Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/fs/proc/root.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | * | |
7 | * proc root directory handling functions | |
8 | */ | |
1da177e4 LT |
9 | #include <linux/errno.h> |
10 | #include <linux/time.h> | |
11 | #include <linux/proc_fs.h> | |
12 | #include <linux/stat.h> | |
1da177e4 | 13 | #include <linux/init.h> |
914e2637 | 14 | #include <linux/sched.h> |
03441a34 | 15 | #include <linux/sched/stat.h> |
1da177e4 LT |
16 | #include <linux/module.h> |
17 | #include <linux/bitops.h> | |
87a8ebd6 | 18 | #include <linux/user_namespace.h> |
66f592e2 | 19 | #include <linux/fs_context.h> |
f6c7a1f3 | 20 | #include <linux/mount.h> |
07543f5c | 21 | #include <linux/pid_namespace.h> |
66f592e2 | 22 | #include <linux/fs_parser.h> |
5b825c3a | 23 | #include <linux/cred.h> |
60a3c3a5 | 24 | #include <linux/magic.h> |
66f592e2 | 25 | #include <linux/slab.h> |
1da177e4 | 26 | |
fee781e6 AB |
27 | #include "internal.h" |
28 | ||
66f592e2 DH |
29 | struct proc_fs_context { |
30 | struct pid_namespace *pid_ns; | |
31 | unsigned int mask; | |
e61bb8b3 | 32 | enum proc_hidepid hidepid; |
66f592e2 | 33 | int gid; |
e61bb8b3 | 34 | enum proc_pidonly pidonly; |
97412950 VK |
35 | }; |
36 | ||
66f592e2 DH |
37 | enum proc_param { |
38 | Opt_gid, | |
39 | Opt_hidepid, | |
6814ef2d | 40 | Opt_subset, |
97412950 VK |
41 | }; |
42 | ||
d7167b14 | 43 | static const struct fs_parameter_spec proc_fs_parameters[] = { |
66f592e2 | 44 | fsparam_u32("gid", Opt_gid), |
1c6c4d11 | 45 | fsparam_string("hidepid", Opt_hidepid), |
6814ef2d | 46 | fsparam_string("subset", Opt_subset), |
66f592e2 DH |
47 | {} |
48 | }; | |
49 | ||
24a71ce5 AG |
50 | static inline int valid_hidepid(unsigned int value) |
51 | { | |
52 | return (value == HIDEPID_OFF || | |
53 | value == HIDEPID_NO_ACCESS || | |
54 | value == HIDEPID_INVISIBLE || | |
55 | value == HIDEPID_NOT_PTRACEABLE); | |
56 | } | |
57 | ||
1c6c4d11 AG |
58 | static int proc_parse_hidepid_param(struct fs_context *fc, struct fs_parameter *param) |
59 | { | |
60 | struct proc_fs_context *ctx = fc->fs_private; | |
61 | struct fs_parameter_spec hidepid_u32_spec = fsparam_u32("hidepid", Opt_hidepid); | |
62 | struct fs_parse_result result; | |
63 | int base = (unsigned long)hidepid_u32_spec.data; | |
64 | ||
65 | if (param->type != fs_value_is_string) | |
66 | return invalf(fc, "proc: unexpected type of hidepid value\n"); | |
67 | ||
68 | if (!kstrtouint(param->string, base, &result.uint_32)) { | |
69 | if (!valid_hidepid(result.uint_32)) | |
70 | return invalf(fc, "proc: unknown value of hidepid - %s\n", param->string); | |
71 | ctx->hidepid = result.uint_32; | |
72 | return 0; | |
73 | } | |
74 | ||
75 | if (!strcmp(param->string, "off")) | |
76 | ctx->hidepid = HIDEPID_OFF; | |
77 | else if (!strcmp(param->string, "noaccess")) | |
78 | ctx->hidepid = HIDEPID_NO_ACCESS; | |
79 | else if (!strcmp(param->string, "invisible")) | |
80 | ctx->hidepid = HIDEPID_INVISIBLE; | |
81 | else if (!strcmp(param->string, "ptraceable")) | |
82 | ctx->hidepid = HIDEPID_NOT_PTRACEABLE; | |
83 | else | |
84 | return invalf(fc, "proc: unknown value of hidepid - %s\n", param->string); | |
85 | ||
86 | return 0; | |
87 | } | |
88 | ||
6814ef2d AG |
89 | static int proc_parse_subset_param(struct fs_context *fc, char *value) |
90 | { | |
91 | struct proc_fs_context *ctx = fc->fs_private; | |
92 | ||
93 | while (value) { | |
