xref: /onnv-gate/usr/src/uts/common/fs/lofs/lofs_vfsops.c (revision 0:68f95e015346)
1*0Sstevel@tonic-gate /*
2*0Sstevel@tonic-gate  * CDDL HEADER START
3*0Sstevel@tonic-gate  *
4*0Sstevel@tonic-gate  * The contents of this file are subject to the terms of the
5*0Sstevel@tonic-gate  * Common Development and Distribution License, Version 1.0 only
6*0Sstevel@tonic-gate  * (the "License").  You may not use this file except in compliance
7*0Sstevel@tonic-gate  * with the License.
8*0Sstevel@tonic-gate  *
9*0Sstevel@tonic-gate  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10*0Sstevel@tonic-gate  * or http://www.opensolaris.org/os/licensing.
11*0Sstevel@tonic-gate  * See the License for the specific language governing permissions
12*0Sstevel@tonic-gate  * and limitations under the License.
13*0Sstevel@tonic-gate  *
14*0Sstevel@tonic-gate  * When distributing Covered Code, include this CDDL HEADER in each
15*0Sstevel@tonic-gate  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16*0Sstevel@tonic-gate  * If applicable, add the following below this CDDL HEADER, with the
17*0Sstevel@tonic-gate  * fields enclosed by brackets "[]" replaced with your own identifying
18*0Sstevel@tonic-gate  * information: Portions Copyright [yyyy] [name of copyright owner]
19*0Sstevel@tonic-gate  *
20*0Sstevel@tonic-gate  * CDDL HEADER END
21*0Sstevel@tonic-gate  */
22*0Sstevel@tonic-gate /*
23*0Sstevel@tonic-gate  * Copyright 2004 Sun Microsystems, Inc.  All rights reserved.
24*0Sstevel@tonic-gate  * Use is subject to license terms.
25*0Sstevel@tonic-gate  */
26*0Sstevel@tonic-gate 
27*0Sstevel@tonic-gate #pragma ident	"%Z%%M%	%I%	%E% SMI"
28*0Sstevel@tonic-gate 
29*0Sstevel@tonic-gate #include <sys/param.h>
30*0Sstevel@tonic-gate #include <sys/errno.h>
31*0Sstevel@tonic-gate #include <sys/vfs.h>
32*0Sstevel@tonic-gate #include <sys/vnode.h>
33*0Sstevel@tonic-gate #include <sys/uio.h>
34*0Sstevel@tonic-gate #include <sys/pathname.h>
35*0Sstevel@tonic-gate #include <sys/kmem.h>
36*0Sstevel@tonic-gate #include <sys/cred.h>
37*0Sstevel@tonic-gate #include <sys/statvfs.h>
38*0Sstevel@tonic-gate #include <sys/fs/lofs_info.h>
39*0Sstevel@tonic-gate #include <sys/fs/lofs_node.h>
40*0Sstevel@tonic-gate #include <sys/mount.h>
41*0Sstevel@tonic-gate #include <sys/mntent.h>
42*0Sstevel@tonic-gate #include <sys/mkdev.h>
43*0Sstevel@tonic-gate #include <sys/sysmacros.h>
44*0Sstevel@tonic-gate #include <sys/systm.h>
45*0Sstevel@tonic-gate #include <sys/cmn_err.h>
46*0Sstevel@tonic-gate #include <sys/policy.h>
47*0Sstevel@tonic-gate #include "fs/fs_subr.h"
48*0Sstevel@tonic-gate 
49*0Sstevel@tonic-gate /*
50*0Sstevel@tonic-gate  * This is the loadable module wrapper.
51*0Sstevel@tonic-gate  */
52*0Sstevel@tonic-gate #include <sys/modctl.h>
53*0Sstevel@tonic-gate 
54*0Sstevel@tonic-gate static mntopts_t lofs_mntopts;
55*0Sstevel@tonic-gate 
56*0Sstevel@tonic-gate static int lofsinit(int, char *);
57*0Sstevel@tonic-gate 
58*0Sstevel@tonic-gate static vfsdef_t vfw = {
59*0Sstevel@tonic-gate 	VFSDEF_VERSION,
60*0Sstevel@tonic-gate 	"lofs",
61*0Sstevel@tonic-gate 	lofsinit,
62*0Sstevel@tonic-gate 	VSW_HASPROTO,
63*0Sstevel@tonic-gate 	&lofs_mntopts
64*0Sstevel@tonic-gate };
65*0Sstevel@tonic-gate 
66*0Sstevel@tonic-gate /*
67*0Sstevel@tonic-gate  * Stuff needed to support "zonedevfs" mode.
