xref: /openbmc/linux/fs/proc/root.c (revision 94cdda6b) 
1 /*
2  *  linux/fs/proc/root.c
3  *
4  *  Copyright (C) 1991, 1992 Linus Torvalds
5  *
6  *  proc root directory handling functions
7  */
8 
9 #include <asm/uaccess.h>
10 
11 #include <linux/errno.h>
12 #include <linux/time.h>
13 #include <linux/proc_fs.h>
14 #include <linux/stat.h>
15 #include <linux/init.h>
16 #include <linux/sched.h>
17 #include <linux/module.h>
18 #include <linux/bitops.h>
19 #include <linux/user_namespace.h>
20 #include <linux/mount.h>
21 #include <linux/pid_namespace.h>
22 #include <linux/parser.h>
23 
24 #include "internal.h"
25 
26 static int proc_test_super(struct super_block *sb, void *data)
27 {
28 	return sb->s_fs_info == data;
29 }
30 
31 static int proc_set_super(struct super_block *sb, void *data)
32 {
33 	int err = set_anon_super(sb, NULL);
34 	if (!err) {
35 		struct pid_namespace *ns = (struct pid_namespace *)data;
36 		sb->s_fs_info = get_pid_ns(ns);
37 	}
38 	return err;
39 }
40 
41 enum {
42 	Opt_gid, Opt_hidepid, Opt_err,
43 };
44 
45 static const match_table_t tokens = {
46 	{Opt_hidepid, "hidepid=%u"},
47 	{Opt_gid, "gid=%u"},
48 	{Opt_err, NULL},
49 };
50 
51 static int proc_parse_options(char *options, struct pid_namespace *pid)
52 {
53 	char *p;
54 	substring_t args[MAX_OPT_ARGS];
55 	int option;
56 
57 	if (!options)
58 		return 1;
59 
60 	while ((p = strsep(&options, ",")) != NULL) {
61 		int token;
62 		if (!*p)
63 			continue;
64 
65 		args[0].to = args[0].from = NULL;
66 		token = match_token(p, tokens, args);
67 		switch (token) {
68 		case Opt_gid:
69 			if (match_int(&args[0], &option))
70 				return 0;
71 			pid->pid_gid = make_kgid(current_user_ns(), option);
72 			break;
73 		case Opt_hidepid:
74 			if (match_int(&args[0], &option))
75 				return 0;
76 			if (option < 0 || option > 2) {
77 				pr_err("proc: hidepid value must be between 0 and 2.\n");
78 				return 0;
79 			}
80 			pid->hide_pid = option;
81 			break;
82 		default:
83 			pr_err("proc: unrecognized mount option \"%s\" "
84 			       "or missing value\n", p);
85 			return 0;
86 		}
87 	}
88 
89 	return 1;
90 }
91 
92 int proc_remount(struct super_block *sb, int *flags, char *data)
93 {
94 	struct pid_namespace *pid = sb->s_fs_info;
95 
96 	sync_filesystem(sb);
97 	return !proc_parse_options(data, pid);
98 }
99 
100 static struct dentry *proc_mount(struct file_system_type *fs_type,
101 	int flags, const char *dev_name, void *data)
102 {
103 	int err;
104 	struct super_block *sb;
105 	struct pid_namespace *ns;
106 	char *options;
107 
108 	if (flags & MS_KERNMOUNT) {
109 		ns = (struct pid_namespace *)data;
110 		options = NULL;
111 	} else {
112 		ns = task_active_pid_ns(current);
113 		options = data;
114 
115 		if (!capable(CAP_SYS_ADMIN) && !fs_fully_visible(fs_type))
116 			return ERR_PTR(-EPERM);
117 
118 		/* Does the mounter have privilege over the pid namespace? */
119 		if (!ns_capable(ns->user_ns, CAP_SYS_ADMIN))
120 			return ERR_PTR(-EPERM);
121 	}
122 
123 	sb = sget(fs_type, proc_test_super, proc_set_super, flags, ns);
124 	if (IS_ERR(sb))
125 		return ERR_CAST(sb);
126 
127 	if (!proc_parse_options(options, ns)) {
128 		deactivate_locked_super(sb);
129 		return ERR_PTR(-EINVAL);
130 	}
131 
132 	if (!sb->s_root) {
133 		err = proc_fill_super(sb);
134 		if (err) {
135 			deactivate_locked_super(sb);
136 			return ERR_PTR(err);
137 		}
138 
139 		sb->s_flags |= MS_ACTIVE;
