| 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 | #include <linux/parser.h> |
| 22 | |
| 23 | #include "internal.h" |
| 24 | |
| 25 | static int proc_test_super(struct super_block *sb, void *data) |
| 26 | { |
| 27 | return sb->s_fs_info == data; |
| 28 | } |
| 29 | |
| 30 | static int proc_set_super(struct super_block *sb, void *data) |
| 31 | { |
| 32 | int err = set_anon_super(sb, NULL); |
| 33 | if (!err) { |
| 34 | struct pid_namespace *ns = (struct pid_namespace *)data; |
| 35 | sb->s_fs_info = get_pid_ns(ns); |
| 36 | } |
| 37 | return err; |
| 38 | } |
| 39 | |
| 40 | enum { |
| 41 | Opt_gid, Opt_hidepid, Opt_err, |
| 42 | }; |
| 43 | |
| 44 | static const match_table_t tokens = { |
| 45 | {Opt_hidepid, "hidepid=%u"}, |
| 46 | {Opt_gid, "gid=%u"}, |
| 47 | {Opt_err, NULL}, |
| 48 | }; |
| 49 | |
| 50 | static int proc_parse_options(char *options, struct pid_namespace *pid) |
| 51 | { |
| 52 | char *p; |
| 53 | substring_t args[MAX_OPT_ARGS]; |
| 54 | int option; |
| 55 | |
| 56 | if (!options) |
| 57 | return 1; |
| 58 | |
| 59 | while ((p = strsep(&options, ",")) != NULL) { |
| 60 | int token; |
| 61 | if (!*p) |
| 62 | continue; |
| 63 | |
| 64 | args[0].to = args[0].from = 0; |
| 65 | token = match_token(p, tokens, args); |
| 66 | switch (token) { |
| 67 | case Opt_gid: |
| 68 | if (match_int(&args[0], &option)) |
| 69 | return 0; |
| 70 | pid->pid_gid = make_kgid(current_user_ns(), option); |
| 71 | break; |
| 72 | case Opt_hidepid: |
| 73 | if (match_int(&args[0], &option)) |
| 74 | return 0; |
| 75 | if (option < 0 || option > 2) { |
| 76 | pr_err("proc: hidepid value must be between 0 and 2.\n"); |
| 77 | return 0; |
| 78 | } |
| 79 | pid->hide_pid = option; |
| 80 | break; |
| 81 | default: |
| 82 | pr_err("proc: unrecognized mount option \"%s\" " |
| 83 | "or missing value\n", p); |
| 84 | return 0; |
| 85 | } |
| 86 | } |
| 87 | |
| 88 | return 1; |
| 89 | } |
| 90 | |
| 91 | int proc_remount(struct super_block *sb, int *flags, char *data) |
| 92 | { |
| 93 | struct pid_namespace *pid = sb->s_fs_info; |
| 94 | return !proc_parse_options(data, pid); |
| 95 | } |
| 96 | |
| 97 | static struct dentry *proc_mount(struct file_system_type *fs_type, |
| 98 | int flags, const char *dev_name, void *data) |
| 99 | { |
| 100 | int err; |
| 101 | struct super_block *sb; |
| 102 | struct pid_namespace *ns; |
| 103 | struct proc_inode *ei; |
| 104 | char *options; |
| 105 | |
| 106 | if (flags & MS_KERNMOUNT) { |
| 107 | ns = (struct pid_namespace *)data; |
| 108 | options = NULL; |
| 109 | } else { |
| 110 | ns = current->nsproxy->pid_ns; |
| 111 | options = data; |
| 112 | } |
| 113 | |
| 114 | sb = sget(fs_type, proc_test_super, proc_set_super, ns); |
| 115 | if (IS_ERR(sb)) |
| 116 | return ERR_CAST(sb); |
| 117 | |
| 118 | if (!proc_parse_options(options, ns)) { |
| 119 | deactivate_locked_super(sb); |
| 120 | return ERR_PTR(-EINVAL); |
| 121 | } |
| 122 | |
| 123 | if (!sb->s_root) { |
| 124 | sb->s_flags = flags; |
| 125 | err = proc_fill_super(sb); |
| 126 | if (err) { |
| 127 | deactivate_locked_super(sb); |
| 128 | return ERR_PTR(err); |
| 129 | } |
| 130 | |
| 131 | sb->s_flags |= MS_ACTIVE; |
| 132 | } |
| 133 | |
| 134 | ei = PROC_I(sb->s_root->d_inode); |
| 135 | if (!ei->pid) { |
| 136 | rcu_read_lock(); |
| 137 | ei->pid = get_pid(find_pid_ns(1, ns)); |
| 138 | rcu_read_unlock(); |
| 139 | } |
| 140 | |
| 141 | return dget(sb->s_root); |
| 142 | } |
| 143 | |
| 144 | static void proc_kill_sb(struct super_block *sb) |
| 145 | { |
| 146 | struct pid_namespace *ns; |
| 147 | |
| 148 | ns = (struct pid_namespace *)sb->s_fs_info; |
| 149 | kill_anon_super(sb); |
| 150 | put_pid_ns(ns); |
| 151 | } |
| 152 | |
| 153 | static struct file_system_type proc_fs_type = { |
| 154 | .name = "proc", |
