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 | */ | |
9 | ||
7c0f6ba6 | 10 | #include <linux/uaccess.h> |
1da177e4 LT |
11 | |
12 | #include <linux/errno.h> | |
13 | #include <linux/time.h> | |
14 | #include <linux/proc_fs.h> | |
15 | #include <linux/stat.h> | |
1da177e4 | 16 | #include <linux/init.h> |
914e2637 | 17 | #include <linux/sched.h> |
03441a34 | 18 | #include <linux/sched/stat.h> |
1da177e4 LT |
19 | #include <linux/module.h> |
20 | #include <linux/bitops.h> | |
87a8ebd6 | 21 | #include <linux/user_namespace.h> |
66f592e2 | 22 | #include <linux/fs_context.h> |
f6c7a1f3 | 23 | #include <linux/mount.h> |
07543f5c | 24 | #include <linux/pid_namespace.h> |
66f592e2 | 25 | #include <linux/fs_parser.h> |
5b825c3a | 26 | #include <linux/cred.h> |
60a3c3a5 | 27 | #include <linux/magic.h> |
66f592e2 | 28 | #include <linux/slab.h> |
1da177e4 | 29 | |
fee781e6 AB |
30 | #include "internal.h" |
31 | ||
66f592e2 DH |
32 | struct proc_fs_context { |
33 | struct pid_namespace *pid_ns; | |
34 | unsigned int mask; | |
e61bb8b3 | 35 | enum proc_hidepid hidepid; |
66f592e2 | 36 | int gid; |
e61bb8b3 | 37 | enum proc_pidonly pidonly; |
97412950 VK |
38 | }; |
39 | ||
66f592e2 DH |
40 | enum proc_param { |
41 | Opt_gid, | |
42 | Opt_hidepid, | |
6814ef2d | 43 | Opt_subset, |
97412950 VK |
44 | }; |
45 | ||
d7167b14 | 46 | static const struct fs_parameter_spec proc_fs_parameters[] = { |
66f592e2 | 47 | fsparam_u32("gid", Opt_gid), |
1c6c4d11 | 48 | fsparam_string("hidepid", Opt_hidepid), |
6814ef2d | 49 | fsparam_string("subset", Opt_subset), |
66f592e2 DH |
50 | {} |
51 | }; | |
52 | ||
24a71ce5 AG |
53 | static inline int valid_hidepid(unsigned int value) |
54 | { | |
55 | return (value == HIDEPID_OFF || | |
56 | value == HIDEPID_NO_ACCESS || | |
57 | value == HIDEPID_INVISIBLE || | |
58 | value == HIDEPID_NOT_PTRACEABLE); | |
59 | } | |
60 | ||
1c6c4d11 AG |
61 | static int proc_parse_hidepid_param(struct fs_context *fc, struct fs_parameter *param) |
62 | { | |
63 | struct proc_fs_context *ctx = fc->fs_private; | |
64 | struct fs_parameter_spec hidepid_u32_spec = fsparam_u32("hidepid", Opt_hidepid); | |
65 | struct fs_parse_result result; | |
66 | int base = (unsigned long)hidepid_u32_spec.data; | |
67 | ||
68 | if (param->type != fs_value_is_string) | |
69 | return invalf(fc, "proc: unexpected type of hidepid value\n"); | |
70 | ||
71 | if (!kstrtouint(param->string, base, &result.uint_32)) { | |
72 | if (!valid_hidepid(result.uint_32)) | |
73 | return invalf(fc, "proc: unknown value of hidepid - %s\n", param->string); | |
74 | ctx->hidepid = result.uint_32; | |
75 | return 0; | |
76 | } | |
77 | ||
78 | if (!strcmp(param->string, "off")) | |
79 | ctx->hidepid = HIDEPID_OFF; | |
80 | else if (!strcmp(param->string, "noaccess")) | |
81 | ctx->hidepid = HIDEPID_NO_ACCESS; | |
82 | else if (!strcmp(param->string, "invisible")) | |
83 | ctx->hidepid = HIDEPID_INVISIBLE; | |
84 | else if (!strcmp(param->string, "ptraceable")) | |
85 | ctx->hidepid = HIDEPID_NOT_PTRACEABLE; | |
86 | else | |
87 | return invalf(fc, "proc: unknown value of hidepid - %s\n", param->string); | |
88 | ||
89 | return 0; | |
90 | } | |
91 | ||
6814ef2d AG |
92 | static int proc_parse_subset_param(struct fs_context *fc, char *value) |
93 | { | |
94 | struct proc_fs_context *ctx = fc->fs_private; | |
95 | ||
96 | while (value) { | |