68*0Sstevel@tonic-gate  */
69*0Sstevel@tonic-gate static major_t lofs_major;
70*0Sstevel@tonic-gate static minor_t lofs_minor;
71*0Sstevel@tonic-gate static kmutex_t lofs_minor_lock;
72*0Sstevel@tonic-gate 
73*0Sstevel@tonic-gate /*
74*0Sstevel@tonic-gate  * LOFS mount options table
75*0Sstevel@tonic-gate  */
76*0Sstevel@tonic-gate static char *xattr_cancel[] = { MNTOPT_NOXATTR, NULL };
77*0Sstevel@tonic-gate static char *noxattr_cancel[] = { MNTOPT_XATTR, NULL };
78*0Sstevel@tonic-gate static char *zonedevfs_cancel[] = { MNTOPT_LOFS_NOZONEDEVFS, NULL };
79*0Sstevel@tonic-gate static char *nozonedevfs_cancel[] = { MNTOPT_LOFS_ZONEDEVFS, NULL };
80*0Sstevel@tonic-gate static char *sub_cancel[] = { MNTOPT_LOFS_NOSUB, NULL };
81*0Sstevel@tonic-gate static char *nosub_cancel[] = { MNTOPT_LOFS_SUB, NULL };
82*0Sstevel@tonic-gate 
83*0Sstevel@tonic-gate static mntopt_t mntopts[] = {
84*0Sstevel@tonic-gate /*
85*0Sstevel@tonic-gate  *	option name		cancel option	default arg	flags
86*0Sstevel@tonic-gate  *		private data
87*0Sstevel@tonic-gate  */
88*0Sstevel@tonic-gate 	{ MNTOPT_XATTR,		xattr_cancel,	NULL,		0,
89*0Sstevel@tonic-gate 		(void *)0 },
90*0Sstevel@tonic-gate 	{ MNTOPT_NOXATTR,	noxattr_cancel,	NULL,		0,
91*0Sstevel@tonic-gate 		(void *)0 },
92*0Sstevel@tonic-gate 	{ MNTOPT_LOFS_ZONEDEVFS,	zonedevfs_cancel,	NULL,	0,
93*0Sstevel@tonic-gate 		(void *)0 },
94*0Sstevel@tonic-gate 	{ MNTOPT_LOFS_NOZONEDEVFS,	nozonedevfs_cancel,	NULL,	0,
95*0Sstevel@tonic-gate 		(void *)0 },
96*0Sstevel@tonic-gate 	{ MNTOPT_LOFS_SUB,	sub_cancel,	NULL,		0,
97*0Sstevel@tonic-gate 		(void *)0 },
98*0Sstevel@tonic-gate 	{ MNTOPT_LOFS_NOSUB,	nosub_cancel,	NULL,		0,
99*0Sstevel@tonic-gate 		(void *)0 },
100*0Sstevel@tonic-gate };
101*0Sstevel@tonic-gate 
102*0Sstevel@tonic-gate static mntopts_t lofs_mntopts = {
103*0Sstevel@tonic-gate 	sizeof (mntopts) / sizeof (mntopt_t),
104*0Sstevel@tonic-gate 	mntopts
105*0Sstevel@tonic-gate };
106*0Sstevel@tonic-gate 
107*0Sstevel@tonic-gate /*
108*0Sstevel@tonic-gate  * Module linkage information for the kernel.
109*0Sstevel@tonic-gate  */
110*0Sstevel@tonic-gate 
111*0Sstevel@tonic-gate static struct modlfs modlfs = {
112*0Sstevel@tonic-gate 	&mod_fsops, "filesystem for lofs", &vfw
113*0Sstevel@tonic-gate };
114*0Sstevel@tonic-gate 
115*0Sstevel@tonic-gate static struct modlinkage modlinkage = {
116*0Sstevel@tonic-gate 	MODREV_1, (void *)&modlfs, NULL
117*0Sstevel@tonic-gate };
118*0Sstevel@tonic-gate 
119*0Sstevel@tonic-gate /*
120*0Sstevel@tonic-gate  * This is the module initialization routine.
121*0Sstevel@tonic-gate  */
122*0Sstevel@tonic-gate int
123*0Sstevel@tonic-gate _init()
124*0Sstevel@tonic-gate {
125*0Sstevel@tonic-gate 	int status;
126*0Sstevel@tonic-gate 
127*0Sstevel@tonic-gate 	lofs_subrinit();
128*0Sstevel@tonic-gate 	status = mod_install(&modlinkage);
129*0Sstevel@tonic-gate 	if (status != 0) {
130*0Sstevel@tonic-gate 		/*
131*0Sstevel@tonic-gate 		 * Cleanup previously initialized work.
132*0Sstevel@tonic-gate 		 */
133*0Sstevel@tonic-gate 		lofs_subrfini();
134*0Sstevel@tonic-gate 	}
135*0Sstevel@tonic-gate 
136*0Sstevel@tonic-gate 	return (status);
137*0Sstevel@tonic-gate }
138*0Sstevel@tonic-gate 
139*0Sstevel@tonic-gate /*
140*0Sstevel@tonic-gate  * Don't allow the lofs module to be unloaded for now.
