pnode.c 9 KB
Newer Older
1 2 3 4 5 6 7 8
/*
 *  linux/fs/pnode.c
 *
 * (C) Copyright IBM Corporation 2005.
 *	Released under GPL v2.
 *	Author : Ram Pai (linuxram@us.ibm.com)
 *
 */
9
#include <linux/mnt_namespace.h>
10 11
#include <linux/mount.h>
#include <linux/fs.h>
12
#include <linux/nsproxy.h>
13
#include "internal.h"
14 15
#include "pnode.h"

Ram Pai's avatar
Ram Pai committed
16
/* return the next shared peer mount of @p */
17
static inline struct mount *next_peer(struct mount *p)
Ram Pai's avatar
Ram Pai committed
18
{
19
	return list_entry(p->mnt_share.next, struct mount, mnt_share);
Ram Pai's avatar
Ram Pai committed
20 21
}

22
static inline struct mount *first_slave(struct mount *p)
Ram Pai's avatar
Ram Pai committed
23
{
24
	return list_entry(p->mnt_slave_list.next, struct mount, mnt_slave);
Ram Pai's avatar
Ram Pai committed
25 26
}

27
static inline struct mount *next_slave(struct mount *p)
Ram Pai's avatar
Ram Pai committed
28
{
29
	return list_entry(p->mnt_slave.next, struct mount, mnt_slave);
Ram Pai's avatar
Ram Pai committed
30 31
}

32 33 34
static struct mount *get_peer_under_root(struct mount *mnt,
					 struct mnt_namespace *ns,
					 const struct path *root)
35
{
36
	struct mount *m = mnt;
37 38 39

	do {
		/* Check the namespace first for optimization */
Al Viro's avatar
Al Viro committed
40
		if (m->mnt_ns == ns && is_path_reachable(m, m->mnt.mnt_root, root))
41
			return m;
42

43
		m = next_peer(m);
44
	} while (m != mnt);
45 46 47 48 49 50 51 52 53 54

	return NULL;
}

/*
 * Get ID of closest dominating peer group having a representative
 * under the given root.
 *
 * Caller must hold namespace_sem
 */
55
int get_dominating_id(struct mount *mnt, const struct path *root)
56
{
57
	struct mount *m;
58

59
	for (m = mnt->mnt_master; m != NULL; m = m->mnt_master) {
Al Viro's avatar
Al Viro committed
60
		struct mount *d = get_peer_under_root(m, mnt->mnt_ns, root);
61
		if (d)
Al Viro's avatar
Al Viro committed
62
			return d->mnt_group_id;
63 64 65 66 67
	}

	return 0;
}

68
static int do_make_slave(struct mount *mnt)
Ram Pai's avatar
Ram Pai committed
69
{
70
	struct mount *peer_mnt = mnt, *master = mnt->mnt_master;
Al Viro's avatar
Al Viro committed
71
	struct mount *slave_mnt;
Ram Pai's avatar
Ram Pai committed
72 73 74

	/*
	 * slave 'mnt' to a peer mount that has the
75
	 * same root dentry. If none is available then
Ram Pai's avatar
Ram Pai committed
76 77
	 * slave it to anything that is available.
	 */
78
	while ((peer_mnt = next_peer(peer_mnt)) != mnt &&
79
	       peer_mnt->mnt.mnt_root != mnt->mnt.mnt_root) ;
Ram Pai's avatar
Ram Pai committed
80 81

	if (peer_mnt == mnt) {
82
		peer_mnt = next_peer(mnt);
Ram Pai's avatar
Ram Pai committed
83 84 85
		if (peer_mnt == mnt)
			peer_mnt = NULL;
	}
86
	if (IS_MNT_SHARED(mnt) && list_empty(&mnt->mnt_share))
87
		mnt_release_group_id(mnt);
88