97 | char *ptr = strchr(value, ','); | |
98 | ||
99 | if (ptr != NULL) | |
100 | *ptr++ = '\0'; | |
101 | ||
102 | if (*value != '\0') { | |
103 | if (!strcmp(value, "pid")) { | |
104 | ctx->pidonly = PROC_PIDONLY_ON; | |
105 | } else { | |
106 | return invalf(fc, "proc: unsupported subset option - %s\n", value); | |
107 | } | |
108 | } | |
109 | value = ptr; | |
110 | } | |
111 | ||
112 | return 0; | |
113 | } | |
114 | ||
66f592e2 | 115 | static int proc_parse_param(struct fs_context *fc, struct fs_parameter *param) |
97412950 | 116 | { |
66f592e2 DH |
117 | struct proc_fs_context *ctx = fc->fs_private; |
118 | struct fs_parse_result result; | |
119 | int opt; | |
120 | ||
d7167b14 | 121 | opt = fs_parse(fc, proc_fs_parameters, param, &result); |
66f592e2 DH |
122 | if (opt < 0) |
123 | return opt; | |
124 | ||
125 | switch (opt) { | |
126 | case Opt_gid: | |
127 | ctx->gid = result.uint_32; | |
128 | break; | |
129 | ||
130 | case Opt_hidepid: | |
1c6c4d11 AG |
131 | if (proc_parse_hidepid_param(fc, param)) |
132 | return -EINVAL; | |
66f592e2 DH |
133 | break; |
134 | ||
6814ef2d AG |
135 | case Opt_subset: |
136 | if (proc_parse_subset_param(fc, param->string) < 0) | |
137 | return -EINVAL; | |
138 | break; | |
139 | ||
66f592e2 DH |
140 | default: |
141 | return -EINVAL; | |
97412950 VK |
142 | } |
143 | ||
66f592e2 DH |
144 | ctx->mask |= 1 << opt; |
145 | return 0; | |
97412950 VK |
146 | } |
147 | ||
fa10fed3 | 148 | static void proc_apply_options(struct proc_fs_info *fs_info, |
66f592e2 | 149 | struct fs_context *fc, |
66f592e2 | 150 | struct user_namespace *user_ns) |
97412950 | 151 | { |
66f592e2 DH |
152 | struct proc_fs_context *ctx = fc->fs_private; |
153 | ||
154 | if (ctx->mask & (1 << Opt_gid)) | |
fa10fed3 | 155 | fs_info->pid_gid = make_kgid(user_ns, ctx->gid); |
66f592e2 | 156 | if (ctx->mask & (1 << Opt_hidepid)) |
fa10fed3 | 157 | fs_info->hide_pid = ctx->hidepid; |
6814ef2d AG |
158 | if (ctx->mask & (1 << Opt_subset)) |
159 | fs_info->pidonly = ctx->pidonly; | |
97412950 VK |
160 | } |
161 | ||
66f592e2 | 162 | static int proc_fill_super(struct super_block *s, struct fs_context *fc) |
60a3c3a5 | 163 | { |
fa10fed3 | 164 | struct proc_fs_context *ctx = fc->fs_private; |
60a3c3a5 | 165 | struct inode *root_inode; |
fa10fed3 | 166 | struct proc_fs_info *fs_info; |
60a3c3a5 DH |
167 | int ret; |
168 | ||
fa10fed3 AG |
169 | fs_info = kzalloc(sizeof(*fs_info), GFP_KERNEL); |
170 | if (!fs_info) | |
171 | return -ENOMEM; | |
172 | ||
173 | fs_info->pid_ns = get_pid_ns(ctx->pid_ns); | |
174 | proc_apply_options(fs_info, fc, current_user_ns()); | |
60a3c3a5 DH |
175 | |
176 | /* User space would break if executables or devices appear on proc */ | |
177 | s->s_iflags |= SB_I_USERNS_VISIBLE | SB_I_NOEXEC | SB_I_NODEV; | |
178 | s->s_flags |= SB_NODIRATIME | SB_NOSUID | SB_NOEXEC; | |
179 | s->s_blocksize = 1024; | |
180 | s->s_blocksize_bits = 10; | |
181 | s->s_magic = PROC_SUPER_MAGIC; | |
182 | s->s_op = &proc_sops; | |
183 | s->s_time_gran = 1; | |
fa10fed3 | 184 | s->s_fs_info = fs_info; |
60a3c3a5 DH |
185 | |
186 | /* | |
187 | * procfs isn't actually a stacking filesystem; however, there is | |
188 | * too much magic going on inside it to permit stacking things on | |
189 | * top of it | |
190 | */ | |
191 | s->s_stack_depth = FILESYSTEM_MAX_STACK_DEPTH; | |
fa10fed3 | 192 | |
