common/fs/vfs.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

/*	Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T	*/
/*	  All Rights Reserved  	*/

/*
 * University Copyright- Copyright (c) 1982, 1986, 1988
 * The Regents of the University of California
 * All Rights Reserved
 *
 * University Acknowledgment- Portions of this document are derived from
 * software developed by the University of California, Berkeley, and its
 * contributors.
 */


#include <sys/types.h>
#include <sys/t_lock.h>
#include <sys/param.h>
#include <sys/errno.h>
#include <sys/user.h>
#include <sys/fstyp.h>
#include <sys/kmem.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/mount.h>
#include <sys/vfs.h>
#include <sys/vfs_opreg.h>
#include <sys/fem.h>
#include <sys/mntent.h>
#include <sys/stat.h>
#include <sys/statvfs.h>
#include <sys/statfs.h>
#include <sys/cred.h>
#include <sys/vnode.h>
#include <sys/rwstlock.h>
#include <sys/dnlc.h>
#include <sys/file.h>
#include <sys/time.h>
#include <sys/atomic.h>
#include <sys/cmn_err.h>
#include <sys/buf.h>
#include <sys/swap.h>
#include <sys/debug.h>
#include <sys/vnode.h>
#include <sys/modctl.h>
#include <sys/ddi.h>
#include <sys/pathname.h>
#include <sys/bootconf.h>
#include <sys/dumphdr.h>
#include <sys/dc_ki.h>
#include <sys/poll.h>
#include <sys/sunddi.h>
#include <sys/sysmacros.h>
#include <sys/zone.h>
#include <sys/policy.h>
#include <sys/ctfs.h>
#include <sys/objfs.h>
#include <sys/console.h>
#include <sys/reboot.h>
#include <sys/attr.h>
#include <sys/spa.h>
#include <sys/lofi.h>

#include <vm/page.h>

#include <fs/fs_subr.h>

/* Private interfaces to create vopstats-related data structures */
extern void		initialize_vopstats(vopstats_t *);
extern vopstats_t	*get_fstype_vopstats(struct vfs *, struct vfssw *);
extern vsk_anchor_t	*get_vskstat_anchor(struct vfs *);

static void vfs_clearmntopt_nolock(mntopts_t *, const char *, int);
static void vfs_setmntopt_nolock(mntopts_t *, const char *,
    const char *, int, int);
static int  vfs_optionisset_nolock(const mntopts_t *, const char *, char **);
static void vfs_freemnttab(struct vfs *);
static void vfs_freeopt(mntopt_t *);
static void vfs_swapopttbl_nolock(mntopts_t *, mntopts_t *);
static void vfs_swapopttbl(mntopts_t *, mntopts_t *);
static void vfs_copyopttbl_extend(const mntopts_t *, mntopts_t *, int);
static void vfs_createopttbl_extend(mntopts_t *, const char *,
    const mntopts_t *);
static char **vfs_copycancelopt_extend(char **const, int);
static void vfs_freecancelopt(char **);
static void getrootfs(char **, char **);
static int getmacpath(dev_info_t *, void *);
static void vfs_mnttabvp_setup(void);

struct ipmnt {
	struct ipmnt	*mip_next;
	dev_t		mip_dev;
	struct vfs	*mip_vfsp;
};

static kmutex_t		vfs_miplist_mutex;
static struct ipmnt	*vfs_miplist = NULL;
static struct ipmnt	*vfs_miplist_end = NULL;

static kmem_cache_t *vfs_cache;	/* Pointer to VFS kmem cache */

/*
 * VFS global data.
 */
vnode_t *rootdir;		/* pointer to root inode vnode. */
vnode_t *devicesdir;		/* pointer to inode of devices root */
vnode_t	*devdir;		/* pointer to inode of dev root */

char *server_rootpath;		/* root path for diskless clients */
char *server_hostname;		/* hostname of diskless server */

static struct vfs root;
static struct vfs devices;
static struct vfs dev;
struct vfs *rootvfs = &root;	/* pointer to root vfs; head of VFS list. */
rvfs_t *rvfs_list;		/* array of vfs ptrs for vfs hash list */
int vfshsz = 512;		/* # of heads/locks in vfs hash arrays */
				/* must be power of 2!	*/
timespec_t vfs_mnttab_ctime;	/* mnttab created time */
timespec_t vfs_mnttab_mtime;	/* mnttab last modified time */
char *vfs_dummyfstype = "\0";
struct pollhead vfs_pollhd;	/* for mnttab pollers */
struct vnode *vfs_mntdummyvp;	/* to fake mnttab read/write for file events */
int	mntfstype;		/* will be set once mnt fs is mounted */