| 155 | .mount = proc_mount, |
| 156 | .kill_sb = proc_kill_sb, |
| 157 | }; |
| 158 | |
| 159 | void __init proc_root_init(void) |
| 160 | { |
| 161 | int err; |
| 162 | |
| 163 | proc_init_inodecache(); |
| 164 | err = register_filesystem(&proc_fs_type); |
| 165 | if (err) |
| 166 | return; |
| 167 | err = pid_ns_prepare_proc(&init_pid_ns); |
| 168 | if (err) { |
| 169 | unregister_filesystem(&proc_fs_type); |
| 170 | return; |
| 171 | } |
| 172 | |
| 173 | proc_symlink("mounts", NULL, "self/mounts"); |
| 174 | |
| 175 | proc_net_init(); |
| 176 | |
| 177 | #ifdef CONFIG_SYSVIPC |
| 178 | proc_mkdir("sysvipc", NULL); |
| 179 | #endif |
| 180 | proc_mkdir("fs", NULL); |
| 181 | proc_mkdir("driver", NULL); |
| 182 | proc_mkdir("fs/nfsd", NULL); /* somewhere for the nfsd filesystem to be mounted */ |
| 183 | #if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE) |
| 184 | /* just give it a mountpoint */ |
| 185 | proc_mkdir("openprom", NULL); |
| 186 | #endif |
| 187 | proc_tty_init(); |
| 188 | #ifdef CONFIG_PROC_DEVICETREE |
| 189 | proc_device_tree_init(); |
| 190 | #endif |
| 191 | proc_mkdir("bus", NULL); |
| 192 | proc_sys_init(); |
| 193 | } |
| 194 | |
| 195 | static int proc_root_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat |
| 196 | ) |
| 197 | { |
| 198 | generic_fillattr(dentry->d_inode, stat); |
| 199 | stat->nlink = proc_root.nlink + nr_processes(); |
| 200 | return 0; |
| 201 | } |
| 202 | |
| 203 | static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, unsigned int flags) |
| 204 | { |
| 205 | if (!proc_lookup(dir, dentry, flags)) |
| 206 | return NULL; |
| 207 | |
| 208 | return proc_pid_lookup(dir, dentry, flags); |
| 209 | } |
| 210 | |
| 211 | static int proc_root_readdir(struct file * filp, |
| 212 | void * dirent, filldir_t filldir) |
| 213 | { |
| 214 | unsigned int nr = filp->f_pos; |
| 215 | int ret; |
| 216 | |
| 217 | if (nr < FIRST_PROCESS_ENTRY) { |
| 218 | int error = proc_readdir(filp, dirent, filldir); |
| 219 | if (error <= 0) |
| 220 | return error; |
| 221 | filp->f_pos = FIRST_PROCESS_ENTRY; |
| 222 | } |
| 223 | |
| 224 | ret = proc_pid_readdir(filp, dirent, filldir); |
| 225 | return ret; |
| 226 | } |
| 227 | |
| 228 | /* |
| 229 | * The root /proc directory is special, as it has the |
| 230 | * <pid> directories. Thus we don't use the generic |
| 231 | * directory handling functions for that.. |
| 232 | */ |
| 233 | static const struct file_operations proc_root_operations = { |
| 234 | .read = generic_read_dir, |
| 235 | .readdir = proc_root_readdir, |
| 236 | .llseek = default_llseek, |
| 237 | }; |
| 238 | |
| 239 | /* |
| 240 | * proc root can do almost nothing.. |
| 241 | */ |
| 242 | static const struct inode_operations proc_root_inode_operations = { |
| 243 | .lookup = proc_root_lookup, |
| 244 | .getattr = proc_root_getattr, |
| 245 | }; |
| 246 | |
| 247 | /* |
| 248 | * This is the root "inode" in the /proc tree.. |
| 249 | */ |
| 250 | struct proc_dir_entry proc_root = { |
| 251 | .low_ino = PROC_ROOT_INO, |
| 252 | .namelen = 5, |
| 253 | .mode = S_IFDIR | S_IRUGO | S_IXUGO, |
| 254 | .nlink = 2, |
| 255 | .count = ATOMIC_INIT(1), |
| 256 | .proc_iops = &proc_root_inode_operations, |
| 257 | .proc_fops = &proc_root_operations, |
| 258 | .parent = &proc_root, |
| 259 | .name = "/proc", |
| 260 | }; |
| 261 | |
| 262 | int pid_ns_prepare_proc(struct pid_namespace *ns) |
| 263 | { |
| 264 | struct vfsmount *mnt; |
| 265 | |
| 266 | mnt = kern_mount_data(&proc_fs_type, ns); |
| 267 | if (IS_ERR(mnt)) |
| 268 | return PTR_ERR(mnt); |
| 269 | |
| 270 | ns->proc_mnt = mnt; |
| 271 | return 0; |
| 272 | } |
| 273 | |
| 274 | void pid_ns_release_proc(struct pid_namespace *ns) |
| 275 | { |
| 276 | kern_unmount(ns->proc_mnt); |
| 277 | } |