141*0Sstevel@tonic-gate  * There is a memory leak if it gets unloaded.
142*0Sstevel@tonic-gate  */
143*0Sstevel@tonic-gate int
144*0Sstevel@tonic-gate _fini()
145*0Sstevel@tonic-gate {
146*0Sstevel@tonic-gate 	return (EBUSY);
147*0Sstevel@tonic-gate }
148*0Sstevel@tonic-gate 
149*0Sstevel@tonic-gate int
150*0Sstevel@tonic-gate _info(struct modinfo *modinfop)
151*0Sstevel@tonic-gate {
152*0Sstevel@tonic-gate 	return (mod_info(&modlinkage, modinfop));
153*0Sstevel@tonic-gate }
154*0Sstevel@tonic-gate 
155*0Sstevel@tonic-gate 
156*0Sstevel@tonic-gate static int lofsfstype;
157*0Sstevel@tonic-gate vfsops_t *lo_vfsops;
158*0Sstevel@tonic-gate 
159*0Sstevel@tonic-gate /*
160*0Sstevel@tonic-gate  * lo mount vfsop
161*0Sstevel@tonic-gate  * Set up mount info record and attach it to vfs struct.
162*0Sstevel@tonic-gate  */
163*0Sstevel@tonic-gate /*ARGSUSED*/
164*0Sstevel@tonic-gate static int
165*0Sstevel@tonic-gate lo_mount(struct vfs *vfsp,
166*0Sstevel@tonic-gate 	struct vnode *vp,
167*0Sstevel@tonic-gate 	struct mounta *uap,
168*0Sstevel@tonic-gate 	struct cred *cr)
169*0Sstevel@tonic-gate {
170*0Sstevel@tonic-gate 	int error;
171*0Sstevel@tonic-gate 	struct vnode *srootvp = NULL;	/* the server's root */
172*0Sstevel@tonic-gate 	struct vnode *realrootvp;
173*0Sstevel@tonic-gate 	struct loinfo *li;
174*0Sstevel@tonic-gate 	int is_zonedevfs = 0;
175*0Sstevel@tonic-gate 	int nodev;
176*0Sstevel@tonic-gate 
177*0Sstevel@tonic-gate 	nodev = vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL);
178*0Sstevel@tonic-gate 
179*0Sstevel@tonic-gate 	if ((error = secpolicy_fs_mount(cr, vp, vfsp)) != 0)
180*0Sstevel@tonic-gate 		return (EPERM);
181*0Sstevel@tonic-gate 
182*0Sstevel@tonic-gate 	/*
183*0Sstevel@tonic-gate 	 * Loopback devices which get "nodevices" added can be done without
184*0Sstevel@tonic-gate 	 * "nodevices" set because we cannot import devices into a zone
185*0Sstevel@tonic-gate 	 * with loopback.  Note that we have all zone privileges when
186*0Sstevel@tonic-gate 	 * this happens; if not, we'd have gotten "nosuid".
187*0Sstevel@tonic-gate 	 */
188*0Sstevel@tonic-gate 	if (!nodev && vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL))
189*0Sstevel@tonic-gate 		vfs_setmntopt(vfsp, MNTOPT_DEVICES, NULL, VFS_NODISPLAY);
190*0Sstevel@tonic-gate 
191*0Sstevel@tonic-gate 	/*
192*0Sstevel@tonic-gate 	 * We must ensure that only the global zone applies the 'zonedevfs'
193*0Sstevel@tonic-gate 	 * option; we don't want non-global zones to be able to establish
194*0Sstevel@tonic-gate 	 * lofs mounts using the special dev_t we use to ensure that the
195*0Sstevel@tonic-gate 	 * contents of a zone's /dev cannot be victim to link(2) or rename(2).
196*0Sstevel@tonic-gate 	 * See below, where we set all of this up.
197*0Sstevel@tonic-gate 	 *
198*0Sstevel@tonic-gate 	 * Since this is more like a privilege check, we use crgetzoneid(cr)
199*0Sstevel@tonic-gate 	 * instead of getzoneid().
200*0Sstevel@tonic-gate 	 */
201*0Sstevel@tonic-gate 	is_zonedevfs = vfs_optionisset(vfsp, MNTOPT_LOFS_ZONEDEVFS, NULL);
202*0Sstevel@tonic-gate 	if (crgetzoneid(cr) != GLOBAL_ZONEID && is_zonedevfs)
203*0Sstevel@tonic-gate 		return (EPERM);
204*0Sstevel@tonic-gate 
205*0Sstevel@tonic-gate 	mutex_enter(&vp->v_lock);
206*0Sstevel@tonic-gate 	if (!(uap->flags & MS_OVERLAY) &&
207*0Sstevel@tonic-gate 		(vp->v_count != 1 || (vp->v_flag & VROOT))) {
208*0Sstevel@tonic-gate 		mutex_exit(&vp->v_lock);
209*0Sstevel@tonic-gate 		return (EBUSY);
210*0Sstevel@tonic-gate 	}
211*0Sstevel@tonic-gate 	mutex_exit(&vp->v_lock);
212*0Sstevel@tonic-gate 
213*0Sstevel@tonic-gate 	/*
214*0Sstevel@tonic-gate 	 * Find real root, and make vfs point to real vfs
215*0Sstevel@tonic-gate 	 */
216*0Sstevel@tonic-gate 	if (error = lookupname(uap->spec, (uap->flags & MS_SYSSPACE) ?
