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 = {
52 .specs = proc_param_specs,
55 static int proc_parse_param(struct fs_context *fc, struct fs_parameter *param)
57 struct proc_fs_context *ctx = fc->fs_private;
58 struct fs_parse_result result;
61 opt = fs_parse(fc, &proc_fs_parameters, param, &result);
67 ctx->gid = result.uint_32;
71 ctx->hidepid = result.uint_32;
72 if (ctx->hidepid < HIDEPID_OFF ||
73 ctx->hidepid > HIDEPID_INVISIBLE)
74 return invalf(fc, "proc: hidepid value must be between 0 and 2.\n");
81 ctx->mask |= 1 << opt;
85 static void proc_apply_options(struct super_block *s,
86 struct fs_context *fc,
87 struct pid_namespace *pid_ns,
88 struct user_namespace *user_ns)
90 struct proc_fs_context *ctx = fc->fs_private;
92 if (ctx->mask & (1 << Opt_gid))
93 pid_ns->pid_gid = make_kgid(user_ns, ctx->gid);
94 if (ctx->mask & (1 << Opt_hidepid))
95 pid_ns->hide_pid = ctx->hidepid;
98 static int proc_fill_super(struct super_block *s, struct fs_context *fc)
100 struct pid_namespace *pid_ns = get_pid_ns(s->s_fs_info);
101 struct inode *root_inode;
104 proc_apply_options(s, fc, pid_ns, current_user_ns());
106 /* User space would break if executables or devices appear on proc */
107 s->s_iflags |= SB_I_USERNS_VISIBLE | SB_I_NOEXEC | SB_I_NODEV;
108 s->s_flags |= SB_NODIRATIME | SB_NOSUID | SB_NOEXEC;
109 s->s_blocksize = 1024;
110 s->s_blocksize_bits = 10;
111 s->s_magic = PROC_SUPER_MAGIC;
112 s->s_op = &proc_sops;
116 * procfs isn't actually a stacking filesystem; however, there is
117 * too much magic going on inside it to permit stacking things on
120 s->s_stack_depth = FILESYSTEM_MAX_STACK_DEPTH;
122 /* procfs dentries and inodes don't require IO to create */
123 s->s_shrink.seeks = 0;
126 root_inode = proc_get_inode(s, &proc_root);
128 pr_err("proc_fill_super: get root inode failed\n");
132 s->s_root = d_make_root(root_inode);
134 pr_err("proc_fill_super: allocate dentry failed\n");
138 ret = proc_setup_self(s);
142 return proc_setup_thread_self(s);
145 static int proc_reconfigure(struct fs_context *fc)
147 struct super_block *sb = fc->root->d_sb;
148 struct pid_namespace *pid = sb->s_fs_info;
152 proc_apply_options(sb, fc, pid, current_user_ns());
156 static int proc_get_tree(struct fs_context *fc)
158 struct proc_fs_context *ctx = fc->fs_private;
160 put_user_ns(fc->user_ns);
161 fc->user_ns = get_user_ns(ctx->pid_ns->user_ns);
162 fc->s_fs_info = ctx->pid_ns;
163 return vfs_get_super(fc, vfs_get_keyed_super, proc_fill_super);
166 static void proc_fs_context_free(struct fs_context *fc)
168 struct proc_fs_context *ctx = fc->fs_private;
171 put_pid_ns(ctx->pid_ns);
175 static const struct fs_context_operations proc_fs_context_ops = {
176 .free = proc_fs_context_free,
177 .parse_param = proc_parse_param,
178 .get_tree = proc_get_tree,
179 .reconfigure = proc_reconfigure,
182 static int proc_init_fs_context(struct fs_context *fc)
184 struct proc_fs_context *ctx;
186 ctx = kzalloc(sizeof(struct proc_fs_context), GFP_KERNEL);
190 ctx->pid_ns = get_pid_ns(task_active_pid_ns(current));
191 fc->fs_private = ctx;
192 fc->ops = &proc_fs_context_ops;
196 static void proc_kill_sb(struct super_block *sb)
198 struct pid_namespace *ns;
200 ns = (struct pid_namespace *)sb->s_fs_info;
203 if (ns->proc_thread_self)
204 dput(ns->proc_thread_self);
209 static struct file_system_type proc_fs_type = {
211 .init_fs_context = proc_init_fs_context,
212 .parameters = &proc_fs_parameters,
213 .kill_sb = proc_kill_sb,
214 .fs_flags = FS_USERNS_MOUNT,
217 void __init proc_root_init(void)
219 proc_init_kmemcache();
220 set_proc_pid_nlink();
222 proc_thread_self_init();
223 proc_symlink("mounts", NULL, "self/mounts");
226 proc_mkdir("fs", NULL);
227 proc_mkdir("driver", NULL);
228 proc_create_mount_point("fs/nfsd"); /* somewhere for the nfsd filesystem to be mounted */
229 #if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE)
230 /* just give it a mountpoint */
231 proc_create_mount_point("openprom");
234 proc_mkdir("bus", NULL);
237 register_filesystem(&proc_fs_type);
240 static int proc_root_getattr(const struct path *path, struct kstat *stat,
241 u32 request_mask, unsigned int query_flags)
243 generic_fillattr(d_inode(path->dentry), stat);
244 stat->nlink = proc_root.nlink + nr_processes();
248 static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, unsigned int flags)
250 if (!proc_pid_lookup(dentry, flags))
253 return proc_lookup(dir, dentry, flags);
256 static int proc_root_readdir(struct file *file, struct dir_context *ctx)
258 if (ctx->pos < FIRST_PROCESS_ENTRY) {
259 int error = proc_readdir(file, ctx);
260 if (unlikely(error <= 0))
262 ctx->pos = FIRST_PROCESS_ENTRY;
265 return proc_pid_readdir(file, ctx);
269 * The root /proc directory is special, as it has the
270 * <pid> directories. Thus we don't use the generic
271 * directory handling functions for that..
273 static const struct file_operations proc_root_operations = {
274 .read = generic_read_dir,
275 .iterate_shared = proc_root_readdir,
276 .llseek = generic_file_llseek,
280 * proc root can do almost nothing..
282 static const struct inode_operations proc_root_inode_operations = {
283 .lookup = proc_root_lookup,
284 .getattr = proc_root_getattr,
288 * This is the root "inode" in the /proc tree..
290 struct proc_dir_entry proc_root = {
291 .low_ino = PROC_ROOT_INO,
293 .mode = S_IFDIR | S_IRUGO | S_IXUGO,
295 .refcnt = REFCOUNT_INIT(1),
296 .proc_iops = &proc_root_inode_operations,
297 .proc_fops = &proc_root_operations,
298 .parent = &proc_root,
303 int pid_ns_prepare_proc(struct pid_namespace *ns)
305 struct proc_fs_context *ctx;
306 struct fs_context *fc;
307 struct vfsmount *mnt;
309 fc = fs_context_for_mount(&proc_fs_type, SB_KERNMOUNT);
313 if (fc->user_ns != ns->user_ns) {
314 put_user_ns(fc->user_ns);
315 fc->user_ns = get_user_ns(ns->user_ns);
318 ctx = fc->fs_private;
319 if (ctx->pid_ns != ns) {
320 put_pid_ns(ctx->pid_ns);
334 void pid_ns_release_proc(struct pid_namespace *ns)
336 kern_unmount(ns->proc_mnt);