1502		flags &= ~FORCECLOSE;
1503	ump = VFSTOUFS(mp);
1504#ifdef QUOTA
1505	if ((error = quota1_umount(mp, flags)) != 0)
1506		return (error);
1507#endif
1508#ifdef QUOTA2
1509	if ((error = quota2_umount(mp, flags)) != 0)
1510		return (error);
1511#endif
1512	if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0)
1513		return (error);
1514	ffs_snapshot_unmount(mp);
1515	/*
1516	 * Flush all the files.
1517	 */
1518	error = vflush(mp, NULLVP, flags);
1519	if (error)
1520		return (error);
1521	/*
1522	 * Flush filesystem metadata.
1523	 */
1524	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1525	error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0);
1526	VOP_UNLOCK(ump->um_devvp);
1527	if (flags & FORCECLOSE) /* XXXDBJ */
1528		error = 0;
1529
1530#ifdef WAPBL
1531	if (error)
1532		return error;
1533	if (mp->mnt_wapbl) {
1534		error = wapbl_flush(mp->mnt_wapbl, 1);
1535		if (flags & FORCECLOSE)
1536			error = 0;
1537	}
1538#endif
1539
1540	return (error);
1541}
1542
1543/*
1544 * Get file system statistics.
1545 */
1546int
1547ffs_statvfs(struct mount *mp, struct statvfs *sbp)
1548{
1549	struct ufsmount *ump;
1550	struct fs *fs;
1551
1552	ump = VFSTOUFS(mp);
1553	fs = ump->um_fs;
1554	mutex_enter(&ump->um_lock);
1555	sbp->f_bsize = fs->fs_bsize;
1556	sbp->f_frsize = fs->fs_fsize;
1557	sbp->f_iosize = fs->fs_bsize;
1558	sbp->f_blocks = fs->fs_dsize;
1559	sbp->f_bfree = blkstofrags(fs, fs->fs_cstotal.cs_nbfree) +
1560	    fs->fs_cstotal.cs_nffree + dbtofsb(fs, fs->fs_pendingblocks);
1561	sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t)
1562	    fs->fs_minfree) / (u_int64_t) 100;
1563	if (sbp->f_bfree > sbp->f_bresvd)
1564		sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
1565	else
1566		sbp->f_bavail = 0;
1567	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - ROOTINO;
1568	sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes;
1569	sbp->f_favail = sbp->f_ffree;
1570	sbp->f_fresvd = 0;
1571	mutex_exit(&ump->um_lock);
1572	copy_statvfs_info(sbp, mp);
1573
1574	return (0);
1575}
1576
1577/*
1578 * Go through the disk queues to initiate sandbagged IO;
1579 * go through the inodes to write those that have been modified;
1580 * initiate the writing of the super block if it has been modified.
1581 *
1582 * Note: we are always called with the filesystem marked `MPBUSY'.
1583 */
1584int
1585ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1586{
1587	struct vnode *vp, *mvp, *nvp;
1588	struct inode *ip;
1589	struct ufsmount *ump = VFSTOUFS(mp);
1590	struct fs *fs;
1591	int error, allerror = 0;
1592	bool is_suspending;
1593
1594	fs = ump->um_fs;
1595	if (fs->fs_fmod != 0 && fs->fs_ronly != 0) {		/* XXX */
1596		printf("fs = %s\n", fs->fs_fsmnt);
1597		panic("update: rofs mod");
1598	}
1599
1600	/* Allocate a marker vnode. */
1601	mvp = vnalloc(mp);
1602
1603	fstrans_start(mp, FSTRANS_SHARED);
1604	is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING);
1605	/*
1606	 * Write back each (modified) inode.
1607	 */
1608	mutex_enter(&mntvnode_lock);
1609loop:
1610	/*
1611	 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
1612	 * and vclean() can be called indirectly
1613	 */
1614	for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = nvp) {
1615		nvp = TAILQ_NEXT(vp, v_mntvnodes);
1616		/*
1617		 * If the vnode that we are about to sync is no longer
1618		 * associated with this mount point, start over.
1619		 */
1620		if (vp->v_mount != mp)
1621			goto loop;
1622		/*
1623		 * Don't interfere with concurrent scans of this FS.
1624		 */
1625		if (vismarker(vp))
1626			continue;
1627		mutex_enter(vp->v_interlock);
1628		ip = VTOI(vp);
1629
1630		/*
1631		 * Skip the vnode/inode if inaccessible.
1632		 */
1633		if (ip == NULL || (vp->v_iflag & (VI_XLOCK | VI_CLEAN)) != 0 ||
1634		    vp->v_type == VNON) {
1635			mutex_exit(vp->v_interlock);
1636			continue;
1637		}
1638
1639		/*
1640		 * We deliberately update inode times here.  This will
1641		 * prevent a massive queue of updates accumulating, only
1642		 * to be handled by a call to unmount.
1643		 *
1644		 * XXX It would be better to have the syncer trickle these
1645		 * out.  Adjustment needed to allow registering vnodes for
1646		 * sync when the vnode is clean, but the inode dirty.  Or
1647		 * have ufs itself trickle out inode updates.
