| 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/user_namespace.h> |
| 20 | #include <linux/mount.h> |
| 21 | #include <linux/pid_namespace.h> |
| 22 | #include <linux/parser.h> |
| 23 | |
| 24 | #include "internal.h" |
| 25 | |
| 26 | static int proc_test_super(struct super_block *sb, void *data) |
| 27 | { |
| 28 | return sb->s_fs_info == data; |
| 29 | } |
| 30 | |
| 31 | static int proc_set_super(struct super_block *sb, void *data) |
| 32 | { |
| 33 | int err = set_anon_super(sb, NULL); |
| 34 | if (!err) { |
| 35 | struct pid_namespace *ns = (struct pid_namespace *)data; |
| 36 | sb->s_fs_info = get_pid_ns(ns); |
| 37 | } |
| 38 | return err; |
| 39 | } |
| 40 | |
| 41 | enum { |
| 42 | Opt_gid, Opt_hidepid, Opt_err, |
| 43 | }; |
| 44 | |
| 45 | static const match_table_t tokens = { |
| 46 | {Opt_hidepid, "hidepid=%u"}, |
| 47 | {Opt_gid, "gid=%u"}, |
| 48 | {Opt_err, NULL}, |
| 49 | }; |
| 50 | |
| 51 | static int proc_parse_options(char *options, struct pid_namespace *pid) |
| 52 | { |
| 53 | char *p; |
| 54 | substring_t args[MAX_OPT_ARGS]; |
| 55 | int option; |
| 56 | |
| 57 | if (!options) |
| 58 | return 1; |
| 59 | |
| 60 | while ((p = strsep(&options, ",")) != NULL) { |
| 61 | int token; |
| 62 | if (!*p) |
| 63 | continue; |
| 64 | |
| 65 | args[0].to = args[0].from = NULL; |
| 66 | token = match_token(p, tokens, args); |
| 67 | switch (token) { |
| 68 | case Opt_gid: |
| 69 | if (match_int(&args[0], &option)) |
| 70 | return 0; |
| 71 | pid->pid_gid = make_kgid(current_user_ns(), option); |
| 72 | break; |
| 73 | case Opt_hidepid: |
| 74 | if (match_int(&args[0], &option)) |
| 75 | return 0; |
| 76 | if (option < 0 || option > 2) { |
| 77 | pr_err("proc: hidepid value must be between 0 and 2.\n"); |
| 78 | return 0; |
| 79 | } |
| 80 | pid->hide_pid = option; |
| 81 | break; |
| 82 | default: |
| 83 | pr_err("proc: unrecognized mount option \"%s\" " |
| 84 | "or missing value\n", p); |
| 85 | return 0; |
| 86 | } |
| 87 | } |
| 88 | |
| 89 | return 1; |
| 90 | } |
| 91 | |
| 92 | int proc_remount(struct super_block *sb, int *flags, char *data) |
| 93 | { |
| 94 | struct pid_namespace *pid = sb->s_fs_info; |
| 95 | |
| 96 | sync_filesystem(sb); |
| 97 | return !proc_parse_options(data, pid); |
| 98 | } |
| 99 | |
| 100 | static struct dentry *proc_mount(struct file_system_type *fs_type, |
| 101 | int flags, const char *dev_name, void *data) |
| 102 | { |
| 103 | int err; |
| 104 | struct super_block *sb; |
| 105 | struct pid_namespace *ns; |
| 106 | char *options; |
| 107 | |
| 108 | if (flags & MS_KERNMOUNT) { |
| 109 | ns = (struct pid_namespace *)data; |
| 110 | options = NULL; |
| 111 | } else { |
| 112 | ns = task_active_pid_ns(current); |
| 113 | options = data; |
| 114 | |
| 115 | /* Does the mounter have privilege over the pid namespace? */ |
| 116 | if (!ns_capable(ns->user_ns, CAP_SYS_ADMIN)) |
| 117 | return ERR_PTR(-EPERM); |
| 118 | } |
| 119 | |
| 120 | sb = sget(fs_type, proc_test_super, proc_set_super, flags, ns); |
| 121 | if (IS_ERR(sb)) |
| 122 | return ERR_CAST(sb); |
| 123 | |
| 124 | if (!proc_parse_options(options, ns)) { |
| 125 | deactivate_locked_super(sb); |
| 126 | return ERR_PTR(-EINVAL); |
| 127 | } |
| 128 | |
| 129 | if (!sb->s_root) { |
| 130 | err = proc_fill_super(sb); |
| 131 | if (err) { |
| 132 | deactivate_locked_super(sb); |
| 133 | return ERR_PTR(err); |
| 134 | } |
| 135 | |
| 136 | sb->s_flags |= MS_ACTIVE; |
| 137 | /* User space would break if executables appear on proc */ |
| 138 | sb->s_iflags |= SB_I_NOEXEC; |
| 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 | if (ns->proc_self) |
| 150 | dput(ns->proc_self); |
| 151 | if (ns->proc_thread_self) |
| 152 | dput(ns->proc_thread_self); |