60a3c3a5 DH |
193 | /* procfs dentries and inodes don't require IO to create */ |
194 | s->s_shrink.seeks = 0; | |
195 | ||
196 | pde_get(&proc_root); | |
197 | root_inode = proc_get_inode(s, &proc_root); | |
198 | if (!root_inode) { | |
199 | pr_err("proc_fill_super: get root inode failed\n"); | |
200 | return -ENOMEM; | |
201 | } | |
202 | ||
203 | s->s_root = d_make_root(root_inode); | |
204 | if (!s->s_root) { | |
205 | pr_err("proc_fill_super: allocate dentry failed\n"); | |
206 | return -ENOMEM; | |
207 | } | |
208 | ||
209 | ret = proc_setup_self(s); | |
210 | if (ret) { | |
211 | return ret; | |
212 | } | |
213 | return proc_setup_thread_self(s); | |
214 | } | |
215 | ||
66f592e2 | 216 | static int proc_reconfigure(struct fs_context *fc) |
97412950 | 217 | { |
66f592e2 | 218 | struct super_block *sb = fc->root->d_sb; |
fa10fed3 | 219 | struct proc_fs_info *fs_info = proc_sb_info(sb); |
02b9984d TT |
220 | |
221 | sync_filesystem(sb); | |
66f592e2 | 222 | |
fa10fed3 | 223 | proc_apply_options(fs_info, fc, current_user_ns()); |
66f592e2 | 224 | return 0; |
97412950 VK |
225 | } |
226 | ||
66f592e2 | 227 | static int proc_get_tree(struct fs_context *fc) |
1da177e4 | 228 | { |
fa10fed3 | 229 | return get_tree_nodev(fc, proc_fill_super); |
66f592e2 | 230 | } |
07543f5c | 231 | |
66f592e2 DH |
232 | static void proc_fs_context_free(struct fs_context *fc) |
233 | { | |
234 | struct proc_fs_context *ctx = fc->fs_private; | |
235 | ||
46cf047a | 236 | put_pid_ns(ctx->pid_ns); |
66f592e2 DH |
237 | kfree(ctx); |
238 | } | |
239 | ||
240 | static const struct fs_context_operations proc_fs_context_ops = { | |
241 | .free = proc_fs_context_free, | |
242 | .parse_param = proc_parse_param, | |
243 | .get_tree = proc_get_tree, | |
244 | .reconfigure = proc_reconfigure, | |
245 | }; | |
246 | ||
247 | static int proc_init_fs_context(struct fs_context *fc) | |
248 | { | |
249 | struct proc_fs_context *ctx; | |
250 | ||
251 | ctx = kzalloc(sizeof(struct proc_fs_context), GFP_KERNEL); | |
252 | if (!ctx) | |
253 | return -ENOMEM; | |
07543f5c | 254 | |
66f592e2 | 255 | ctx->pid_ns = get_pid_ns(task_active_pid_ns(current)); |
46cf047a AV |
256 | put_user_ns(fc->user_ns); |
257 | fc->user_ns = get_user_ns(ctx->pid_ns->user_ns); | |
66f592e2 DH |
258 | fc->fs_private = ctx; |
259 | fc->ops = &proc_fs_context_ops; | |
260 | return 0; | |
07543f5c PE |
261 | } |
262 | ||
263 | static void proc_kill_sb(struct super_block *sb) | |
264 | { | |
fa10fed3 | 265 | struct proc_fs_info *fs_info = proc_sb_info(sb); |
07543f5c | 266 | |
058f2e4d AG |
267 | if (!fs_info) { |
268 | kill_anon_super(sb); | |
269 | return; | |
270 | } | |
4fa3b1c4 | 271 | |
058f2e4d AG |
272 | dput(fs_info->proc_self); |
273 | dput(fs_info->proc_thread_self); | |
4fa3b1c4 | 274 | |
fa10fed3 AG |
275 | kill_anon_super(sb); |
276 | put_pid_ns(fs_info->pid_ns); | |
277 | kfree(fs_info); | |
1da177e4 LT |
278 | } |
279 | ||
c2319540 | 280 | static struct file_system_type proc_fs_type = { |
66f592e2 DH |
281 | .name = "proc", |
282 | .init_fs_context = proc_init_fs_context, | |
d7167b14 | 283 | .parameters = proc_fs_parameters, |
66f592e2 | 284 | .kill_sb = proc_kill_sb, |
0b3b094a | 285 | .fs_flags = FS_USERNS_MOUNT | FS_DISALLOW_NOTIFY_PERM, |
1da177e4 LT |
286 | }; |
287 | ||
1da177e4 LT |
288 | void __init proc_root_init(void) |
289 | { | |