94 | char *ptr = strchr(value, ','); | |
95 | ||
96 | if (ptr != NULL) | |
97 | *ptr++ = '\0'; | |
98 | ||
99 | if (*value != '\0') { | |
100 | if (!strcmp(value, "pid")) { | |
101 | ctx->pidonly = PROC_PIDONLY_ON; | |
102 | } else { | |
103 | return invalf(fc, "proc: unsupported subset option - %s\n", value); | |
104 | } | |
105 | } | |
106 | value = ptr; | |
107 | } | |
108 | ||
109 | return 0; | |
110 | } | |
111 | ||
66f592e2 | 112 | static int proc_parse_param(struct fs_context *fc, struct fs_parameter *param) |
97412950 | 113 | { |
66f592e2 DH |
114 | struct proc_fs_context *ctx = fc->fs_private; |
115 | struct fs_parse_result result; | |
116 | int opt; | |
117 | ||
d7167b14 | 118 | opt = fs_parse(fc, proc_fs_parameters, param, &result); |
66f592e2 DH |
119 | if (opt < 0) |
120 | return opt; | |
121 | ||
122 | switch (opt) { | |
123 | case Opt_gid: | |
124 | ctx->gid = result.uint_32; | |
125 | break; | |
126 | ||
127 | case Opt_hidepid: | |
1c6c4d11 AG |
128 | if (proc_parse_hidepid_param(fc, param)) |
129 | return -EINVAL; | |
66f592e2 DH |
130 | break; |
131 | ||
6814ef2d AG |
132 | case Opt_subset: |
133 | if (proc_parse_subset_param(fc, param->string) < 0) | |
134 | return -EINVAL; | |
135 | break; | |
136 | ||
66f592e2 DH |
137 | default: |
138 | return -EINVAL; | |
97412950 VK |
139 | } |
140 | ||
66f592e2 DH |
141 | ctx->mask |= 1 << opt; |
142 | return 0; | |
97412950 VK |
143 | } |
144 | ||
fa10fed3 | 145 | static void proc_apply_options(struct proc_fs_info *fs_info, |
66f592e2 | 146 | struct fs_context *fc, |
66f592e2 | 147 | struct user_namespace *user_ns) |
97412950 | 148 | { |
66f592e2 DH |
149 | struct proc_fs_context *ctx = fc->fs_private; |
150 | ||
151 | if (ctx->mask & (1 << Opt_gid)) | |
fa10fed3 | 152 | fs_info->pid_gid = make_kgid(user_ns, ctx->gid); |
66f592e2 | 153 | if (ctx->mask & (1 << Opt_hidepid)) |
fa10fed3 | 154 | fs_info->hide_pid = ctx->hidepid; |
6814ef2d AG |
155 | if (ctx->mask & (1 << Opt_subset)) |
156 | fs_info->pidonly = ctx->pidonly; | |
97412950 VK |
157 | } |
158 | ||
66f592e2 | 159 | static int proc_fill_super(struct super_block *s, struct fs_context *fc) |
60a3c3a5 | 160 | { |
fa10fed3 | 161 | struct proc_fs_context *ctx = fc->fs_private; |
60a3c3a5 | 162 | struct inode *root_inode; |
fa10fed3 | 163 | struct proc_fs_info *fs_info; |
60a3c3a5 DH |
164 | int ret; |
165 | ||
fa10fed3 AG |
166 | fs_info = kzalloc(sizeof(*fs_info), GFP_KERNEL); |
167 | if (!fs_info) | |
168 | return -ENOMEM; | |
169 | ||
170 | fs_info->pid_ns = get_pid_ns(ctx->pid_ns); | |
171 | proc_apply_options(fs_info, fc, current_user_ns()); | |
60a3c3a5 DH |
172 | |
173 | /* User space would break if executables or devices appear on proc */ | |
174 | s->s_iflags |= SB_I_USERNS_VISIBLE | SB_I_NOEXEC | SB_I_NODEV; | |
175 | s->s_flags |= SB_NODIRATIME | SB_NOSUID | SB_NOEXEC; | |
176 | s->s_blocksize = 1024; | |
177 | s->s_blocksize_bits = 10; | |
178 | s->s_magic = PROC_SUPER_MAGIC; | |
179 | s->s_op = &proc_sops; | |
180 | s->s_time_gran = 1; | |
fa10fed3 | 181 | s->s_fs_info = fs_info; |
60a3c3a5 DH |
182 | |
183 | /* | |
184 | * procfs isn't actually a stacking filesystem; however, there is | |
185 | * too much magic going on inside it to permit stacking things on | |
186 | * top of it | |
187 | */ | |
188 | s->s_stack_depth = FILESYSTEM_MAX_STACK_DEPTH; | |
fa10fed3 | 189 | |
60a3c3a5 DH |
190 | /* procfs dentries and inodes don't require IO to create */ |