89
	list_del_init(&mnt->mnt_share);
Al Viro's avatar
Al Viro committed
90
	mnt->mnt_group_id = 0;
Ram Pai's avatar
Ram Pai committed
91 92 93 94 95

	if (peer_mnt)
		master = peer_mnt;

	if (master) {
96
		list_for_each_entry(slave_mnt, &mnt->mnt_slave_list, mnt_slave)
97
			slave_mnt->mnt_master = master;
98 99 100
		list_move(&mnt->mnt_slave, &master->mnt_slave_list);
		list_splice(&mnt->mnt_slave_list, master->mnt_slave_list.prev);
		INIT_LIST_HEAD(&mnt->mnt_slave_list);
Ram Pai's avatar
Ram Pai committed
101
	} else {
102
		struct list_head *p = &mnt->mnt_slave_list;
Ram Pai's avatar
Ram Pai committed
103
		while (!list_empty(p)) {
104
                        slave_mnt = list_first_entry(p,
105 106
					struct mount, mnt_slave);
			list_del_init(&slave_mnt->mnt_slave);
Ram Pai's avatar
Ram Pai committed
107 108 109
			slave_mnt->mnt_master = NULL;
		}
	}
110
	mnt->mnt_master = master;
111
	CLEAR_MNT_SHARED(mnt);
Ram Pai's avatar
Ram Pai committed
112 113 114
	return 0;
}

Nick Piggin's avatar
Nick Piggin committed
115 116 117
/*
 * vfsmount lock must be held for write
 */
118
void change_mnt_propagation(struct mount *mnt, int type)
119
{
Ram Pai's avatar
Ram Pai committed
120
	if (type == MS_SHARED) {
121
		set_mnt_shared(mnt);
Ram Pai's avatar
Ram Pai committed
122 123
		return;
	}
124
	do_make_slave(mnt);
Ram Pai's avatar
Ram Pai committed
125
	if (type != MS_SLAVE) {
126
		list_del_init(&mnt->mnt_slave);
Al Viro's avatar
Al Viro committed
127
		mnt->mnt_master = NULL;
Ram Pai's avatar
Ram Pai committed
128
		if (type == MS_UNBINDABLE)
129
			mnt->mnt.mnt_flags |= MNT_UNBINDABLE;
Andries E. Brouwer's avatar
Andries E. Brouwer committed
130
		else
131
			mnt->mnt.mnt_flags &= ~MNT_UNBINDABLE;
Ram Pai's avatar
Ram Pai committed
132
	}
133
}
134 135 136 137 138

/*
 * get the next mount in the propagation tree.
 * @m: the mount seen last
 * @origin: the original mount from where the tree walk initiated
139 140 141 142 143
 *
 * Note that peer groups form contiguous segments of slave lists.
 * We rely on that in get_source() to be able to find out if
 * vfsmount found while iterating with propagation_next() is
 * a peer of one we'd found earlier.
144
 */
145 146
static struct mount *propagation_next(struct mount *m,
					 struct mount *origin)
147
{
Ram Pai's avatar
Ram Pai committed
148
	/* are there any slaves of this mount? */
Al Viro's avatar
Al Viro committed
149
	if (!IS_MNT_NEW(m) && !list_empty(&m->mnt_slave_list))
Ram Pai's avatar
Ram Pai committed
150 151 152
		return first_slave(m);

	while (1) {
153
		struct mount *master = m->mnt_master;
Ram Pai's avatar
Ram Pai committed
154

155
		if (master == origin->mnt_master) {
156 157
			struct mount *next = next_peer(m);
			return (next == origin) ? NULL : next;
158
		} else if (m->mnt_slave.next != &master->mnt_slave_list)
Ram Pai's avatar
Ram Pai committed
159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
			return next_slave(m);