195b8cf0 | 290 | proc_init_kmemcache(); |
1270dd8d | 291 | set_proc_pid_nlink(); |
e656d8a6 | 292 | proc_self_init(); |
0097875b | 293 | proc_thread_self_init(); |
155134fe | 294 | proc_symlink("mounts", NULL, "self/mounts"); |
457c4cbc EB |
295 | |
296 | proc_net_init(); | |
36a5aeb8 | 297 | proc_mkdir("fs", NULL); |
928b4d8c | 298 | proc_mkdir("driver", NULL); |
eb6d38d5 | 299 | proc_create_mount_point("fs/nfsd"); /* somewhere for the nfsd filesystem to be mounted */ |
1da177e4 LT |
300 | #if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE) |
301 | /* just give it a mountpoint */ | |
eb6d38d5 | 302 | proc_create_mount_point("openprom"); |
1da177e4 LT |
303 | #endif |
304 | proc_tty_init(); | |
9c37066d | 305 | proc_mkdir("bus", NULL); |
77b14db5 | 306 | proc_sys_init(); |
1539d584 AD |
307 | |
308 | register_filesystem(&proc_fs_type); | |
1da177e4 LT |
309 | } |
310 | ||
a528d35e DH |
311 | static int proc_root_getattr(const struct path *path, struct kstat *stat, |
312 | u32 request_mask, unsigned int query_flags) | |
1da177e4 | 313 | { |
0d56a451 | 314 | generic_fillattr(&init_user_ns, d_inode(path->dentry), stat); |
76b6159b AV |
315 | stat->nlink = proc_root.nlink + nr_processes(); |
316 | return 0; | |
317 | } | |
1da177e4 | 318 | |
00cd8dd3 | 319 | static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, unsigned int flags) |
76b6159b | 320 | { |
867aaccf | 321 | if (!proc_pid_lookup(dentry, flags)) |
1da177e4 | 322 | return NULL; |
66f592e2 | 323 | |
335eb531 | 324 | return proc_lookup(dir, dentry, flags); |
1da177e4 LT |
325 | } |
326 | ||
f0c3b509 | 327 | static int proc_root_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 328 | { |
f0c3b509 | 329 | if (ctx->pos < FIRST_PROCESS_ENTRY) { |
94fc5d9d RG |
330 | int error = proc_readdir(file, ctx); |
331 | if (unlikely(error <= 0)) | |
332 | return error; | |
f0c3b509 | 333 | ctx->pos = FIRST_PROCESS_ENTRY; |
1da177e4 | 334 | } |
1da177e4 | 335 | |
f0c3b509 | 336 | return proc_pid_readdir(file, ctx); |
1da177e4 LT |
337 | } |
338 | ||
339 | /* | |
340 | * The root /proc directory is special, as it has the | |
341 | * <pid> directories. Thus we don't use the generic | |
342 | * directory handling functions for that.. | |
343 | */ | |
00977a59 | 344 | static const struct file_operations proc_root_operations = { |
1da177e4 | 345 | .read = generic_read_dir, |
f50752ea AV |
346 | .iterate_shared = proc_root_readdir, |
347 | .llseek = generic_file_llseek, | |
1da177e4 LT |
348 | }; |
349 | ||
350 | /* | |
351 | * proc root can do almost nothing.. | |
352 | */ | |
c5ef1c42 | 353 | static const struct inode_operations proc_root_inode_operations = { |
1da177e4 | 354 | .lookup = proc_root_lookup, |
76b6159b | 355 | .getattr = proc_root_getattr, |
1da177e4 LT |
356 | }; |
357 | ||
358 | /* | |
359 | * This is the root "inode" in the /proc tree.. | |
360 | */ | |
361 | struct proc_dir_entry proc_root = { | |
362 | .low_ino = PROC_ROOT_INO, | |
363 | .namelen = 5, | |
1da177e4 LT |
364 | .mode = S_IFDIR | S_IRUGO | S_IXUGO, |
365 | .nlink = 2, | |
9cdd83e3 | 366 | .refcnt = REFCOUNT_INIT(1), |
1da177e4 | 367 | .proc_iops = &proc_root_inode_operations, |
d56c0d45 | 368 | .proc_dir_ops = &proc_root_operations, |
1da177e4 | 369 | .parent = &proc_root, |
4f113437 | 370 | .subdir = RB_ROOT, |
24074a35 | 371 | .name = "/proc", |
1da177e4 | 372 | }; |