1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1991, 1992 Linus Torvalds
7 * proc root directory handling functions
10 #include <linux/uaccess.h>
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>
32 struct proc_fs_context {
33 struct pid_namespace *pid_ns;
44 static const struct fs_parameter_spec proc_param_specs[] = {
45 fsparam_u32("gid", Opt_gid),
46 fsparam_u32("hidepid", Opt_hidepid),
50 static const struct fs_parameter_description proc_fs_parameters = {
51 .specs = proc_param_specs,
54 static int proc_parse_param(struct fs_context *fc, struct fs_parameter *param)
56 struct proc_fs_context *ctx = fc->fs_private;
57 struct fs_parse_result result;
60 opt = fs_parse(fc, &proc_fs_parameters, param, &result);
66 ctx->gid = result.uint_32;
70 ctx->hidepid = result.uint_32;
71 if (ctx->hidepid < HIDEPID_OFF ||
72 ctx->hidepid > HIDEPID_INVISIBLE)
73 return invalf(fc, "proc: hidepid value must be between 0 and 2.\n");
80 ctx->mask |= 1 << opt;
84 static void proc_apply_options(struct super_block *s,
85 struct fs_context *fc,
86 struct pid_namespace *pid_ns,
87 struct user_namespace *user_ns)
89 struct proc_fs_context *ctx = fc->fs_private;
91 if (ctx->mask & (1 << Opt_gid))
92 pid_ns->pid_gid = make_kgid(user_ns, ctx->gid);
93 if (ctx->mask & (1 << Opt_hidepid))
94 pid_ns->hide_pid = ctx->hidepid;
97 static int proc_fill_super(struct super_block *s, struct fs_context *fc)
99 struct pid_namespace *pid_ns = get_pid_ns(s->s_fs_info);
100 struct inode *root_inode;
103 proc_apply_options(s, fc, pid_ns, current_user_ns());
105 /* User space would break if executables or devices appear on proc */
106 s->s_iflags |= SB_I_USERNS_VISIBLE | SB_I_NOEXEC | SB_I_NODEV;
107 s->s_flags |= SB_NODIRATIME | SB_NOSUID | SB_NOEXEC;
108 s->s_blocksize = 1024;
109 s->s_blocksize_bits = 10;
110 s->s_magic = PROC_SUPER_MAGIC;
111 s->s_op = &proc_sops;
115 * procfs isn't actually a stacking filesystem; however, there is
116 * too much magic going on inside it to permit stacking things on
119 s->s_stack_depth = FILESYSTEM_MAX_STACK_DEPTH;
121 /* procfs dentries and inodes don't require IO to create */
122 s->s_shrink.seeks = 0;
125 root_inode = proc_get_inode(s, &proc_root);
127 pr_err("proc_fill_super: get root inode failed\n");
131 s->s_root = d_make_root(root_inode);
133 pr_err("proc_fill_super: allocate dentry failed\n");
137 ret = proc_setup_self(s);
141 return proc_setup_thread_self(s);
144 static int proc_reconfigure(struct fs_context *fc)
146 struct super_block *sb = fc->root->d_sb;
147 struct pid_namespace *pid = sb->s_fs_info;
151 proc_apply_options(sb, fc, pid, current_user_ns());
155 static int proc_get_tree(struct fs_context *fc)
157 struct proc_fs_context *ctx = fc->fs_private;
159 return get_tree_keyed(fc, proc_fill_super, ctx->pid_ns);
162 static void proc_fs_context_free(struct fs_context *fc)
164 struct proc_fs_context *ctx = fc->fs_private;
166 put_pid_ns(ctx->pid_ns);
170 static const struct fs_context_operations proc_fs_context_ops = {
171 .free = proc_fs_context_free,
172 .parse_param = proc_parse_param,
173 .get_tree = proc_get_tree,
174 .reconfigure = proc_reconfigure,
177 static int proc_init_fs_context(struct fs_context *fc)