		/* back at master */
		m = master;
	}
}

/*
 * return the source mount to be used for cloning
 *
 * @dest 	the current destination mount
 * @last_dest  	the last seen destination mount
 * @last_src  	the last seen source mount
 * @type	return CL_SLAVE if the new mount has to be
 * 		cloned as a slave.
 */
175 176 177 178
static struct mount *get_source(struct mount *dest,
				struct mount *last_dest,
				struct mount *last_src,
				int *type)
Ram Pai's avatar
Ram Pai committed
179
{
180 181
	struct mount *p_last_src = NULL;
	struct mount *p_last_dest = NULL;
Ram Pai's avatar
Ram Pai committed
182

183
	while (last_dest != dest->mnt_master) {
Ram Pai's avatar
Ram Pai committed
184 185
		p_last_dest = last_dest;
		p_last_src = last_src;
186 187
		last_dest = last_dest->mnt_master;
		last_src = last_src->mnt_master;
Ram Pai's avatar
Ram Pai committed
188 189 190 191 192
	}

	if (p_last_dest) {
		do {
			p_last_dest = next_peer(p_last_dest);
Al Viro's avatar
Al Viro committed
193
		} while (IS_MNT_NEW(p_last_dest));
194 195 196 197 198
		/* is that a peer of the earlier? */
		if (dest == p_last_dest) {
			*type = CL_MAKE_SHARED;
			return p_last_src;
		}
Ram Pai's avatar
Ram Pai committed
199
	}
200 201 202
	/* slave of the earlier, then */
	*type = CL_SLAVE;
	/* beginning of peer group among the slaves? */
203
	if (IS_MNT_SHARED(dest))
204 205
		*type |= CL_MAKE_SHARED;
	return last_src;
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
}

/*
 * mount 'source_mnt' under the destination 'dest_mnt' at
 * dentry 'dest_dentry'. And propagate that mount to
 * all the peer and slave mounts of 'dest_mnt'.
 * Link all the new mounts into a propagation tree headed at
 * source_mnt. Also link all the new mounts using ->mnt_list
 * headed at source_mnt's ->mnt_list
 *
 * @dest_mnt: destination mount.
 * @dest_dentry: destination dentry.
 * @source_mnt: source mount.
 * @tree_list : list of heads of trees to be attached.
 */
221
int propagate_mnt(struct mount *dest_mnt, struct mountpoint *dest_mp,
222
		    struct mount *source_mnt, struct list_head *tree_list)
223
{
224
	struct user_namespace *user_ns = current->nsproxy->mnt_ns->user_ns;
225
	struct mount *m, *child;
226
	int ret = 0;
227 228
	struct mount *prev_dest_mnt = dest_mnt;
	struct mount *prev_src_mnt  = source_mnt;
229 230
	LIST_HEAD(tmp_list);

231 232
	for (m = propagation_next(dest_mnt, dest_mnt); m;
			m = propagation_next(m, dest_mnt)) {
Ram Pai's avatar
Ram Pai committed
233
		int type;
234
		struct mount *source;
235

Al Viro's avatar
Al Viro committed
236
		if (IS_MNT_NEW(m))
237 238
			continue;

Ram Pai's avatar
Ram Pai committed
239
		source =  get_source(m, prev_dest_mnt, prev_src_mnt, &type);
240

241 242 243 244
		/* Notice when we are propagating across user namespaces */
		if (m->mnt_ns->user_ns != user_ns)
			type |= CL_UNPRIVILEGED;

245 246 247
		child = copy_tree(source, source->mnt.mnt_root, type);
		if (IS_ERR(child)) {
			ret = PTR_ERR(child);
248 249 250 251
			list_splice(tree_list, tmp_list.prev);
			goto out;
		}