217*0Sstevel@tonic-gate 		UIO_SYSSPACE : UIO_USERSPACE, FOLLOW, NULLVPP,
218*0Sstevel@tonic-gate 	    &realrootvp))
219*0Sstevel@tonic-gate 		return (error);
220*0Sstevel@tonic-gate 
221*0Sstevel@tonic-gate 	/*
222*0Sstevel@tonic-gate 	 * realrootvp may be an AUTOFS node, in which case we
223*0Sstevel@tonic-gate 	 * perform a VOP_ACCESS() to trigger the mount of the
224*0Sstevel@tonic-gate 	 * intended filesystem, so we loopback mount the intended
225*0Sstevel@tonic-gate 	 * filesystem instead of the AUTOFS filesystem.
226*0Sstevel@tonic-gate 	 */
227*0Sstevel@tonic-gate 	(void) VOP_ACCESS(realrootvp, 0, 0, cr);
228*0Sstevel@tonic-gate 
229*0Sstevel@tonic-gate 	/*
230*0Sstevel@tonic-gate 	 * We're interested in the top most filesystem.
231*0Sstevel@tonic-gate 	 * This is specially important when uap->spec is a trigger
232*0Sstevel@tonic-gate 	 * AUTOFS node, since we're really interested in mounting the
233*0Sstevel@tonic-gate 	 * filesystem AUTOFS mounted as result of the VOP_ACCESS()
234*0Sstevel@tonic-gate 	 * call not the AUTOFS node itself.
235*0Sstevel@tonic-gate 	 */
236*0Sstevel@tonic-gate 	if (vn_mountedvfs(realrootvp) != NULL) {
237*0Sstevel@tonic-gate 		if (error = traverse(&realrootvp)) {
238*0Sstevel@tonic-gate 			VN_RELE(realrootvp);
239*0Sstevel@tonic-gate 			return (error);
240*0Sstevel@tonic-gate 		}
241*0Sstevel@tonic-gate 	}
242*0Sstevel@tonic-gate 
243*0Sstevel@tonic-gate 	/*
244*0Sstevel@tonic-gate 	 * Allocate a vfs info struct and attach it
245*0Sstevel@tonic-gate 	 */
246*0Sstevel@tonic-gate 	li = kmem_zalloc(sizeof (struct loinfo), KM_SLEEP);
247*0Sstevel@tonic-gate 	li->li_realvfs = realrootvp->v_vfsp;
248*0Sstevel@tonic-gate 	li->li_mountvfs = vfsp;
249*0Sstevel@tonic-gate 
250*0Sstevel@tonic-gate 	/*
251*0Sstevel@tonic-gate 	 * Set mount flags to be inherited by loopback vfs's
252*0Sstevel@tonic-gate 	 */
253*0Sstevel@tonic-gate 	if (vfs_optionisset(vfsp, MNTOPT_RO, NULL)) {
254*0Sstevel@tonic-gate 		li->li_mflag |= VFS_RDONLY;
255*0Sstevel@tonic-gate 	}
256*0Sstevel@tonic-gate 	if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) {
257*0Sstevel@tonic-gate 		li->li_mflag |= (VFS_NOSETUID|VFS_NODEVICES);
258*0Sstevel@tonic-gate 	}
259*0Sstevel@tonic-gate 	if (vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL)) {
260*0Sstevel@tonic-gate 		li->li_mflag |= VFS_NODEVICES;
261*0Sstevel@tonic-gate 	}
262*0Sstevel@tonic-gate 	if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) {
263*0Sstevel@tonic-gate 		li->li_mflag |= VFS_NOSETUID;
264*0Sstevel@tonic-gate 	}
265*0Sstevel@tonic-gate 	/*
266*0Sstevel@tonic-gate 	 * Permissive flags are added to the "deny" bitmap.
267*0Sstevel@tonic-gate 	 */
268*0Sstevel@tonic-gate 	if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) {
269*0Sstevel@tonic-gate 		li->li_dflag |= VFS_XATTR;
270*0Sstevel@tonic-gate 	}
271*0Sstevel@tonic-gate 	if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) {
272*0Sstevel@tonic-gate 		li->li_dflag |= VFS_NBMAND;
273*0Sstevel@tonic-gate 	}
274*0Sstevel@tonic-gate 
275*0Sstevel@tonic-gate 	/*
276*0Sstevel@tonic-gate 	 * Propagate inheritable mount flags from the real vfs.
