Commit | Line | Data |
---|---|---|
e149ed2b AV |
1 | #include <linux/mount.h> |
2 | #include <linux/file.h> | |
3 | #include <linux/fs.h> | |
4 | #include <linux/proc_ns.h> | |
5 | #include <linux/magic.h> | |
6 | #include <linux/ktime.h> | |
75509fd8 | 7 | #include <linux/seq_file.h> |
6786741d AV |
8 | #include <linux/user_namespace.h> |
9 | #include <linux/nsfs.h> | |
e149ed2b AV |
10 | |
11 | static struct vfsmount *nsfs_mnt; | |
12 | ||
6786741d AV |
13 | static long ns_ioctl(struct file *filp, unsigned int ioctl, |
14 | unsigned long arg); | |
e149ed2b AV |
15 | static const struct file_operations ns_file_operations = { |
16 | .llseek = no_llseek, | |
6786741d | 17 | .unlocked_ioctl = ns_ioctl, |
e149ed2b AV |
18 | }; |
19 | ||
20 | static char *ns_dname(struct dentry *dentry, char *buffer, int buflen) | |
21 | { | |
75c3cfa8 | 22 | struct inode *inode = d_inode(dentry); |
e149ed2b AV |
23 | const struct proc_ns_operations *ns_ops = dentry->d_fsdata; |
24 | ||
25 | return dynamic_dname(dentry, buffer, buflen, "%s:[%lu]", | |
26 | ns_ops->name, inode->i_ino); | |
27 | } | |
28 | ||
29 | static void ns_prune_dentry(struct dentry *dentry) | |
30 | { | |
75c3cfa8 | 31 | struct inode *inode = d_inode(dentry); |
e149ed2b AV |
32 | if (inode) { |
33 | struct ns_common *ns = inode->i_private; | |
34 | atomic_long_set(&ns->stashed, 0); | |
35 | } | |
36 | } | |
37 | ||
38 | const struct dentry_operations ns_dentry_operations = | |
39 | { | |
40 | .d_prune = ns_prune_dentry, | |
41 | .d_delete = always_delete_dentry, | |
42 | .d_dname = ns_dname, | |
43 | }; | |
44 | ||
45 | static void nsfs_evict(struct inode *inode) | |
46 | { | |
47 | struct ns_common *ns = inode->i_private; | |
48 | clear_inode(inode); | |
49 | ns->ops->put(ns); | |
50 | } | |
51 | ||
6786741d | 52 | static void *__ns_get_path(struct path *path, struct ns_common *ns) |
e149ed2b AV |
53 | { |
54 | struct vfsmount *mnt = mntget(nsfs_mnt); | |
55 | struct qstr qname = { .name = "", }; | |
56 | struct dentry *dentry; | |
57 | struct inode *inode; | |
e149ed2b AV |
58 | unsigned long d; |
59 | ||
e149ed2b AV |
60 | rcu_read_lock(); |
61 | d = atomic_long_read(&ns->stashed); | |
62 | if (!d) | |
63 | goto slow; | |
64 | dentry = (struct dentry *)d; | |
65 | if (!lockref_get_not_dead(&dentry->d_lockref)) | |
66 | goto slow; | |
67 | rcu_read_unlock(); | |
6786741d | 68 | ns->ops->put(ns); |
e149ed2b AV |
69 | got_it: |
70 | path->mnt = mnt; | |
71 | path->dentry = dentry; | |
72 | return NULL; | |
73 | slow: | |
74 | rcu_read_unlock(); | |
75 | inode = new_inode_pseudo(mnt->mnt_sb); | |
76 | if (!inode) { | |
6786741d | 77 | ns->ops->put(ns); |
e149ed2b AV |
78 | mntput(mnt); |
79 | return ERR_PTR(-ENOMEM); | |
80 | } | |
81 | inode->i_ino = ns->inum; | |
82 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
83 | inode->i_flags |= S_IMMUTABLE; | |
84 | inode->i_mode = S_IFREG | S_IRUGO; | |
85 | inode->i_fop = &ns_file_operations; | |
86 | inode->i_private = ns; | |
87 | ||
88 | dentry = d_alloc_pseudo(mnt->mnt_sb, &qname); | |
89 | if (!dentry) { | |
90 | iput(inode); | |
91 | mntput(mnt); | |
92 | return ERR_PTR(-ENOMEM); | |
93 | } | |
94 | d_instantiate(dentry, inode); | |
6786741d | 95 | dentry->d_fsdata = (void *)ns->ops; |
e149ed2b AV |
96 | d = atomic_long_cmpxchg(&ns->stashed, 0, (unsigned long)dentry); |
97 | if (d) { | |
98 | d_delete(dentry); /* make sure ->d_prune() does nothing */ | |
99 | dput(dentry); | |
6786741d | 100 | mntput(mnt); |
e149ed2b | 101 | cpu_relax(); |
6786741d | 102 | return ERR_PTR(-EAGAIN); |
e149ed2b AV |
103 | } |
104 | goto got_it; | |
105 | } | |
106 | ||
6786741d AV |
107 | void *ns_get_path(struct path *path, struct task_struct *task, |
108 | const struct proc_ns_operations *ns_ops) | |
109 | { | |
110 | struct ns_common *ns; | |