191 | s->s_shrink.seeks = 0; | |
192 | ||
193 | pde_get(&proc_root); | |
194 | root_inode = proc_get_inode(s, &proc_root); | |
195 | if (!root_inode) { | |
196 | pr_err("proc_fill_super: get root inode failed\n"); | |
197 | return -ENOMEM; | |
198 | } | |
199 | ||
200 | s->s_root = d_make_root(root_inode); | |
201 | if (!s->s_root) { | |
202 | pr_err("proc_fill_super: allocate dentry failed\n"); | |
203 | return -ENOMEM; | |
204 | } | |
205 | ||
206 | ret = proc_setup_self(s); | |
207 | if (ret) { | |
208 | return ret; | |
209 | } | |
210 | return proc_setup_thread_self(s); | |
211 | } | |
212 | ||
66f592e2 | 213 | static int proc_reconfigure(struct fs_context *fc) |
97412950 | 214 | { |
66f592e2 | 215 | struct super_block *sb = fc->root->d_sb; |
fa10fed3 | 216 | struct proc_fs_info *fs_info = proc_sb_info(sb); |
02b9984d TT |
217 | |
218 | sync_filesystem(sb); | |
66f592e2 | 219 | |
fa10fed3 | 220 | proc_apply_options(fs_info, fc, current_user_ns()); |
66f592e2 | 221 | return 0; |
97412950 VK |
222 | } |
223 | ||
66f592e2 | 224 | static int proc_get_tree(struct fs_context *fc) |
1da177e4 | 225 | { |
fa10fed3 | 226 | return get_tree_nodev(fc, proc_fill_super); |
66f592e2 | 227 | } |
07543f5c | 228 | |
66f592e2 DH |
229 | static void proc_fs_context_free(struct fs_context *fc) |
230 | { | |
231 | struct proc_fs_context *ctx = fc->fs_private; | |
232 | ||
46cf047a | 233 | put_pid_ns(ctx->pid_ns); |
66f592e2 DH |
234 | kfree(ctx); |
235 | } | |
236 | ||
237 | static const struct fs_context_operations proc_fs_context_ops = { | |
238 | .free = proc_fs_context_free, | |
239 | .parse_param = proc_parse_param, | |
240 | .get_tree = proc_get_tree, | |
241 | .reconfigure = proc_reconfigure, | |
242 | }; | |
243 | ||
244 | static int proc_init_fs_context(struct fs_context *fc) | |
245 | { | |
246 | struct proc_fs_context *ctx; | |
247 | ||
248 | ctx = kzalloc(sizeof(struct proc_fs_context), GFP_KERNEL); | |
249 | if (!ctx) | |
250 | return -ENOMEM; | |
07543f5c | 251 | |
66f592e2 | 252 | ctx->pid_ns = get_pid_ns(task_active_pid_ns(current)); |
46cf047a AV |
253 | put_user_ns(fc->user_ns); |
254 | fc->user_ns = get_user_ns(ctx->pid_ns->user_ns); | |
66f592e2 DH |
255 | fc->fs_private = ctx; |
256 | fc->ops = &proc_fs_context_ops; | |
257 | return 0; | |
07543f5c PE |
258 | } |
259 | ||
260 | static void proc_kill_sb(struct super_block *sb) | |
261 | { | |
fa10fed3 | 262 | struct proc_fs_info *fs_info = proc_sb_info(sb); |
07543f5c | 263 | |
058f2e4d AG |
264 | if (!fs_info) { |
265 | kill_anon_super(sb); | |
266 | return; | |
267 | } | |
4fa3b1c4 | 268 | |
058f2e4d AG |
269 | dput(fs_info->proc_self); |
270 | dput(fs_info->proc_thread_self); | |
4fa3b1c4 | 271 | |
fa10fed3 AG |
272 | kill_anon_super(sb); |
273 | put_pid_ns(fs_info->pid_ns); | |
274 | kfree(fs_info); | |
1da177e4 LT |
275 | } |
276 | ||
c2319540 | 277 | static struct file_system_type proc_fs_type = { |
66f592e2 DH |
278 | .name = "proc", |
279 | .init_fs_context = proc_init_fs_context, | |
d7167b14 | 280 | .parameters = proc_fs_parameters, |
66f592e2 | 281 | .kill_sb = proc_kill_sb, |
0b3b094a | 282 | .fs_flags = FS_USERNS_MOUNT | FS_DISALLOW_NOTIFY_PERM, |
1da177e4 LT |
283 | }; |
284 | ||
1da177e4 LT |
285 | void __init proc_root_init(void) |
286 | { | |
195b8cf0 | 287 | proc_init_kmemcache(); |
1270dd8d | 288 | set_proc_pid_nlink(); |
e656d8a6 | 289 | proc_self_init(); |
0097875b | 290 | proc_thread_self_init(); |