277*0Sstevel@tonic-gate 	 */
278*0Sstevel@tonic-gate 	if ((li->li_realvfs->vfs_flag & VFS_RDONLY) &&
279*0Sstevel@tonic-gate 	    !vfs_optionisset(vfsp, MNTOPT_RO, NULL))
280*0Sstevel@tonic-gate 		vfs_setmntopt(vfsp, MNTOPT_RO, NULL,
281*0Sstevel@tonic-gate 		    VFS_NODISPLAY);
282*0Sstevel@tonic-gate 	if ((li->li_realvfs->vfs_flag & VFS_NOSETUID) &&
283*0Sstevel@tonic-gate 	    !vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL))
284*0Sstevel@tonic-gate 		vfs_setmntopt(vfsp, MNTOPT_NOSETUID, NULL,
285*0Sstevel@tonic-gate 		    VFS_NODISPLAY);
286*0Sstevel@tonic-gate 	if ((li->li_realvfs->vfs_flag & VFS_NODEVICES) &&
287*0Sstevel@tonic-gate 	    !vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL))
288*0Sstevel@tonic-gate 		vfs_setmntopt(vfsp, MNTOPT_NODEVICES, NULL,
289*0Sstevel@tonic-gate 		    VFS_NODISPLAY);
290*0Sstevel@tonic-gate 	/*
291*0Sstevel@tonic-gate 	 * Permissive flags such as VFS_XATTR, as opposed to restrictive flags
292*0Sstevel@tonic-gate 	 * such as VFS_RDONLY, are handled differently.  An explicit
293*0Sstevel@tonic-gate 	 * MNTOPT_NOXATTR should override the underlying filesystem's VFS_XATTR.
294*0Sstevel@tonic-gate 	 */
295*0Sstevel@tonic-gate 	if ((li->li_realvfs->vfs_flag & VFS_XATTR) &&
296*0Sstevel@tonic-gate 	    !vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL) &&
297*0Sstevel@tonic-gate 	    !vfs_optionisset(vfsp, MNTOPT_XATTR, NULL))
298*0Sstevel@tonic-gate 		vfs_setmntopt(vfsp, MNTOPT_XATTR, NULL,
299*0Sstevel@tonic-gate 		    VFS_NODISPLAY);
300*0Sstevel@tonic-gate 	if ((li->li_realvfs->vfs_flag & VFS_NBMAND) &&
301*0Sstevel@tonic-gate 	    !vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL) &&
302*0Sstevel@tonic-gate 	    !vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL))
303*0Sstevel@tonic-gate 		vfs_setmntopt(vfsp, MNTOPT_NBMAND, NULL,
304*0Sstevel@tonic-gate 		    VFS_NODISPLAY);
305*0Sstevel@tonic-gate 
306*0Sstevel@tonic-gate 	li->li_refct = 0;
307*0Sstevel@tonic-gate 	vfsp->vfs_data = (caddr_t)li;
308*0Sstevel@tonic-gate 	vfsp->vfs_bcount = 0;
309*0Sstevel@tonic-gate 	vfsp->vfs_fstype = lofsfstype;
310*0Sstevel@tonic-gate 	vfsp->vfs_bsize = li->li_realvfs->vfs_bsize;
311*0Sstevel@tonic-gate 
312*0Sstevel@tonic-gate 	/*
313*0Sstevel@tonic-gate 	 * Test to see if we need to be in "zone /dev" mode.  In zonedevfs
314*0Sstevel@tonic-gate 	 * mode, we pull a nasty trick; we make sure that the lofs dev_t does
315*0Sstevel@tonic-gate 	 * *not* reflect the underlying device, so that no renames or links
316*0Sstevel@tonic-gate 	 * can occur to or from the /dev hierarchy.