1648		 *
1649		 * If doing a lazy sync, we don't care about metadata or
1650		 * data updates, because they are handled by each vnode's
1651		 * synclist entry.  In this case we are only interested in
1652		 * writing back modified inodes.
1653		 */
1654		if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE |
1655		    IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 &&
1656		    (waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) &&
1657		    UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) {
1658			mutex_exit(vp->v_interlock);
1659			continue;
1660		}
1661		if (vp->v_type == VBLK && is_suspending) {
1662			mutex_exit(vp->v_interlock);
1663			continue;
1664		}
1665		vmark(mvp, vp);
1666		mutex_exit(&mntvnode_lock);
1667		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT);
1668		if (error) {
1669			mutex_enter(&mntvnode_lock);
1670			nvp = vunmark(mvp);
1671			if (error == ENOENT) {
1672				goto loop;
1673			}
1674			continue;
1675		}
1676		if (waitfor == MNT_LAZY) {
1677			error = UFS_WAPBL_BEGIN(vp->v_mount);
1678			if (!error) {
1679				error = ffs_update(vp, NULL, NULL,
1680				    UPDATE_CLOSE);
1681				UFS_WAPBL_END(vp->v_mount);
1682			}
1683		} else {
1684			error = VOP_FSYNC(vp, cred, FSYNC_NOLOG |
1685			    (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0);
1686		}
1687		if (error)
1688			allerror = error;
1689		vput(vp);
1690		mutex_enter(&mntvnode_lock);
1691		nvp = vunmark(mvp);
1692	}
1693	mutex_exit(&mntvnode_lock);
1694	/*
1695	 * Force stale file system control information to be flushed.
1696	 */
1697	if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 ||
1698	    !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) {
1699		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1700		if ((error = VOP_FSYNC(ump->um_devvp, cred,
1701		    (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG,
1702		    0, 0)) != 0)
1703			allerror = error;
1704		VOP_UNLOCK(ump->um_devvp);
1705		if (allerror == 0 && waitfor == MNT_WAIT && !mp->mnt_wapbl) {
1706			mutex_enter(&mntvnode_lock);
1707			goto loop;
1708		}
1709	}
1710#if defined(QUOTA) || defined(QUOTA2)
1711	qsync(mp);
1712#endif
1713	/*
1714	 * Write back modified superblock.
1715	 */
1716	if (fs->fs_fmod != 0) {
1717		fs->fs_fmod = 0;
1718		fs->fs_time = time_second;
1719		error = UFS_WAPBL_BEGIN(mp);
1720		if (error)
1721			allerror = error;
1722		else {
1723			if ((error = ffs_cgupdate(ump, waitfor)))
1724				allerror = error;
1725			UFS_WAPBL_END(mp);
1726		}
1727	}
1728
1729#ifdef WAPBL
1730	if (mp->mnt_wapbl) {
1731		error = wapbl_flush(mp->mnt_wapbl, 0);
1732		if (error)
1733			allerror = error;
1734	}
1735#endif
1736
1737	fstrans_done(mp);
1738	vnfree(mvp);
1739	return (allerror);
1740}
1741
1742/*
1743 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1744 * in from disk.  If it is in core, wait for the lock bit to clear, then
1745 * return the inode locked.  Detection and handling of mount points must be
1746 * done by the calling routine.
1747 */
1748int
1749ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1750{
1751	struct fs *fs;
1752	struct inode *ip;
1753	struct ufsmount *ump;
1754	struct buf *bp;
1755	struct vnode *vp;
1756	dev_t dev;
1757	int error;
1758
1759	ump = VFSTOUFS(mp);
1760	dev = ump->um_dev;
1761
1762 retry:
1763	if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
1764		return (0);
1765
1766	/* Allocate a new vnode/inode. */
1767	error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, NULL, &vp);
1768	if (error) {
1769		*vpp = NULL;
1770		return (error);
1771	}
1772	ip = pool_cache_get(ffs_inode_cache, PR_WAITOK);
1773
1774	/*
1775	 * If someone beat us to it, put back the freshly allocated
1776	 * vnode/inode pair and retry.
1777	 */
1778	mutex_enter(&ufs_hashlock);
1779	if (ufs_ihashget(dev, ino, 0) != NULL) {
1780		mutex_exit(&ufs_hashlock);
1781		ungetnewvnode(vp);
1782		pool_cache_put(ffs_inode_cache, ip);
1783		goto retry;
1784	}
1785
1786	vp->v_vflag |= VV_LOCKSWORK;
1787
1788	/*
1789	 * XXX MFS ends up here, too, to allocate an inode.  Should we
1790	 * XXX create another pool for MFS inodes?
1791	 */
1792
1793	memset(ip, 0, sizeof(struct inode));
1794	vp->v_data = ip;
1795	ip->i_vnode = vp;
1796	ip->i_ump = ump;
1797	ip->i_fs = fs = ump->um_fs;
1798	ip->i_dev = dev;
1799	ip->i_number = ino;
1800#if defined(QUOTA) || defined(QUOTA2)
1801	ufsquota_init(ip);
1802#endif
1803
1804	/*
1805	 * Initialize genfs node, we might proceed to destroy it in
1806	 * error branches.