155134fe | 291 | proc_symlink("mounts", NULL, "self/mounts"); |
457c4cbc EB |
292 | |
293 | proc_net_init(); | |
36a5aeb8 | 294 | proc_mkdir("fs", NULL); |
928b4d8c | 295 | proc_mkdir("driver", NULL); |
eb6d38d5 | 296 | proc_create_mount_point("fs/nfsd"); /* somewhere for the nfsd filesystem to be mounted */ |
1da177e4 LT |
297 | #if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE) |
298 | /* just give it a mountpoint */ | |
eb6d38d5 | 299 | proc_create_mount_point("openprom"); |
1da177e4 LT |
300 | #endif |
301 | proc_tty_init(); | |
9c37066d | 302 | proc_mkdir("bus", NULL); |
77b14db5 | 303 | proc_sys_init(); |
1539d584 | 304 | |
ed8fb78d AD |
305 | /* |
306 | * Last things last. It is not like userspace processes eager | |
307 | * to open /proc files exist at this point but register last | |
308 | * anyway. | |
309 | */ | |
1539d584 | 310 | register_filesystem(&proc_fs_type); |
1da177e4 LT |
311 | } |
312 | ||
b74d24f7 | 313 | static int proc_root_getattr(struct mnt_idmap *idmap, |
549c7297 | 314 | const struct path *path, struct kstat *stat, |
a528d35e | 315 | u32 request_mask, unsigned int query_flags) |
1da177e4 | 316 | { |
b74d24f7 | 317 | generic_fillattr(&nop_mnt_idmap, d_inode(path->dentry), stat); |
76b6159b AV |
318 | stat->nlink = proc_root.nlink + nr_processes(); |
319 | return 0; | |
320 | } | |
1da177e4 | 321 | |
00cd8dd3 | 322 | static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, unsigned int flags) |
76b6159b | 323 | { |
867aaccf | 324 | if (!proc_pid_lookup(dentry, flags)) |
1da177e4 | 325 | return NULL; |
66f592e2 | 326 | |
335eb531 | 327 | return proc_lookup(dir, dentry, flags); |
1da177e4 LT |
328 | } |
329 | ||
f0c3b509 | 330 | static int proc_root_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 331 | { |
f0c3b509 | 332 | if (ctx->pos < FIRST_PROCESS_ENTRY) { |
94fc5d9d RG |
333 | int error = proc_readdir(file, ctx); |
334 | if (unlikely(error <= 0)) | |
335 | return error; | |
f0c3b509 | 336 | ctx->pos = FIRST_PROCESS_ENTRY; |
1da177e4 | 337 | } |
1da177e4 | 338 | |
f0c3b509 | 339 | return proc_pid_readdir(file, ctx); |
1da177e4 LT |
340 | } |
341 | ||
342 | /* | |
343 | * The root /proc directory is special, as it has the | |
344 | * <pid> directories. Thus we don't use the generic | |
345 | * directory handling functions for that.. | |
346 | */ | |
00977a59 | 347 | static const struct file_operations proc_root_operations = { |
1da177e4 | 348 | .read = generic_read_dir, |
f50752ea AV |
349 | .iterate_shared = proc_root_readdir, |
350 | .llseek = generic_file_llseek, | |
1da177e4 LT |
351 | }; |
352 | ||
353 | /* | |
354 | * proc root can do almost nothing.. | |
355 | */ | |
c5ef1c42 | 356 | static const struct inode_operations proc_root_inode_operations = { |
1da177e4 | 357 | .lookup = proc_root_lookup, |
76b6159b | 358 | .getattr = proc_root_getattr, |
1da177e4 LT |
359 | }; |
360 | ||
361 | /* | |
362 | * This is the root "inode" in the /proc tree.. | |
363 | */ | |
364 | struct proc_dir_entry proc_root = { | |
365 | .low_ino = PROC_ROOT_INO, | |
366 | .namelen = 5, | |
1da177e4 LT |
367 | .mode = S_IFDIR | S_IRUGO | S_IXUGO, |
368 | .nlink = 2, | |
9cdd83e3 | 369 | .refcnt = REFCOUNT_INIT(1), |
1da177e4 | 370 | .proc_iops = &proc_root_inode_operations, |
d56c0d45 | 371 | .proc_dir_ops = &proc_root_operations, |
1da177e4 | 372 | .parent = &proc_root, |
4f113437 | 373 | .subdir = RB_ROOT, |
24074a35 | 374 | .name = "/proc", |
1da177e4 | 375 | }; |