317*0Sstevel@tonic-gate 	 */
318*0Sstevel@tonic-gate 	if (is_zonedevfs) {
319*0Sstevel@tonic-gate 		dev_t dev;
320*0Sstevel@tonic-gate 
321*0Sstevel@tonic-gate 		mutex_enter(&lofs_minor_lock);
322*0Sstevel@tonic-gate 		do {
323*0Sstevel@tonic-gate 			lofs_minor = (lofs_minor + 1) & MAXMIN32;
324*0Sstevel@tonic-gate 			dev = makedevice(lofs_major, lofs_minor);
325*0Sstevel@tonic-gate 		} while (vfs_devismounted(dev));
326*0Sstevel@tonic-gate 		mutex_exit(&lofs_minor_lock);
327*0Sstevel@tonic-gate 
328*0Sstevel@tonic-gate 		vfsp->vfs_dev = dev;
329*0Sstevel@tonic-gate 		vfs_make_fsid(&vfsp->vfs_fsid, dev, lofsfstype);
330*0Sstevel@tonic-gate 
331*0Sstevel@tonic-gate 		li->li_flag |= LO_ZONEDEVFS;
332*0Sstevel@tonic-gate 	} else {
333*0Sstevel@tonic-gate 		vfsp->vfs_dev = li->li_realvfs->vfs_dev;
334*0Sstevel@tonic-gate 		vfsp->vfs_fsid.val[0] = li->li_realvfs->vfs_fsid.val[0];
335*0Sstevel@tonic-gate 		vfsp->vfs_fsid.val[1] = li->li_realvfs->vfs_fsid.val[1];
336*0Sstevel@tonic-gate 	}
337*0Sstevel@tonic-gate 
338*0Sstevel@tonic-gate 	if (vfs_optionisset(vfsp, MNTOPT_LOFS_NOSUB, NULL)) {
339*0Sstevel@tonic-gate 		li->li_flag |= LO_NOSUB;
340*0Sstevel@tonic-gate 	}
341*0Sstevel@tonic-gate 
342*0Sstevel@tonic-gate 	/*
343*0Sstevel@tonic-gate 	 * Setup the hashtable. If the root of this mount isn't a directory,
344*0Sstevel@tonic-gate 	 * there's no point in allocating a large hashtable. A table with one
345*0Sstevel@tonic-gate 	 * bucket is sufficient.
346*0Sstevel@tonic-gate 	 */
347*0Sstevel@tonic-gate 	if (realrootvp->v_type != VDIR)
348*0Sstevel@tonic-gate 		lsetup(li, 1);
349*0Sstevel@tonic-gate 	else
350*0Sstevel@tonic-gate 		lsetup(li, 0);
351*0Sstevel@tonic-gate 
352*0Sstevel@tonic-gate 	/*
353*0Sstevel@tonic-gate 	 * Make the root vnode
354*0Sstevel@tonic-gate 	 */
355*0Sstevel@tonic-gate 	srootvp = makelonode(realrootvp, li);
356*0Sstevel@tonic-gate 	srootvp->v_flag |= VROOT;
357*0Sstevel@tonic-gate 	li->li_rootvp = srootvp;
358*0Sstevel@tonic-gate 
359*0Sstevel@tonic-gate #ifdef LODEBUG
360*0Sstevel@tonic-gate 	lo_dprint(4, "lo_mount: vfs %p realvfs %p root %p realroot %p li %p\n",
361*0Sstevel@tonic-gate 	    vfsp, li->li_realvfs, srootvp, realrootvp, li);
362*0Sstevel@tonic-gate #endif
363*0Sstevel@tonic-gate 	return (0);
364*0Sstevel@tonic-gate }
365*0Sstevel@tonic-gate 
366*0Sstevel@tonic-gate /*
367*0Sstevel@tonic-gate  * Undo loopback mount
368*0Sstevel@tonic-gate  */
369*0Sstevel@tonic-gate static int
370*0Sstevel@tonic-gate lo_unmount(struct vfs *vfsp, int flag, struct cred *cr)
371*0Sstevel@tonic-gate {
372*0Sstevel@tonic-gate 	struct loinfo *li;
373*0Sstevel@tonic-gate 
374*0Sstevel@tonic-gate 	if (secpolicy_fs_unmount(cr, vfsp) != 0)
375*0Sstevel@tonic-gate 		return (EPERM);
376*0Sstevel@tonic-gate 
377*0Sstevel@tonic-gate 	/*
378*0Sstevel@tonic-gate 	 * Forced unmount is not supported by this file system
379*0Sstevel@tonic-gate 	 * and thus, ENOTSUP, is being returned.
380*0Sstevel@tonic-gate 	 */
381*0Sstevel@tonic-gate 	if (flag & MS_FORCE)
382*0Sstevel@tonic-gate 		return (ENOTSUP);
383*0Sstevel@tonic-gate 
384*0Sstevel@tonic-gate 	li = vtoli(vfsp);
385*0Sstevel@tonic-gate #ifdef LODEBUG
386*0Sstevel@tonic-gate 	lo_dprint(4, "lo_unmount(%p) li %p\n", vfsp, li);
387*0Sstevel@tonic-gate #endif
388*0Sstevel@tonic-gate 	if (li->li_refct != 1 || li->li_rootvp->v_count != 1) {
389*0Sstevel@tonic-gate #ifdef LODEBUG
390*0Sstevel@tonic-gate 		lo_dprint(4, "refct %d v_ct %d\n", li->li_refct,
391*0Sstevel@tonic-gate 		    li->li_rootvp->v_count);
392*0Sstevel@tonic-gate #endif
393*0Sstevel@tonic-gate 		return (EBUSY);
394*0Sstevel@tonic-gate 	}
395*0Sstevel@tonic-gate 	VN_RELE(li->li_rootvp);
396*0Sstevel@tonic-gate 	return (0);
397*0Sstevel@tonic-gate }
398*0Sstevel@tonic-gate 
399*0Sstevel@tonic-gate /*
400*0Sstevel@tonic-gate  * Find root of lofs mount.