/*
 * Table for generic options recognized in the VFS layer and acted
 * on at this level before parsing file system specific options.
 * The nosuid option is stronger than any of the devices and setuid
 * options, so those are canceled when nosuid is seen.
 *
 * All options which are added here need to be added to the
 * list of standard options in usr/src/cmd/fs.d/fslib.c as well.
 */
/*
 * VFS Mount options table
 */
static char *ro_cancel[] = { MNTOPT_RW, NULL };
static char *rw_cancel[] = { MNTOPT_RO, NULL };
static char *suid_cancel[] = { MNTOPT_NOSUID, NULL };
static char *nosuid_cancel[] = { MNTOPT_SUID, MNTOPT_DEVICES, MNTOPT_NODEVICES,
    MNTOPT_NOSETUID, MNTOPT_SETUID, NULL };
static char *devices_cancel[] = { MNTOPT_NODEVICES, NULL };
static char *nodevices_cancel[] = { MNTOPT_DEVICES, NULL };
static char *setuid_cancel[] = { MNTOPT_NOSETUID, NULL };
static char *nosetuid_cancel[] = { MNTOPT_SETUID, NULL };
static char *nbmand_cancel[] = { MNTOPT_NONBMAND, NULL };
static char *nonbmand_cancel[] = { MNTOPT_NBMAND, NULL };
static char *exec_cancel[] = { MNTOPT_NOEXEC, NULL };
static char *noexec_cancel[] = { MNTOPT_EXEC, NULL };

static const mntopt_t mntopts[] = {
/*
 *	option name		cancel options		default arg	flags
 */
	{ MNTOPT_REMOUNT,	NULL,			NULL,
		MO_NODISPLAY, (void *)0 },
	{ MNTOPT_RO,		ro_cancel,		NULL,		0,
		(void *)0 },
	{ MNTOPT_RW,		rw_cancel,		NULL,		0,
		(void *)0 },
	{ MNTOPT_SUID,		suid_cancel,		NULL,		0,
		(void *)0 },
	{ MNTOPT_NOSUID,	nosuid_cancel,		NULL,		0,
		(void *)0 },
	{ MNTOPT_DEVICES,	devices_cancel,		NULL,		0,
		(void *)0 },
	{ MNTOPT_NODEVICES,	nodevices_cancel,	NULL,		0,
		(void *)0 },
	{ MNTOPT_SETUID,	setuid_cancel,		NULL,		0,
		(void *)0 },
	{ MNTOPT_NOSETUID,	nosetuid_cancel,	NULL,		0,
		(void *)0 },
	{ MNTOPT_NBMAND,	nbmand_cancel,		NULL,		0,
		(void *)0 },
	{ MNTOPT_NONBMAND,	nonbmand_cancel,	NULL,		0,
		(void *)0 },
	{ MNTOPT_EXEC,		exec_cancel,		NULL,		0,
		(void *)0 },
	{ MNTOPT_NOEXEC,	noexec_cancel,		NULL,		0,
		(void *)0 },
};

const mntopts_t vfs_mntopts = {
	sizeof (mntopts) / sizeof (mntopt_t),
	(mntopt_t *)&mntopts[0]
};

/*
 * File system operation dispatch functions.
 */

int
fsop_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr)
{
	return (*(vfsp)->vfs_op->vfs_mount)(vfsp, mvp, uap, cr);
}

int
fsop_unmount(vfs_t *vfsp, int flag, cred_t *cr)
{
	return (*(vfsp)->vfs_op->vfs_unmount)(vfsp, flag, cr);
}

int
fsop_root(vfs_t *vfsp, vnode_t **vpp)
{
	refstr_t *mntpt;
	int ret = (*(vfsp)->vfs_op->vfs_root)(vfsp, vpp);
	/*
	 * Make sure this root has a path.  With lofs, it is possible to have
	 * a NULL mountpoint.
	 */
	if (ret == 0 && vfsp->vfs_mntpt != NULL && (*vpp)->v_path == NULL) {
		mntpt = vfs_getmntpoint(vfsp);
		vn_setpath_str(*vpp, refstr_value(mntpt),
		    strlen(refstr_value(mntpt)));
		refstr_rele(mntpt);
	}

	return (ret);
}

int
fsop_statfs(vfs_t *vfsp, statvfs64_t *sp)
{
	return (*(vfsp)->vfs_op->vfs_statvfs)(vfsp, sp);
}