1807	 */
1808	genfs_node_init(vp, &ffs_genfsops);
1809
1810	/*
1811	 * Put it onto its hash chain and lock it so that other requests for
1812	 * this inode will block if they arrive while we are sleeping waiting
1813	 * for old data structures to be purged or for the contents of the
1814	 * disk portion of this inode to be read.
1815	 */
1816
1817	ufs_ihashins(ip);
1818	mutex_exit(&ufs_hashlock);
1819
1820	/* Read in the disk contents for the inode, copy into the inode. */
1821	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1822		      (int)fs->fs_bsize, NOCRED, 0, &bp);
1823	if (error) {
1824
1825		/*
1826		 * The inode does not contain anything useful, so it would
1827		 * be misleading to leave it on its hash chain. With mode
1828		 * still zero, it will be unlinked and returned to the free
1829		 * list by vput().
1830		 */
1831
1832		vput(vp);
1833		brelse(bp, 0);
1834		*vpp = NULL;
1835		return (error);
1836	}
1837	if (ip->i_ump->um_fstype == UFS1)
1838		ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache,
1839		    PR_WAITOK);
1840	else
1841		ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache,
1842		    PR_WAITOK);
1843	ffs_load_inode(bp, ip, fs, ino);
1844	brelse(bp, 0);
1845
1846	/*
1847	 * Initialize the vnode from the inode, check for aliases.
1848	 * Note that the underlying vnode may have changed.
1849	 */
1850
1851	ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
1852
1853	/*
1854	 * Finish inode initialization now that aliasing has been resolved.
1855	 */
1856
1857	ip->i_devvp = ump->um_devvp;
1858	vref(ip->i_devvp);
1859
1860	/*
1861	 * Ensure that uid and gid are correct. This is a temporary
1862	 * fix until fsck has been changed to do the update.
1863	 */
1864
1865	if (fs->fs_old_inodefmt < FS_44INODEFMT) {		/* XXX */
1866		ip->i_uid = ip->i_ffs1_ouid;			/* XXX */
1867		ip->i_gid = ip->i_ffs1_ogid;			/* XXX */
1868	}							/* XXX */
1869	uvm_vnp_setsize(vp, ip->i_size);
1870	*vpp = vp;
1871	return (0);
1872}
1873
1874/*
1875 * File handle to vnode
1876 *
1877 * Have to be really careful about stale file handles:
1878 * - check that the inode number is valid
1879 * - call ffs_vget() to get the locked inode
1880 * - check for an unallocated inode (i_mode == 0)
1881 * - check that the given client host has export rights and return
1882 *   those rights via. exflagsp and credanonp
1883 */
1884int
1885ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1886{
1887	struct ufid ufh;
1888	struct fs *fs;
1889
1890	if (fhp->fid_len != sizeof(struct ufid))
1891		return EINVAL;
1892
1893	memcpy(&ufh, fhp, sizeof(ufh));
1894	fs = VFSTOUFS(mp)->um_fs;
1895	if (ufh.ufid_ino < ROOTINO ||
1896	    ufh.ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1897		return (ESTALE);
1898	return (ufs_fhtovp(mp, &ufh, vpp));
1899}
1900
1901/*
1902 * Vnode pointer to File handle
1903 */
1904/* ARGSUSED */
1905int
1906ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1907{
1908	struct inode *ip;
1909	struct ufid ufh;
1910
1911	if (*fh_size < sizeof(struct ufid)) {
1912		*fh_size = sizeof(struct ufid);
1913		return E2BIG;
1914	}
1915	ip = VTOI(vp);
1916	*fh_size = sizeof(struct ufid);
1917	memset(&ufh, 0, sizeof(ufh));
1918	ufh.ufid_len = sizeof(struct ufid);
1919	ufh.ufid_ino = ip->i_number;
1920	ufh.ufid_gen = ip->i_gen;
1921	memcpy(fhp, &ufh, sizeof(ufh));
1922	return (0);
1923}
1924
1925void
1926ffs_init(void)
1927{
1928	if (ffs_initcount++ > 0)
1929		return;
1930
1931	ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0,
1932	    "ffsino", NULL, IPL_NONE, NULL, NULL, NULL);
1933	ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0,
1934	    "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL);
1935	ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0,
1936	    "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL);
1937	ufs_init();
1938}
1939
1940void
1941ffs_reinit(void)
1942{
1943
1944	ufs_reinit();
1945}
1946
1947void
1948ffs_done(void)
1949{
1950	if (--ffs_initcount > 0)
1951		return;
1952
1953	ufs_done();
1954	pool_cache_destroy(ffs_dinode2_cache);
1955	pool_cache_destroy(ffs_dinode1_cache);
1956	pool_cache_destroy(ffs_inode_cache);
1957}
1958
1959/*
1960 * Write a superblock and associated information back to disk.
1961 */
1962int
1963ffs_sbupdate(struct ufsmou

Large files files are truncated, but you can click here to view the full file