111 | void *ret; | |
112 | ||
113 | again: | |
114 | ns = ns_ops->get(task); | |
115 | if (!ns) | |
116 | return ERR_PTR(-ENOENT); | |
117 | ||
118 | ret = __ns_get_path(path, ns); | |
119 | if (IS_ERR(ret) && PTR_ERR(ret) == -EAGAIN) | |
120 | goto again; | |
121 | return ret; | |
122 | } | |
123 | ||
124 | static int open_related_ns(struct ns_common *ns, | |
125 | struct ns_common *(*get_ns)(struct ns_common *ns)) | |
126 | { | |
127 | struct path path = {}; | |
128 | struct file *f; | |
129 | void *err; | |
130 | int fd; | |
131 | ||
132 | fd = get_unused_fd_flags(O_CLOEXEC); | |
133 | if (fd < 0) | |
134 | return fd; | |
135 | ||
136 | while (1) { | |
137 | struct ns_common *relative; | |
138 | ||
139 | relative = get_ns(ns); | |
140 | if (IS_ERR(relative)) { | |
141 | put_unused_fd(fd); | |
142 | return PTR_ERR(relative); | |
143 | } | |
144 | ||
145 | err = __ns_get_path(&path, relative); | |
146 | if (IS_ERR(err) && PTR_ERR(err) == -EAGAIN) | |
147 | continue; | |
148 | break; | |
149 | } | |
150 | if (IS_ERR(err)) { | |
151 | put_unused_fd(fd); | |
152 | return PTR_ERR(err); | |
153 | } | |
154 | ||
155 | f = dentry_open(&path, O_RDONLY, current_cred()); | |
156 | path_put(&path); | |
157 | if (IS_ERR(f)) { | |
158 | put_unused_fd(fd); | |
159 | fd = PTR_ERR(f); | |
160 | } else | |
161 | fd_install(fd, f); | |
162 | ||
163 | return fd; | |
164 | } | |
165 | ||
166 | static long ns_ioctl(struct file *filp, unsigned int ioctl, | |
167 | unsigned long arg) | |
168 | { | |
169 | struct ns_common *ns = get_proc_ns(file_inode(filp)); | |
170 | ||
171 | switch (ioctl) { | |
172 | case NS_GET_USERNS: | |
173 | return open_related_ns(ns, ns_get_owner); | |
174 | default: | |
175 | return -ENOTTY; | |
176 | } | |
177 | } | |
178 | ||
e149ed2b AV |
179 | int ns_get_name(char *buf, size_t size, struct task_struct *task, |
180 | const struct proc_ns_operations *ns_ops) | |
181 | { | |
182 | struct ns_common *ns; | |
183 | int res = -ENOENT; | |
184 | ns = ns_ops->get(task); | |
185 | if (ns) { | |
186 | res = snprintf(buf, size, "%s:[%u]", ns_ops->name, ns->inum); | |
187 | ns_ops->put(ns); | |
188 | } | |
189 | return res; | |
190 | } | |
191 | ||
192 | struct file *proc_ns_fget(int fd) | |
193 | { | |
194 | struct file *file; | |
195 | ||
196 | file = fget(fd); | |
197 | if (!file) | |
198 | return ERR_PTR(-EBADF); | |
199 | ||
200 | if (file->f_op != &ns_file_operations) | |
201 | goto out_invalid; | |
202 | ||
203 | return file; | |
204 | ||
205 | out_invalid: | |
206 | fput(file); | |
207 | return ERR_PTR(-EINVAL); | |
208 | } | |
209 | ||
75509fd8 EB |
210 | static int nsfs_show_path(struct seq_file *seq, struct dentry *dentry) |
211 | { | |
212 | struct inode *inode = d_inode(dentry); | |
213 | const struct proc_ns_operations *ns_ops = dentry->d_fsdata; | |
214 | ||
6798a8ca JP |
215 | seq_printf(seq, "%s:[%lu]", ns_ops->name, inode->i_ino); |
216 | return 0; | |
75509fd8 EB |
217 | } |
218 | ||
e149ed2b AV |
219 | static const struct super_operations nsfs_ops = { |
220 | .statfs = simple_statfs, | |
221 | .evict_inode = nsfs_evict, | |
75509fd8 | 222 | .show_path = nsfs_show_path, |
e149ed2b AV |
223 | }; |
224 | static struct dentry *nsfs_mount(struct file_system_type *fs_type, | |
225 | int flags, const char *dev_name, void *data) | |
226 | { | |
227 | return mount_pseudo(fs_type, "nsfs:", &nsfs_ops, | |
228 | &ns_dentry_operations, NSFS_MAGIC); | |
229 | } | |
230 | static struct file_system_type nsfs = { | |
231 | .name = "nsfs", | |
232 | .mount = nsfs_mount, | |
233 | .kill_sb = kill_anon_super, | |
234 | }; | |
235 | ||
236 | void __init nsfs_init(void) | |
237 | { | |
238 | nsfs_mnt = kern_mount(&nsfs); | |
239 | if (IS_ERR(nsfs_mnt)) | |
240 | panic("can't set nsfs up\n"); | |
241 | nsfs_mnt->mnt_sb->s_flags &= ~MS_NOUSER; | |
242 | } |