int
fsop_sync(vfs_t *vfsp, short flag, cred_t *cr)
{
	return (*(vfsp)->vfs_op->vfs_sync)(vfsp, flag, cr);
}

int
fsop_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp)
{
	/*
	 * In order to handle system attribute fids in a manner
	 * transparent to the underlying fs, we embed the fid for
	 * the sysattr parent object in the sysattr fid and tack on
	 * some extra bytes that only the sysattr layer knows about.
	 *
	 * This guarantees that sysattr fids are larger than other fids
	 * for this vfs. If the vfs supports the sysattr view interface
	 * (as indicated by VFSFT_SYSATTR_VIEWS), we cannot have a size
	 * collision with XATTR_FIDSZ.
	 */
	if (vfs_has_feature(vfsp, VFSFT_SYSATTR_VIEWS) &&
	    fidp->fid_len == XATTR_FIDSZ)
		return (xattr_dir_vget(vfsp, vpp, fidp));

	return (*(vfsp)->vfs_op->vfs_vget)(vfsp, vpp, fidp);
}

int
fsop_mountroot(vfs_t *vfsp, enum whymountroot reason)
{
	return (*(vfsp)->vfs_op->vfs_mountroot)(vfsp, reason);
}

void
fsop_freefs(vfs_t *vfsp)
{
	(*(vfsp)->vfs_op->vfs_freevfs)(vfsp);
}

int
fsop_vnstate(vfs_t *vfsp, vnode_t *vp, vntrans_t nstate)
{
	return ((*(vfsp)->vfs_op->vfs_vnstate)(vfsp, vp, nstate));
}

int
fsop_sync_by_kind(int fstype, short flag, cred_t *cr)
{
	ASSERT((fstype >= 0) && (fstype < nfstype));

	if (ALLOCATED_VFSSW(&vfssw[fstype]) && VFS_INSTALLED(&vfssw[fstype]))
		return (*vfssw[fstype].vsw_vfsops.vfs_sync) (NULL, flag, cr);
	else
		return (ENOTSUP);
}

/*
 * File system initialization.  vfs_setfsops() must be called from a file
 * system's init routine.
 */

static int
fs_copyfsops(const fs_operation_def_t *template, vfsops_t *actual,
    int *unused_ops)
{
	static const fs_operation_trans_def_t vfs_ops_table[] = {
		VFSNAME_MOUNT, offsetof(vfsops_t, vfs_mount),
			fs_nosys, fs_nosys,

		VFSNAME_UNMOUNT, offsetof(vfsops_t, vfs_unmount),
			fs_nosys, fs_nosys,

		VFSNAME_ROOT, offsetof(vfsops_t, vfs_root),
			fs_nosys, fs_nosys,

		VFSNAME_STATVFS, offsetof(vfsops_t, vfs_statvfs),
			fs_nosys, fs_nosys,

		VFSNAME_SYNC, offsetof(vfsops_t, vfs_sync),
			(fs_generic_func_p) fs_sync,
			(fs_generic_func_p) fs_sync,	/* No errors allowed */

		VFSNAME_VGET, offsetof(vfsops_t, vfs_vget),
			fs_nosys, fs_nosys,

		VFSNAME_MOUNTROOT, offsetof(vfsops_t, vfs_mountroot),
			fs_nosys, fs_nosys,

		VFSNAME_FREEVFS, offsetof(vfsops_t, vfs_freevfs),
			(fs_generic_func_p)fs_freevfs,
			(fs_generic_func_p)fs_freevfs,	/* Shouldn't fail */

		VFSNAME_VNSTATE, offsetof(vfsops_t, vfs_vnstate),
			(fs_generic_func_p)fs_nosys,
			(fs_generic_func_p)fs_nosys,

		NULL, 0, NULL, NULL
	};

	return (fs_build_vector(actual, unused_ops, vfs_ops_table, template));
}

void
zfs_boot_init() {

	if (strcmp(rootfs.bo_fstype, MNTTYPE_ZFS) == 0)
		spa_boot_init();
}

int
vfs_setfsops(int fstype, const fs_operation_def_t *template, vfsops_t **actual)
{
	int error;
	int unused_ops;

	/*
	 * Verify that fstype refers to a valid fs.  Note that
	 * 0 is valid since it's used to set "stray" ops.
	 */
	if ((fstype < 0) || (fstype >= nfstype))
		return (EINVAL);

	if (!ALLOCATED_VFSSW(&vfssw[fstype]))
		return (EINVAL);

