Loading...
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/mount.h>
20#include <linux/pid_namespace.h>
21
22#include "internal.h"
23
24static int proc_test_super(struct super_block *sb, void *data)
25{
26 return sb->s_fs_info == data;
27}
28
29static int proc_set_super(struct super_block *sb, void *data)
30{
31 int err = set_anon_super(sb, NULL);
32 if (!err) {
33 struct pid_namespace *ns = (struct pid_namespace *)data;
34 sb->s_fs_info = get_pid_ns(ns);
35 }
36 return err;
37}
38
39static struct dentry *proc_mount(struct file_system_type *fs_type,
40 int flags, const char *dev_name, void *data)
41{
42 int err;
43 struct super_block *sb;
44 struct pid_namespace *ns;
45 struct proc_inode *ei;
46
47 if (flags & MS_KERNMOUNT)
48 ns = (struct pid_namespace *)data;
49 else
50 ns = current->nsproxy->pid_ns;
51
52 sb = sget(fs_type, proc_test_super, proc_set_super, ns);
53 if (IS_ERR(sb))
54 return ERR_CAST(sb);
55
56 if (!sb->s_root) {
57 sb->s_flags = flags;
58 err = proc_fill_super(sb);
59 if (err) {
60 deactivate_locked_super(sb);
61 return ERR_PTR(err);
62 }
63
64 sb->s_flags |= MS_ACTIVE;
65 }
66
67 ei = PROC_I(sb->s_root->d_inode);
68 if (!ei->pid) {
69 rcu_read_lock();
70 ei->pid = get_pid(find_pid_ns(1, ns));
71 rcu_read_unlock();
72 }
73
74 return dget(sb->s_root);
75}
76
77static void proc_kill_sb(struct super_block *sb)
78{
79 struct pid_namespace *ns;
80
81 ns = (struct pid_namespace *)sb->s_fs_info;
82 kill_anon_super(sb);
83 put_pid_ns(ns);
84}
85
86static struct file_system_type proc_fs_type = {
87 .name = "proc",
88 .mount = proc_mount,
89 .kill_sb = proc_kill_sb,
90};
91
92void __init proc_root_init(void)
93{
94 struct vfsmount *mnt;
95 int err;
96
97 proc_init_inodecache();
98 err = register_filesystem(&proc_fs_type);
99 if (err)
100 return;
101 mnt = kern_mount_data(&proc_fs_type, &init_pid_ns);
102 if (IS_ERR(mnt)) {
103 unregister_filesystem(&proc_fs_type);
104 return;
105 }
106
107 init_pid_ns.proc_mnt = mnt;
108 proc_symlink("mounts", NULL, "self/mounts");
109
110 proc_net_init();
111
112#ifdef CONFIG_SYSVIPC
113 proc_mkdir("sysvipc", NULL);
114#endif
115 proc_mkdir("fs", NULL);
116 proc_mkdir("driver", NULL);
117 proc_mkdir("fs/nfsd", NULL); /* somewhere for the nfsd filesystem to be mounted */
118#if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE)
119 /* just give it a mountpoint */
120 proc_mkdir("openprom", NULL);
121#endif
122 proc_tty_init();
123#ifdef CONFIG_PROC_DEVICETREE
124 proc_device_tree_init();
125#endif
126 proc_mkdir("bus", NULL);
127 proc_sys_init();
128}
129
130static int proc_root_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat
131)
132{
133 generic_fillattr(dentry->d_inode, stat);
134 stat->nlink = proc_root.nlink + nr_processes();
135 return 0;
136}
137
138static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
139{
140 if (!proc_lookup(dir, dentry, nd)) {
141 return NULL;
142 }
143
144 return proc_pid_lookup(dir, dentry, nd);
145}
146
147static int proc_root_readdir(struct file * filp,
148 void * dirent, filldir_t filldir)
149{
150 unsigned int nr = filp->f_pos;
151 int ret;
152
153 if (nr < FIRST_PROCESS_ENTRY) {
154 int error = proc_readdir(filp, dirent, filldir);
155 if (error <= 0)
156 return error;
157 filp->f_pos = FIRST_PROCESS_ENTRY;
158 }
159
160 ret = proc_pid_readdir(filp, dirent, filldir);
161 return ret;
162}
163
164/*
165 * The root /proc directory is special, as it has the
166 * <pid> directories. Thus we don't use the generic
167 * directory handling functions for that..