401*0Sstevel@tonic-gate  */
402*0Sstevel@tonic-gate static int
403*0Sstevel@tonic-gate lo_root(struct vfs *vfsp, struct vnode **vpp)
404*0Sstevel@tonic-gate {
405*0Sstevel@tonic-gate 	*vpp = vtoli(vfsp)->li_rootvp;
406*0Sstevel@tonic-gate #ifdef LODEBUG
407*0Sstevel@tonic-gate 	lo_dprint(4, "lo_root(0x%p) = %p\n", vfsp, *vpp);
408*0Sstevel@tonic-gate #endif
409*0Sstevel@tonic-gate 	/*
410*0Sstevel@tonic-gate 	 * If the root of the filesystem is a special file, return the specvp
411*0Sstevel@tonic-gate 	 * version of the vnode. We don't save the specvp vnode in our
412*0Sstevel@tonic-gate 	 * hashtable since that's exclusively for lnodes.
413*0Sstevel@tonic-gate 	 */
414*0Sstevel@tonic-gate 	if (IS_DEVVP(*vpp)) {
415*0Sstevel@tonic-gate 		struct vnode *svp;
416*0Sstevel@tonic-gate 
417*0Sstevel@tonic-gate 		svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, kcred);
418*0Sstevel@tonic-gate 		if (svp == NULL)
419*0Sstevel@tonic-gate 			return (ENOSYS);
420*0Sstevel@tonic-gate 		*vpp = svp;
421*0Sstevel@tonic-gate 	} else {
422*0Sstevel@tonic-gate 		VN_HOLD(*vpp);
423*0Sstevel@tonic-gate 	}
424*0Sstevel@tonic-gate 
425*0Sstevel@tonic-gate 	return (0);
426*0Sstevel@tonic-gate }
427*0Sstevel@tonic-gate 
428*0Sstevel@tonic-gate /*
429*0Sstevel@tonic-gate  * Get file system statistics.
430*0Sstevel@tonic-gate  */
431*0Sstevel@tonic-gate static int
432*0Sstevel@tonic-gate lo_statvfs(register struct vfs *vfsp, struct statvfs64 *sbp)
433*0Sstevel@tonic-gate {
434*0Sstevel@tonic-gate 	vnode_t *realrootvp;
435*0Sstevel@tonic-gate 
436*0Sstevel@tonic-gate #ifdef LODEBUG
437*0Sstevel@tonic-gate 	lo_dprint(4, "lostatvfs %p\n", vfsp);
438*0Sstevel@tonic-gate #endif
439*0Sstevel@tonic-gate 	/*
440*0Sstevel@tonic-gate 	 * Using realrootvp->v_vfsp (instead of the realvfsp that was
441*0Sstevel@tonic-gate 	 * cached) is necessary to make lofs work woth forced UFS unmounts.
442*0Sstevel@tonic-gate 	 * In the case of a forced unmount, UFS stores a set of dummy vfsops
443*0Sstevel@tonic-gate 	 * in all the (i)vnodes in the filesystem. The dummy ops simply
444*0Sstevel@tonic-gate 	 * returns back EIO.
445*0Sstevel@tonic-gate 	 */
446*0Sstevel@tonic-gate 	(void) lo_realvfs(vfsp, &realrootvp);
447*0Sstevel@tonic-gate 	if (realrootvp != NULL)
448*0Sstevel@tonic-gate 		return (VFS_STATVFS(realrootvp->v_vfsp, sbp));
449*0Sstevel@tonic-gate 	else
450*0Sstevel@tonic-gate 		return (EIO);
451*0Sstevel@tonic-gate }
452*0Sstevel@tonic-gate 
453*0Sstevel@tonic-gate /*
454*0Sstevel@tonic-gate  * LOFS doesn't have any data or metadata to flush, pending I/O on the
455*0Sstevel@tonic-gate  * underlying filesystem will be flushed when such filesystem is synched.
456*0Sstevel@tonic-gate  */
457*0Sstevel@tonic-gate /* ARGSUSED */
458*0Sstevel@tonic-gate static int
459*0Sstevel@tonic-gate lo_sync(struct vfs *vfsp,
460*0Sstevel@tonic-gate 	short flag,
461*0Sstevel@tonic-gate 	struct cred *cr)
462*0Sstevel@tonic-gate {
463*0Sstevel@tonic-gate #ifdef LODEBUG
464*0Sstevel@tonic-gate 	lo_dprint(4, "lo_sync: %p\n", vfsp);
465*0Sstevel@tonic-gate #endif
466*0Sstevel@tonic-gate 	return (0);
467*0Sstevel@tonic-gate }
468*0Sstevel@tonic-gate 
469*0Sstevel@tonic-gate /*
470*0Sstevel@tonic-gate  * Obtain the vnode from the underlying filesystem.