	/* Set up the operations vector. */

	error = fs_copyfsops(template, &vfssw[fstype].vsw_vfsops, &unused_ops);

	if (error != 0)
		return (error);

	vfssw[fstype].vsw_flag |= VSW_INSTALLED;

	if (actual != NULL)
		*actual = &vfssw[fstype].vsw_vfsops;

#if DEBUG
	if (unused_ops != 0)
		cmn_err(CE_WARN, "vfs_setfsops: %s: %d operations supplied "
		    "but not used", vfssw[fstype].vsw_name, unused_ops);
#endif

	return (0);
}

int
vfs_makefsops(const fs_operation_def_t *template, vfsops_t **actual)
{
	int error;
	int unused_ops;

	*actual = (vfsops_t *)kmem_alloc(sizeof (vfsops_t), KM_SLEEP);

	error = fs_copyfsops(template, *actual, &unused_ops);
	if (error != 0) {
		kmem_free(*actual, sizeof (vfsops_t));
		*actual = NULL;
		return (error);
	}

	return (0);
}

/*
 * Free a vfsops structure created as a result of vfs_makefsops().
 * NOTE: For a vfsops structure initialized by vfs_setfsops(), use
 * vfs_freevfsops_by_type().
 */
void
vfs_freevfsops(vfsops_t *vfsops)
{
	kmem_free(vfsops, sizeof (vfsops_t));
}

/*
 * Since the vfsops structure is part of the vfssw table and wasn't
 * really allocated, we're not really freeing anything.  We keep
 * the name for consistency with vfs_freevfsops().  We do, however,
 * need to take care of a little bookkeeping.
 * NOTE: For a vfsops structure created by vfs_setfsops(), use
 * vfs_freevfsops_by_type().
 */
int
vfs_freevfsops_by_type(int fstype)
{

	/* Verify that fstype refers to a loaded fs (and not fsid 0). */
	if ((fstype <= 0) || (fstype >= nfstype))
		return (EINVAL);

	WLOCK_VFSSW();
	if ((vfssw[fstype].vsw_flag & VSW_INSTALLED) == 0) {
		WUNLOCK_VFSSW();
		return (EINVAL);
	}

	vfssw[fstype].vsw_flag &= ~VSW_INSTALLED;
	WUNLOCK_VFSSW();

	return (0);
}

/* Support routines used to reference vfs_op */

/* Set the operations vector for a vfs */
void
vfs_setops(vfs_t *vfsp, vfsops_t *vfsops)
{
	vfsops_t	*op;

	ASSERT(vfsp != NULL);
	ASSERT(vfsops != NULL);

	op = vfsp->vfs_op;
	membar_consumer();
	if (vfsp->vfs_femhead == NULL &&
	    casptr(&vfsp->vfs_op, op, vfsops) == op) {
		return;
	}
	fsem_setvfsops(vfsp, vfsops);
}

/* Retrieve the operations vector for a vfs */
vfsops_t *
vfs_getops(vfs_t *vfsp)
{
	vfsops_t	*op;

	ASSERT(vfsp != NULL);

	op = vfsp->vfs_op;
	membar_consumer();
	if (vfsp->vfs_femhead == NULL && op == vfsp->vfs_op) {
		return (op);
	} else {
		return (fsem_getvfsops(vfsp));
	}
}

/*
 * Returns non-zero (1) if the vfsops matches that of the vfs.
 * Returns zero (0) if not.
 */
int
vfs_matchops(vfs_t *vfsp, vfsops_t *vfsops)
{
	return (vfs_getops(vfsp) == vfsops);
}

/*
 * Returns non-zero (1) if the file system has installed a non-default,
 * non-error vfs_sync routine.  Returns zero (0) otherwise.
 */
int
vfs_can_sync(vfs_t *vfsp)
{
	/* vfs_sync() routine is not the default/error function */
	return (vfs_getops(vfsp)->vfs_sync != fs_sync);
}

/*
 * Initialize a vfs structure.
 */
void
vfs_init(vfs_t *vfsp, vfsops_t *op, void *data)
{
	/* Other initialization has been moved to vfs_alloc() */
	vfsp->vfs_count = 0;
	vfsp->vfs_next = vfsp;
	vfsp->vfs_prev = vfsp;
	vfsp->vfs_zone_next = vfsp;
	vfsp->vfs_zone_prev = vfsp;
	vfsp->vfs_lofi_minor = 0;
	sema_init(&vfsp->vfs_reflock, 1, NULL, SEMA_DEFAULT, NULL);
	vfsimpl_setup(vfsp);
	vfsp->vfs_data = (data);
	vfs_setops((vfsp), (op));
}