168 */
169static const struct file_operations proc_root_operations = {
170 .read = generic_read_dir,
171 .readdir = proc_root_readdir,
172 .llseek = default_llseek,
173};
174
175/*
176 * proc root can do almost nothing..
177 */
178static const struct inode_operations proc_root_inode_operations = {
179 .lookup = proc_root_lookup,
180 .getattr = proc_root_getattr,
181};
182
183/*
184 * This is the root "inode" in the /proc tree..
185 */
186struct proc_dir_entry proc_root = {
187 .low_ino = PROC_ROOT_INO,
188 .namelen = 5,
189 .mode = S_IFDIR | S_IRUGO | S_IXUGO,
190 .nlink = 2,
191 .count = ATOMIC_INIT(1),
192 .proc_iops = &proc_root_inode_operations,
193 .proc_fops = &proc_root_operations,
194 .parent = &proc_root,
195 .name = "/proc",
196};
197
198int pid_ns_prepare_proc(struct pid_namespace *ns)
199{
200 struct vfsmount *mnt;
201
202 mnt = kern_mount_data(&proc_fs_type, ns);
203 if (IS_ERR(mnt))
204 return PTR_ERR(mnt);
205
206 ns->proc_mnt = mnt;
207 return 0;
208}
209
210void pid_ns_release_proc(struct pid_namespace *ns)
211{
212 mntput(ns->proc_mnt);
213}
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * linux/fs/proc/root.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * proc root directory handling functions
8 */
9
10#include <linux/uaccess.h>
11
12#include <linux/errno.h>
13#include <linux/time.h>
14#include <linux/proc_fs.h>
15#include <linux/stat.h>
16#include <linux/init.h>
17#include <linux/sched.h>
18#include <linux/sched/stat.h>
19#include <linux/module.h>
20#include <linux/bitops.h>
21#include <linux/user_namespace.h>
22#include <linux/fs_context.h>
23#include <linux/mount.h>
24#include <linux/pid_namespace.h>
25#include <linux/fs_parser.h>
26#include <linux/cred.h>
27#include <linux/magic.h>
28#include <linux/slab.h>
29
30#include "internal.h"
31
32struct proc_fs_context {
33 struct pid_namespace *pid_ns;
34 unsigned int mask;
35 enum proc_hidepid hidepid;
36 int gid;
37 enum proc_pidonly pidonly;
38};
39
40enum proc_param {
41 Opt_gid,
42 Opt_hidepid,
43 Opt_subset,
44};
45
46static const struct fs_parameter_spec proc_fs_parameters[] = {
47 fsparam_u32("gid", Opt_gid),
48 fsparam_string("hidepid", Opt_hidepid),
49 fsparam_string("subset", Opt_subset),
50 {}
51};
52
53static 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
61static 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
92static 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
115static int proc_parse_param(struct fs_context *fc, struct fs_parameter *param)
116{
117 struct proc_fs_context *ctx = fc->fs_private;
118 struct fs_parse_result result;
119 int opt;
120
121 opt = fs_parse(fc, proc_fs_parameters, param, &result);
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:
131 if (proc_parse_hidepid_param(fc, param))
132 return -EINVAL;
133 break;
134
135 case Opt_subset:
136 if (proc_parse_subset_param(fc, param->string) < 0)
137 return -EINVAL;
138 break;
139
140 default:
141 return -EINVAL;
142 }
143
144 ctx->mask |= 1 << opt;
145 return 0;
146}
147
148static void proc_apply_options(struct proc_fs_info *fs_info,
149 struct fs_context *fc,
150 struct user_namespace *user_ns)
151{
152 struct proc_fs_context *ctx = fc->fs_private;
153
154 if (ctx->mask & (1 << Opt_gid))
155 fs_info->pid_gid = make_kgid(user_ns, ctx->gid);
156 if (ctx->mask & (1 << Opt_hidepid))
157 fs_info->hide_pid = ctx->hidepid;
158 if (ctx->mask & (1 << Opt_subset))
159 fs_info->pidonly = ctx->pidonly;
160}
161
162static int proc_fill_super(struct super_block *s, struct fs_context *fc)
163{
164 struct proc_fs_context *ctx = fc->fs_private;
165 struct inode *root_inode;
166 struct proc_fs_info *fs_info;
167 int ret;
168
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());
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;
184 s->s_fs_info = fs_info;
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;
192
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