471*0Sstevel@tonic-gate  */
472*0Sstevel@tonic-gate static int
473*0Sstevel@tonic-gate lo_vget(struct vfs *vfsp, struct vnode **vpp, struct fid *fidp)
474*0Sstevel@tonic-gate {
475*0Sstevel@tonic-gate 	vnode_t *realrootvp;
476*0Sstevel@tonic-gate 
477*0Sstevel@tonic-gate #ifdef LODEBUG
478*0Sstevel@tonic-gate 	lo_dprint(4, "lo_vget: %p\n", vfsp);
479*0Sstevel@tonic-gate #endif
480*0Sstevel@tonic-gate 	(void) lo_realvfs(vfsp, &realrootvp);
481*0Sstevel@tonic-gate 	if (realrootvp != NULL)
482*0Sstevel@tonic-gate 		return (VFS_VGET(realrootvp->v_vfsp, vpp, fidp));
483*0Sstevel@tonic-gate 	else
484*0Sstevel@tonic-gate 		return (EIO);
485*0Sstevel@tonic-gate }
486*0Sstevel@tonic-gate 
487*0Sstevel@tonic-gate /*
488*0Sstevel@tonic-gate  * Free mount-specific data.
489*0Sstevel@tonic-gate  */
490*0Sstevel@tonic-gate static void
491*0Sstevel@tonic-gate lo_freevfs(struct vfs *vfsp)
492*0Sstevel@tonic-gate {
493*0Sstevel@tonic-gate 	struct loinfo *li = vtoli(vfsp);
494*0Sstevel@tonic-gate 
495*0Sstevel@tonic-gate 	ldestroy(li);
496*0Sstevel@tonic-gate 	kmem_free(li, sizeof (struct loinfo));
497*0Sstevel@tonic-gate }
498*0Sstevel@tonic-gate 
499*0Sstevel@tonic-gate static int
500*0Sstevel@tonic-gate lofsinit(int fstyp, char *name)
501*0Sstevel@tonic-gate {
502*0Sstevel@tonic-gate 	static const fs_operation_def_t lo_vfsops_template[] = {
503*0Sstevel@tonic-gate 		VFSNAME_MOUNT, lo_mount,
504*0Sstevel@tonic-gate 		VFSNAME_UNMOUNT, lo_unmount,
505*0Sstevel@tonic-gate 		VFSNAME_ROOT, lo_root,
506*0Sstevel@tonic-gate 		VFSNAME_STATVFS, lo_statvfs,
507*0Sstevel@tonic-gate 		VFSNAME_SYNC, (fs_generic_func_p) lo_sync,
508*0Sstevel@tonic-gate 		VFSNAME_VGET, lo_vget,
509*0Sstevel@tonic-gate 		VFSNAME_FREEVFS, (fs_generic_func_p) lo_freevfs,
510*0Sstevel@tonic-gate 		NULL, NULL
511*0Sstevel@tonic-gate 	};
512*0Sstevel@tonic-gate 	int error;
513*0Sstevel@tonic-gate 
514*0Sstevel@tonic-gate 	error = vfs_setfsops(fstyp, lo_vfsops_template, &lo_vfsops);
515*0Sstevel@tonic-gate 	if (error != 0) {
516*0Sstevel@tonic-gate 		cmn_err(CE_WARN, "lofsinit: bad vfs ops template");
517*0Sstevel@tonic-gate 		return (error);
518*0Sstevel@tonic-gate 	}
519*0Sstevel@tonic-gate 
520*0Sstevel@tonic-gate 	error = vn_make_ops(name, lo_vnodeops_template, &lo_vnodeops);
521*0Sstevel@tonic-gate 	if (error != 0) {
522*0Sstevel@tonic-gate 		(void) vfs_freevfsops_by_type(fstyp);
523*0Sstevel@tonic-gate 		cmn_err(CE_WARN, "lofsinit: bad vnode ops template");
524*0Sstevel@tonic-gate 		return (error);
525*0Sstevel@tonic-gate 	}
526*0Sstevel@tonic-gate 
527*0Sstevel@tonic-gate 	lofsfstype = fstyp;
528*0Sstevel@tonic-gate 
529*0Sstevel@tonic-gate 	if ((lofs_major = getudev()) == (major_t)-1) {
530*0Sstevel@tonic-gate 		(void) vfs_freevfsops_by_type(fstyp);
531*0Sstevel@tonic-gate 		cmn_err(CE_WARN, "lofsinit: Can't get unique device number.");
532*0Sstevel@tonic-gate 		return (ENXIO);
533*0Sstevel@tonic-gate 	}
534*0Sstevel@tonic-gate 
535*0Sstevel@tonic-gate 	lofs_minor = 0;
536*0Sstevel@tonic-gate 	mutex_init(&lofs_minor_lock, NULL, MUTEX_DEFAULT, NULL);
537*0Sstevel@tonic-gate 
538*0Sstevel@tonic-gate 	return (0);
539*0Sstevel@tonic-gate }
540