/*
 * Allocate and initialize the vfs implementation private data
 * structure, vfs_impl_t.
 */
void
vfsimpl_setup(vfs_t *vfsp)
{
	int i;

	if (vfsp->vfs_implp != NULL) {
		return;
	}

	vfsp->vfs_implp = kmem_alloc(sizeof (vfs_impl_t), KM_SLEEP);
	/* Note that these are #define'd in vfs.h */
	vfsp->vfs_vskap = NULL;
	vfsp->vfs_fstypevsp = NULL;

	/* Set size of counted array, then zero the array */
	vfsp->vfs_featureset[0] = VFS_FEATURE_MAXSZ - 1;
	for (i = 1; i <  VFS_FEATURE_MAXSZ; i++) {
		vfsp->vfs_featureset[i] = 0;
	}
}

/*
 * Release the vfs_impl_t structure, if it exists. Some unbundled
 * filesystems may not use the newer version of vfs and thus
 * would not contain this implementation private data structure.
 */
void
vfsimpl_teardown(vfs_t *vfsp)
{
	vfs_impl_t	*vip = vfsp->vfs_implp;

	if (vip == NULL)
		return;

	kmem_free(vfsp->vfs_implp, sizeof (vfs_impl_t));
	vfsp->vfs_implp = NULL;
}

/*
 * VFS system calls: mount, umount, syssync, statfs, fstatfs, statvfs,
 * fstatvfs, and sysfs moved to common/syscall.
 */

/*
 * Update every mounted file system.  We call the vfs_sync operation of
 * each file system type, passing it a NULL vfsp to indicate that all
 * mounted file systems of that type should be updated.
 */
void
vfs_sync(int flag)
{
	struct vfssw *vswp;
	RLOCK_VFSSW();
	for (vswp = &vfssw[1]; vswp < &vfssw[nfstype]; vswp++) {
		if (ALLOCATED_VFSSW(vswp) && VFS_INSTALLED(vswp)) {
			vfs_refvfssw(vswp);
			RUNLOCK_VFSSW();
			(void) (*vswp->vsw_vfsops.vfs_sync)(NULL, flag,
			    CRED());
			vfs_unrefvfssw(vswp);
			RLOCK_VFSSW();
		}
	}
	RUNLOCK_VFSSW();
}

void
sync(void)
{
	vfs_sync(0);
}

/*
 * External routines.
 */

krwlock_t vfssw_lock;	/* lock accesses to vfssw */

/*
 * Lock for accessing the vfs linked list.  Initialized in vfs_mountroot(),
 * but otherwise should be accessed only via vfs_list_lock() and
 * vfs_list_unlock().  Also used to protect the timestamp for mods to the list.
 */
static krwlock_t vfslist;

/*
 * Mount devfs on /devices. This is done right after root is mounted
 * to provide device access support for the system
 */
static void
vfs_mountdevices(void)
{
	struct vfssw *vsw;
	struct vnode *mvp;
	struct mounta mounta = {	/* fake mounta for devfs_mount() */
		NULL,
		NULL,
		MS_SYSSPACE,
		NULL,
		NULL,
		0,
		NULL,
		0
	};

	/*
	 * _init devfs module to fill in the vfssw
	 */
	if (modload("fs", "devfs") == -1)
		panic("Cannot _init devfs module");

	/*
	 * Hold vfs
	 */
	RLOCK_VFSSW();
	vsw = vfs_getvfsswbyname("devfs");
	VFS_INIT(&devices, &vsw->vsw_vfsops, NULL);
	VFS_HOLD(&devices);

	/*
	 * Locate mount point
	 */
	if (lookupname("/devices", UIO_SYSSPACE, FOLLOW, NULLVPP, &mvp))
		panic("Cannot find /devices");

	/*
	 * Perform the mount of /devices
	 */
	if (VFS_MOUNT(&devices, mvp, &mounta, CRED()))
		panic("Cannot mount /devices");

	RUNLOCK_VFSSW();

	/*
	 * Set appropriate members and add to vfs list for mnttab display
	 */
	vfs_setresource(&devices, "/devices");
	vfs_setmntpoint(&devices, "/devices");

	/*
	 * Hold the root of /devices so it won't go away
	 */
	if (VFS_ROOT(&devices, &devicesdir))
		panic("vfs_mountdevices: not devices root");

	if (vfs_lock(&devices) != 0) {
		VN_RELE(devicesdir);
		cmn_err(CE_NOTE, "Cannot acquire vfs_lock of /devices");
		return;
	}

	if (vn_vfswlock(mvp) != 0) {
		vfs_unlock(&devices);
		VN_RELE(devicesdir);
		cmn_err(CE_NOTE, "Cannot acquire vfswlock of /devices");
		return;
	}

	vfs_add(mvp, &devices, 0);
	vn_vfsunlock(mvp);
	vfs_unlock(&devices);
	VN_RELE(devicesdir);
}