252 253
		if (is_subdir(dest_mp->m_dentry, m->mnt.mnt_root)) {
			mnt_set_mountpoint(m, dest_mp, child);
254
			list_add_tail(&child->mnt_hash, tree_list);
255 256 257 258 259
		} else {
			/*
			 * This can happen if the parent mount was bind mounted
			 * on some subdirectory of a shared/slave mount.
			 */
260
			list_add_tail(&child->mnt_hash, &tmp_list);
261 262
		}
		prev_dest_mnt = m;
263
		prev_src_mnt  = child;
264 265
	}
out:
Andi Kleen's avatar
Andi Kleen committed
266
	br_write_lock(&vfsmount_lock);
267
	while (!list_empty(&tmp_list)) {
268
		child = list_first_entry(&tmp_list, struct mount, mnt_hash);
269
		umount_tree(child, 0);
270
	}
Andi Kleen's avatar
Andi Kleen committed
271
	br_write_unlock(&vfsmount_lock);
272 273
	return ret;
}
274 275 276 277

/*
 * return true if the refcount is greater than count
 */
278
static inline int do_refcount_check(struct mount *mnt, int count)
279
{
280
	int mycount = mnt_get_count(mnt) - mnt->mnt_ghosts;
281 282 283 284 285 286 287 288 289 290
	return (mycount > count);
}

/*
 * check if the mount 'mnt' can be unmounted successfully.
 * @mnt: the mount to be checked for unmount
 * NOTE: unmounting 'mnt' would naturally propagate to all
 * other mounts its parent propagates to.
 * Check if any of these mounts that **do not have submounts**
 * have more references than 'refcnt'. If so return busy.
Nick Piggin's avatar
Nick Piggin committed
291
 *
Nick Piggin's avatar
Nick Piggin committed
292
 * vfsmount lock must be held for write
293
 */
294
int propagate_mount_busy(struct mount *mnt, int refcnt)
295
{
296
	struct mount *m, *child;
297
	struct mount *parent = mnt->mnt_parent;
298 299
	int ret = 0;

300
	if (mnt == parent)
301 302 303 304 305 306 307
		return do_refcount_check(mnt, refcnt);

	/*
	 * quickly check if the current mount can be unmounted.
	 * If not, we don't have to go checking for all other
	 * mounts
	 */
308
	if (!list_empty(&mnt->mnt_mounts) || do_refcount_check(mnt, refcnt))
309 310
		return 1;

311 312 313
	for (m = propagation_next(parent, parent); m;
	     		m = propagation_next(m, parent)) {
		child = __lookup_mnt(&m->mnt, mnt->mnt_mountpoint, 0);
314
		if (child && list_empty(&child->mnt_mounts) &&
315
		    (ret = do_refcount_check(child, 1)))
316 317 318 319 320 321 322 323 324
			break;
	}
	return ret;
}

/*
 * NOTE: unmounting 'mnt' naturally propagates to all other mounts its
 * parent propagates to.
 */
325
static void __propagate_umount(struct mount *mnt)
326
{
327
	struct mount *parent = mnt->mnt_parent;
328
	struct mount *m;
329

330
	BUG_ON(parent == mnt);
331

332 333
	for (m = propagation_next(parent, parent); m;
			m = propagation_next(m, parent)) {
334

335
		struct mount *child = __lookup_mnt(&m->mnt,
336
					mnt->mnt_mountpoint, 0);
337 338 339 340
		/*
		 * umount the child only if the child has no
		 * other children
		 */
341
		if (child && list_empty(&child->mnt_mounts))
342
			list_move_tail(&child->mnt_hash, &mnt->mnt_hash);
343 344 345 346 347 348 349
	}
}

/*
 * collect all mounts that receive propagation from the mount in @list,
 * and return these additional mounts in the same list.
 * @list: the list of mounts to be unmounted.
Nick Piggin's avatar
Nick Piggin committed
350 351
 *
 * vfsmount lock must be held for write
352 353 354
 */
int propagate_umount(struct list_head *list)
{
355
	struct mount *mnt;
356

357
	list_for_each_entry(mnt, list, mnt_hash)
358 359 360
		__propagate_umount(mnt);
	return 0;
}