216static int proc_reconfigure(struct fs_context *fc)
217{
218 struct super_block *sb = fc->root->d_sb;
219 struct proc_fs_info *fs_info = proc_sb_info(sb);
220
221 sync_filesystem(sb);
222
223 proc_apply_options(fs_info, fc, current_user_ns());
224 return 0;
225}
226
227static int proc_get_tree(struct fs_context *fc)
228{
229 return get_tree_nodev(fc, proc_fill_super);
230}
231
232static void proc_fs_context_free(struct fs_context *fc)
233{
234 struct proc_fs_context *ctx = fc->fs_private;
235
236 put_pid_ns(ctx->pid_ns);
237 kfree(ctx);
238}
239
240static 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
247static 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;
254
255 ctx->pid_ns = get_pid_ns(task_active_pid_ns(current));
256 put_user_ns(fc->user_ns);
257 fc->user_ns = get_user_ns(ctx->pid_ns->user_ns);
258 fc->fs_private = ctx;
259 fc->ops = &proc_fs_context_ops;
260 return 0;
261}
262
263static void proc_kill_sb(struct super_block *sb)
264{
265 struct proc_fs_info *fs_info = proc_sb_info(sb);
266
267 if (!fs_info) {
268 kill_anon_super(sb);
269 return;
270 }
271
272 dput(fs_info->proc_self);
273 dput(fs_info->proc_thread_self);
274
275 kill_anon_super(sb);
276 put_pid_ns(fs_info->pid_ns);
277 kfree(fs_info);
278}
279
280static struct file_system_type proc_fs_type = {
281 .name = "proc",
282 .init_fs_context = proc_init_fs_context,
283 .parameters = proc_fs_parameters,
284 .kill_sb = proc_kill_sb,
285 .fs_flags = FS_USERNS_MOUNT | FS_DISALLOW_NOTIFY_PERM,
286};
287
288void __init proc_root_init(void)
289{
290 proc_init_kmemcache();
291 set_proc_pid_nlink();
292 proc_self_init();
293 proc_thread_self_init();
294 proc_symlink("mounts", NULL, "self/mounts");
295
296 proc_net_init();
297 proc_mkdir("fs", NULL);
298 proc_mkdir("driver", NULL);
299 proc_create_mount_point("fs/nfsd"); /* somewhere for the nfsd filesystem to be mounted */
300#if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE)
301 /* just give it a mountpoint */
302 proc_create_mount_point("openprom");
303#endif
304 proc_tty_init();
305 proc_mkdir("bus", NULL);
306 proc_sys_init();
307
308 register_filesystem(&proc_fs_type);
309}
310
311static int proc_root_getattr(struct user_namespace *mnt_userns,
312 const struct path *path, struct kstat *stat,
313 u32 request_mask, unsigned int query_flags)
314{
315 generic_fillattr(&init_user_ns, d_inode(path->dentry), stat);
316 stat->nlink = proc_root.nlink + nr_processes();
317 return 0;
318}
319
320static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, unsigned int flags)
321{
322 if (!proc_pid_lookup(dentry, flags))
323 return NULL;
324
325 return proc_lookup(dir, dentry, flags);
326}
327
328static int proc_root_readdir(struct file *file, struct dir_context *ctx)
329{
330 if (ctx->pos < FIRST_PROCESS_ENTRY) {
331 int error = proc_readdir(file, ctx);
332 if (unlikely(error <= 0))
333 return error;
334 ctx->pos = FIRST_PROCESS_ENTRY;
335 }
336
337 return proc_pid_readdir(file, ctx);
338}
339
340/*
341 * The root /proc directory is special, as it has the
342 * <pid> directories. Thus we don't use the generic
343 * directory handling functions for that..
344 */
345static const struct file_operations proc_root_operations = {
346 .read = generic_read_dir,
347 .iterate_shared = proc_root_readdir,
348 .llseek = generic_file_llseek,
349};
350
351/*
352 * proc root can do almost nothing..
353 */
354static const struct inode_operations proc_root_inode_operations = {
355 .lookup = proc_root_lookup,
356 .getattr = proc_root_getattr,
357};
358
359/*
360 * This is the root "inode" in the /proc tree..
361 */
362struct proc_dir_entry proc_root = {
363 .low_ino = PROC_ROOT_INO,
364 .namelen = 5,
365 .mode = S_IFDIR | S_IRUGO | S_IXUGO,
366 .nlink = 2,
367 .refcnt = REFCOUNT_INIT(1),
368 .proc_iops = &proc_root_inode_operations,
369 .proc_dir_ops = &proc_root_operations,
370 .parent = &proc_root,
371 .subdir = RB_ROOT,
372 .name = "/proc",
373};