/*
 * mount the first instance of /dev  to root and remain mounted
 */
static void
vfs_mountdev1(void)
{
	struct vfssw *vsw;
	struct vnode *mvp;
	struct mounta mounta = {	/* fake mounta for sdev_mount() */
		NULL,
		NULL,
		MS_SYSSPACE | MS_OVERLAY,
		NULL,
		NULL,
		0,
		NULL,
		0
	};

	/*
	 * _init dev module to fill in the vfssw
	 */
	if (modload("fs", "dev") == -1)
		cmn_err(CE_PANIC, "Cannot _init dev module\n");

	/*
	 * Hold vfs
	 */
	RLOCK_VFSSW();
	vsw = vfs_getvfsswbyname("dev");
	VFS_INIT(&dev, &vsw->vsw_vfsops, NULL);
	VFS_HOLD(&dev);

	/*
	 * Locate mount point
	 */
	if (lookupname("/dev", UIO_SYSSPACE, FOLLOW, NULLVPP, &mvp))
		cmn_err(CE_PANIC, "Cannot find /dev\n");

	/*
	 * Perform the mount of /dev
	 */
	if (VFS_MOUNT(&dev, mvp, &mounta, CRED()))
		cmn_err(CE_PANIC, "Cannot mount /dev 1\n");

	RUNLOCK_VFSSW();

	/*
	 * Set appropriate members and add to vfs list for mnttab display
	 */
	vfs_setresource(&dev, "/dev");
	vfs_setmntpoint(&dev, "/dev");

	/*
	 * Hold the root of /dev so it won't go away
	 */
	if (VFS_ROOT(&dev, &devdir))
		cmn_err(CE_PANIC, "vfs_mountdev1: not dev root");

	if (vfs_lock(&dev) != 0) {
		VN_RELE(devdir);
		cmn_err(CE_NOTE, "Cannot acquire vfs_lock of /dev");
		return;
	}

	if (vn_vfswlock(mvp) != 0) {
		vfs_unlock(&dev);
		VN_RELE(devdir);
		cmn_err(CE_NOTE, "Cannot acquire vfswlock of /dev");
		return;
	}

	vfs_add(mvp, &dev, 0);
	vn_vfsunlock(mvp);
	vfs_unlock(&dev);
	VN_RELE(devdir);
}

/*
 * Mount required filesystem. This is done right after root is mounted.
 */
static void
vfs_mountfs(char *module, char *spec, char *path)
{
	struct vnode *mvp;
	struct mounta mounta;
	vfs_t *vfsp;

	mounta.flags = MS_SYSSPACE | MS_DATA;
	mounta.fstype = module;
	mounta.spec = spec;
	mounta.dir = path;
	if (lookupname(path, UIO_SYSSPACE, FOLLOW, NULLVPP, &mvp)) {
		cmn_err(CE_WARN, "Cannot find %s", path);
		return;
	}
	if (domount(NULL, &mounta, mvp, CRED(), &vfsp))
		cmn_err(CE_WARN, "Cannot mount %s", path);
	else
		VFS_RELE(vfsp);
	VN_RELE(mvp);
}

/*
 * vfs_mountroot is called by main() to mount the root filesystem.
 */
void
vfs_mountroot(void)
{
	struct vnode	*rvp = NULL;
	char		*path;
	size_t		plen;
	struct vfssw	*vswp;

	rw_init(&vfssw_lock, NULL, RW_DEFAULT, NULL);
	rw_init(&vfslist, NULL, RW_DEFAULT, NULL);

	/*
	 * Alloc the vfs hash bucket array and locks
	 */
	rvfs_list = kmem_zalloc(vfshsz * sizeof (rvfs_t), KM_SLEEP);

	/*
	 * Call machine-dependent routine "rootconf" to choose a root
	 * file system type.
	 */
	if (rootconf())
		panic("vfs_mountroot: cannot mount root");
	/*
	 * Get vnode for '/'.  Set up rootdir, u.u_rdir and u.u_cdir
	 * to point to it.  These are used by lookuppn() so that it
	 * knows where to start from ('/' or '.').
	 */
	vfs_setmntpoint(rootvfs, "/");
	if (VFS_ROOT(rootvfs, &rootdir))
		panic("vfs_mountroot: no root vnode");
	PTOU(curproc)->u_cdir = rootdir;
	VN_HOLD(PTOU(curproc)->u_cdir);
	PTOU(curproc)->u_rdir = NULL;