140 	}
141 
142 	return dget(sb->s_root);
143 }
144 
145 static void proc_kill_sb(struct super_block *sb)
146 {
147 	struct pid_namespace *ns;
148 
149 	ns = (struct pid_namespace *)sb->s_fs_info;
150 	if (ns->proc_self)
151 		dput(ns->proc_self);
152 	if (ns->proc_thread_self)
153 		dput(ns->proc_thread_self);
154 	kill_anon_super(sb);
155 	put_pid_ns(ns);
156 }
157 
158 static struct file_system_type proc_fs_type = {
159 	.name		= "proc",
160 	.mount		= proc_mount,
161 	.kill_sb	= proc_kill_sb,
162 	.fs_flags	= FS_USERNS_MOUNT,
163 };
164 
165 void __init proc_root_init(void)
166 {
167 	int err;
168 
169 	proc_init_inodecache();
170 	err = register_filesystem(&proc_fs_type);
171 	if (err)
172 		return;
173 
174 	proc_self_init();
175 	proc_thread_self_init();
176 	proc_symlink("mounts", NULL, "self/mounts");
177 
178 	proc_net_init();
179 
180 #ifdef CONFIG_SYSVIPC
181 	proc_mkdir("sysvipc", NULL);
182 #endif
183 	proc_mkdir("fs", NULL);
184 	proc_mkdir("driver", NULL);
185 	proc_mkdir("fs/nfsd", NULL); /* somewhere for the nfsd filesystem to be mounted */
186 #if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE)
187 	/* just give it a mountpoint */
188 	proc_mkdir("openprom", NULL);
189 #endif
190 	proc_tty_init();
191 	proc_mkdir("bus", NULL);
192 	proc_sys_init();
193 }
194 
195 static int proc_root_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat
196 )
197 {
198 	generic_fillattr(d_inode(dentry), stat);
199 	stat->nlink = proc_root.nlink + nr_processes();
200 	return 0;
201 }
202 
203 static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, unsigned int flags)
204 {
205 	if (!proc_pid_lookup(dir, dentry, flags))
206 		return NULL;
207 
208 	return proc_lookup(dir, dentry, flags);
209 }
210 
211 static int proc_root_readdir(struct file *file, struct dir_context *ctx)
212 {
213 	if (ctx->pos < FIRST_PROCESS_ENTRY) {
214 		int error = proc_readdir(file, ctx);
215 		if (unlikely(error <= 0))
216 			return error;
217 		ctx->pos = FIRST_PROCESS_ENTRY;
218 	}
219 
220 	return proc_pid_readdir(file, ctx);
221 }
222 
223 /*
224  * The root /proc directory is special, as it has the
225  * <pid> directories. Thus we don't use the generic
226  * directory handling functions for that..
227  */
228 static const struct file_operations proc_root_operations = {
229 	.read		 = generic_read_dir,
230 	.iterate	 = proc_root_readdir,
231 	.llseek		= default_llseek,
232 };
233 
234 /*
235  * proc root can do almost nothing..
236  */
237 static const struct inode_operations proc_root_inode_operations = {
238 	.lookup		= proc_root_lookup,
239 	.getattr	= proc_root_getattr,
240 };
241 
242 /*
243  * This is the root "inode" in the /proc tree..
244  */
245 struct proc_dir_entry proc_root = {
246 	.low_ino	= PROC_ROOT_INO,
247 	.namelen	= 5,
248 	.mode		= S_IFDIR | S_IRUGO | S_IXUGO,
249 	.nlink		= 2,
250 	.count		= ATOMIC_INIT(1),
251 	.proc_iops	= &proc_root_inode_operations,
252 	.proc_fops	= &proc_root_operations,
253 	.parent		= &proc_root,
254 	.subdir		= RB_ROOT,
255 	.name		= "/proc",
256 };
257 
258 int pid_ns_prepare_proc(struct pid_namespace *ns)
259 {
260 	struct vfsmount *mnt;
261 
262 	mnt = kern_mount_data(&proc_fs_type, ns);
263 	if (IS_ERR(mnt))
264 		return PTR_ERR(mnt);
265 
266 	ns->proc_mnt = mnt;
267 	return 0;
268 }
269 
270 void pid_ns_release_proc(struct pid_namespace *ns)
271 {
272 	kern_unmount(ns->proc_mnt);
273 }
274