179 struct proc_fs_context *ctx;
181 ctx = kzalloc(sizeof(struct proc_fs_context), GFP_KERNEL);
185 ctx->pid_ns = get_pid_ns(task_active_pid_ns(current));
186 put_user_ns(fc->user_ns);
187 fc->user_ns = get_user_ns(ctx->pid_ns->user_ns);
188 fc->fs_private = ctx;
189 fc->ops = &proc_fs_context_ops;
193 static void proc_kill_sb(struct super_block *sb)
195 struct pid_namespace *ns;
197 ns = (struct pid_namespace *)sb->s_fs_info;
200 if (ns->proc_thread_self)
201 dput(ns->proc_thread_self);
206 static struct file_system_type proc_fs_type = {
208 .init_fs_context = proc_init_fs_context,
209 .parameters = &proc_fs_parameters,
210 .kill_sb = proc_kill_sb,
211 .fs_flags = FS_USERNS_MOUNT | FS_DISALLOW_NOTIFY_PERM,
214 void __init proc_root_init(void)
216 proc_init_kmemcache();
217 set_proc_pid_nlink();
219 proc_thread_self_init();
220 proc_symlink("mounts", NULL, "self/mounts");
223 proc_mkdir("fs", NULL);
224 proc_mkdir("driver", NULL);
225 proc_create_mount_point("fs/nfsd"); /* somewhere for the nfsd filesystem to be mounted */
226 #if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE)
227 /* just give it a mountpoint */
228 proc_create_mount_point("openprom");
231 proc_mkdir("bus", NULL);
234 register_filesystem(&proc_fs_type);
237 static int proc_root_getattr(const struct path *path, struct kstat *stat,
238 u32 request_mask, unsigned int query_flags)
240 generic_fillattr(d_inode(path->dentry), stat);
241 stat->nlink = proc_root.nlink + nr_processes();
245 static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, unsigned int flags)
247 if (!proc_pid_lookup(dentry, flags))
250 return proc_lookup(dir, dentry, flags);
253 static int proc_root_readdir(struct file *file, struct dir_context *ctx)
255 if (ctx->pos < FIRST_PROCESS_ENTRY) {
256 int error = proc_readdir(file, ctx);
257 if (unlikely(error <= 0))
259 ctx->pos = FIRST_PROCESS_ENTRY;
262 return proc_pid_readdir(file, ctx);
266 * The root /proc directory is special, as it has the
267 * <pid> directories. Thus we don't use the generic
268 * directory handling functions for that..
270 static const struct file_operations proc_root_operations = {
271 .read = generic_read_dir,
272 .iterate_shared = proc_root_readdir,
273 .llseek = generic_file_llseek,
277 * proc root can do almost nothing..
279 static const struct inode_operations proc_root_inode_operations = {
280 .lookup = proc_root_lookup,
281 .getattr = proc_root_getattr,
285 * This is the root "inode" in the /proc tree..
287 struct proc_dir_entry proc_root = {
288 .low_ino = PROC_ROOT_INO,
290 .mode = S_IFDIR | S_IRUGO | S_IXUGO,
292 .refcnt = REFCOUNT_INIT(1),
293 .proc_iops = &proc_root_inode_operations,
294 .proc_fops = &proc_root_operations,
295 .parent = &proc_root,
300 int pid_ns_prepare_proc(struct pid_namespace *ns)
302 struct proc_fs_context *ctx;
303 struct fs_context *fc;
304 struct vfsmount *mnt;
306 fc = fs_context_for_mount(&proc_fs_type, SB_KERNMOUNT);
310 if (fc->user_ns != ns->user_ns) {
311 put_user_ns(fc->user_ns);
312 fc->user_ns = get_user_ns(ns->user_ns);
315 ctx = fc->fs_private;
316 if (ctx->pid_ns != ns) {
317 put_pid_ns(ctx->pid_ns);
331 void pid_ns_release_proc(struct pid_namespace *ns)
333 kern_unmount(ns->proc_mnt);