	/*
	 * Setup the global zone's rootvp, now that it exists.
	 */
	global_zone->zone_rootvp = rootdir;
	VN_HOLD(global_zone->zone_rootvp);

	/*
	 * Notify the module code that it can begin using the
	 * root filesystem instead of the boot program's services.
	 */
	modrootloaded = 1;

	/*
	 * Special handling for a ZFS root file system.
	 */
	zfs_boot_init();

	/*
	 * Set up mnttab information for root
	 */
	vfs_setresource(rootvfs, rootfs.bo_name);

	/*
	 * Notify cluster software that the root filesystem is available.
	 */
	clboot_mountroot();

	/* Now that we're all done with the root FS, set up its vopstats */
	if ((vswp = vfs_getvfsswbyvfsops(vfs_getops(rootvfs))) != NULL) {
		/* Set flag for statistics collection */
		if (vswp->vsw_flag & VSW_STATS) {
			initialize_vopstats(&rootvfs->vfs_vopstats);
			rootvfs->vfs_flag |= VFS_STATS;
			rootvfs->vfs_fstypevsp =
			    get_fstype_vopstats(rootvfs, vswp);
			rootvfs->vfs_vskap = get_vskstat_anchor(rootvfs);
		}
		vfs_unrefvfssw(vswp);
	}

	/*
	 * Mount /devices, /dev instance 1, /system/contract, /etc/mnttab,
	 * /etc/svc/volatile, /etc/dfs/sharetab, /system/object, and /proc.
	 */
	vfs_mountdevices();
	vfs_mountdev1();

	vfs_mountfs("ctfs", "ctfs", CTFS_ROOT);
	vfs_mountfs("proc", "/proc", "/proc");
	vfs_mountfs("mntfs", "/etc/mnttab", "/etc/mnttab");
	vfs_mountfs("tmpfs", "/etc/svc/volatile", "/etc/svc/volatile");
	vfs_mountfs("objfs", "objfs", OBJFS_ROOT);

	if (getzoneid() == GLOBAL_ZONEID) {
		vfs_mountfs("sharefs", "sharefs", "/etc/dfs/sharetab");
	}

#ifdef __sparc
	/*
	 * This bit of magic can go away when we convert sparc to
	 * the new boot architecture based on ramdisk.
	 *
	 * Booting off a mirrored root volume:
	 * At this point, we have booted and mounted root on a
	 * single component of the mirror.  Complete the boot
	 * by configuring SVM and converting the root to the
	 * dev_t of the mirrored root device.  This dev_t conversion
	 * only works because the underlying device doesn't change.
	 */
	if (root_is_svm) {
		if (svm_rootconf()) {
			panic("vfs_mountroot: cannot remount root");
		}

		/*
		 * mnttab should reflect the new root device
		 */
		vfs_lock_wait(rootvfs);
		vfs_setresource(rootvfs, rootfs.bo_name);
		vfs_unlock(rootvfs);
	}
#endif /* __sparc */

	/*
	 * Look up the root device via devfs so that a dv_node is
	 * created for it. The vnode is never VN_RELE()ed.
	 * We allocate more than MAXPATHLEN so that the
	 * buffer passed to i_ddi_prompath_to_devfspath() is
	 * exactly MAXPATHLEN (the function expects a buffer
	 * of that length).
	 */
	plen = strlen("/devices");
	path = kmem_alloc(plen + MAXPATHLEN, KM_SLEEP);
	(void) strcpy(path, "/devices");

	if (i_ddi_prompath_to_devfspath(rootfs.bo_name, path + plen)
	    != DDI_SUCCESS ||
	    lookupname(path, UIO_SYSSPACE, FOLLOW, NULLVPP, &rvp)) {

		/* NUL terminate in case "path" has garbage */
		path[plen + MAXPATHLEN - 1] = '\0';
#ifdef	DEBUG
		cmn_err(CE_WARN, "!Cannot lookup root device: %s", path);
#endif
	}
	kmem_free(path, plen + MAXPATHLEN);
	vfs_mnttabvp_setup();
}