| 153 | kill_anon_super(sb); |
| 154 | put_pid_ns(ns); |
| 155 | } |
| 156 | |
| 157 | static struct file_system_type proc_fs_type = { |
| 158 | .name = "proc", |
| 159 | .mount = proc_mount, |
| 160 | .kill_sb = proc_kill_sb, |
| 161 | .fs_flags = FS_USERNS_VISIBLE | FS_USERNS_MOUNT, |
| 162 | }; |
| 163 | |
| 164 | void __init proc_root_init(void) |
| 165 | { |
| 166 | int err; |
| 167 | |
| 168 | proc_init_inodecache(); |
| 169 | err = register_filesystem(&proc_fs_type); |
| 170 | if (err) |
| 171 | return; |
| 172 | |
| 173 | proc_self_init(); |
| 174 | proc_thread_self_init(); |
| 175 | proc_symlink("mounts", NULL, "self/mounts"); |
| 176 | |
| 177 | proc_net_init(); |
| 178 | |
| 179 | #ifdef CONFIG_SYSVIPC |
| 180 | proc_mkdir("sysvipc", NULL); |
| 181 | #endif |
| 182 | proc_mkdir("fs", NULL); |
| 183 | proc_mkdir("driver", NULL); |
| 184 | proc_create_mount_point("fs/nfsd"); /* somewhere for the nfsd filesystem to be mounted */ |
| 185 | #if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE) |
| 186 | /* just give it a mountpoint */ |
| 187 | proc_create_mount_point("openprom"); |
| 188 | #endif |
| 189 | proc_tty_init(); |
| 190 | proc_mkdir("bus", NULL); |
| 191 | proc_sys_init(); |
| 192 | } |
| 193 | |
| 194 | static int proc_root_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat |
| 195 | ) |
| 196 | { |
| 197 | generic_fillattr(d_inode(dentry), stat); |
| 198 | stat->nlink = proc_root.nlink + nr_processes(); |
| 199 | return 0; |
| 200 | } |
| 201 | |
| 202 | static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, unsigned int flags) |
| 203 | { |
| 204 | if (!proc_pid_lookup(dir, dentry, flags)) |
| 205 | return NULL; |
| 206 | |
| 207 | return proc_lookup(dir, dentry, flags); |
| 208 | } |
| 209 | |
| 210 | static int proc_root_readdir(struct file *file, struct dir_context *ctx) |
| 211 | { |
| 212 | if (ctx->pos < FIRST_PROCESS_ENTRY) { |
| 213 | int error = proc_readdir(file, ctx); |
| 214 | if (unlikely(error <= 0)) |
| 215 | return error; |
| 216 | ctx->pos = FIRST_PROCESS_ENTRY; |
| 217 | } |
| 218 | |
| 219 | return proc_pid_readdir(file, ctx); |
| 220 | } |
| 221 | |
| 222 | /* |
| 223 | * The root /proc directory is special, as it has the |
| 224 | * <pid> directories. Thus we don't use the generic |
| 225 | * directory handling functions for that.. |
| 226 | */ |
| 227 | static const struct file_operations proc_root_operations = { |
| 228 | .read = generic_read_dir, |
| 229 | .iterate_shared = proc_root_readdir, |
| 230 | .llseek = generic_file_llseek, |
| 231 | }; |
| 232 | |
| 233 | /* |
| 234 | * proc root can do almost nothing.. |
| 235 | */ |
| 236 | static const struct inode_operations proc_root_inode_operations = { |
| 237 | .lookup = proc_root_lookup, |
| 238 | .getattr = proc_root_getattr, |
| 239 | }; |
| 240 | |
| 241 | /* |
| 242 | * This is the root "inode" in the /proc tree.. |
| 243 | */ |
| 244 | struct proc_dir_entry proc_root = { |
| 245 | .low_ino = PROC_ROOT_INO, |
| 246 | .namelen = 5, |
| 247 | .mode = S_IFDIR | S_IRUGO | S_IXUGO, |
| 248 | .nlink = 2, |
| 249 | .count = ATOMIC_INIT(1), |
| 250 | .proc_iops = &proc_root_inode_operations, |
| 251 | .proc_fops = &proc_root_operations, |
| 252 | .parent = &proc_root, |
| 253 | .subdir = RB_ROOT, |
| 254 | .name = "/proc", |
| 255 | }; |
| 256 | |
| 257 | int pid_ns_prepare_proc(struct pid_namespace *ns) |
| 258 | { |
| 259 | struct vfsmount *mnt; |
| 260 | |
| 261 | mnt = kern_mount_data(&proc_fs_type, ns); |
| 262 | if (IS_ERR(mnt)) |
| 263 | return PTR_ERR(mnt); |
| 264 | |
| 265 | ns->proc_mnt = mnt; |
| 266 | return 0; |
| 267 | } |
| 268 | |
| 269 | void pid_ns_release_proc(struct pid_namespace *ns) |
| 270 | { |
| 271 | kern_unmount(ns->proc_mnt); |
| 272 | } |