/*
 * If remount failed and we're in a zone we need to check for the zone
 * root path and strip it before the call to vfs_setpath().
 *
 * If strpath doesn't begin with the zone_rootpath the original
 * strpath is returned unchanged.
 */
static const char *
stripzonepath(const char *strpath)
{
	char *str1, *str2;
	int i;
	zone_t *zonep = curproc->p_zone;

	if (zonep->zone_rootpath == NULL || strpath == NULL) {
		return (NULL);
	}

	/*
	 * we check for the end of the string at one past the
	 * current position because the zone_rootpath always
	 * ends with "/" but we don't want to strip that off.
	 */
	str1 = zonep->zone_rootpath;
	str2 = (char *)strpath;
	ASSERT(str1[0] != '\0');
	for (i = 0; str1[i + 1] != '\0'; i++) {
		if (str1[i] != str2[i])
			return ((char *)strpath);
	}
	return (&str2[i]);
}

/*
 * Check to see if our "block device" is actually a file.  If so,
 * automatically add a lofi device, and keep track of this fact.
 */
static int
lofi_add(const char *fsname, struct vfs *vfsp,
    mntopts_t *mntopts, struct mounta *uap)
{
	int fromspace = (uap->flags & MS_SYSSPACE) ?
	    UIO_SYSSPACE : UIO_USERSPACE;
	struct lofi_ioctl *li = NULL;
	struct vnode *vp = NULL;
	struct pathname	pn = { NULL };
	ldi_ident_t ldi_id;
	ldi_handle_t ldi_hdl;
	vfssw_t *vfssw;
	int minor;
	int err = 0;

	if (fsname == NULL ||
	    (vfssw = vfs_getvfssw(fsname)) == NULL)
		return (0);

	if (!(vfssw->vsw_flag & VSW_CANLOFI)) {
		vfs_unrefvfssw(vfssw);
		return (0);
	}

	vfs_unrefvfssw(vfssw);
	vfssw = NULL;

	if (pn_get(uap->spec, fromspace, &pn) != 0)
		return (0);

	if (lookupname(uap->spec, fromspace, FOLLOW, NULL, &vp) != 0)
		goto out;

	if (vp->v_type != VREG)
		goto out;

	/* OK, this is a lofi mount. */

	if ((uap->flags & (MS_REMOUNT|MS_GLOBAL)) ||
	    vfs_optionisset_nolock(mntopts, MNTOPT_SUID, NULL) ||
	    vfs_optionisset_nolock(mntopts, MNTOPT_SETUID, NULL) ||
	    vfs_optionisset_nolock(mntopts, MNTOPT_DEVICES, NULL)) {
		err = EINVAL;
		goto out;
	}

	ldi_id = ldi_ident_from_anon();
	li = kmem_zalloc(sizeof (*li), KM_SLEEP);
	(void) strlcpy(li->li_filename, pn.pn_path, MAXPATHLEN + 1);

	/*
	 * The lofi control node is currently exclusive-open.  We'd like
	 * to improve this, but in the meantime, we'll loop waiting for
	 * access.
	 */
	for (;;) {
		err = ldi_open_by_name("/dev/lofictl", FREAD | FWRITE | FEXCL,
		    kcred, &ldi_hdl, ldi_id);

		if (err != EBUSY)
			break;

		if ((err = delay_sig(hz / 8)) == EINTR)
			break;
	}

	if (err)
		goto out2;

	err = ldi_ioctl(ldi_hdl, LOFI_MAP_FILE, (intptr_t)li,
	    FREAD | FWRITE | FEXCL | FKIOCTL, kcred, &minor);

	(void) ldi_close(ldi_hdl, FREAD | FWRITE | FEXCL, kcred);

	if (!err)
		vfsp->vfs_lofi_minor = minor;

out2:
	ldi_ident_release(ldi_id);
out:
	if (li != NULL)
		kmem_free(li, sizeof (*li));
	if (vp != NULL)
		VN_RELE(vp);
	pn_free(&pn);
	return (err);
}

static void
lofi_remove(struct vfs *vfsp)
{
	struct lofi_ioctl *li = NULL;
	ldi_ident_t ldi_id;
	ldi_handle_t ldi_hdl;
	int err;

	if (vfsp->vfs_lofi_minor == 0)
		return;

	ldi_id = ldi_ident_from_anon();

	li = kmem_zalloc(sizeof (*li), KM_SLEEP);
	li->li_minor = vfsp->vfs_lofi_minor;
	li->li_cleanup = B_TRUE;

	do {
		err = ldi_open_by_name("/dev/lofictl", FREAD | FWRITE | FEXCL,
		    kcred, &ldi_hdl, ldi_id);
	} while (err == EBUSY);

	if (err)
		goto out;

	err = ldi_ioctl(ldi_hdl, LOFI_UNMAP_FILE_MINOR, (intptr_t)li,
	    FREAD | FWRITE | FEXCL | FKIOCTL, kcred,