| 1 | // SPDX-License-Identifier: GPL-2.0-only |
| 2 | /* Copyright (c) 2025 Christian Brauner <brauner@kernel.org> */ |
| 3 | |
| 4 | #include <linux/nstree.h> |
| 5 | #include <linux/proc_ns.h> |
| 6 | #include <linux/rculist.h> |
| 7 | #include <linux/vfsdebug.h> |
| 8 | #include <linux/syscalls.h> |
| 9 | #include <linux/user_namespace.h> |
| 10 | |
| 11 | static __cacheline_aligned_in_smp DEFINE_SEQLOCK(ns_tree_lock); |
| 12 | |
| 13 | DEFINE_LOCK_GUARD_0(ns_tree_writer, |
| 14 | write_seqlock(&ns_tree_lock), |
| 15 | write_sequnlock(&ns_tree_lock)) |
| 16 | |
| 17 | DEFINE_LOCK_GUARD_0(ns_tree_locked_reader, |
| 18 | read_seqlock_excl(&ns_tree_lock), |
| 19 | read_sequnlock_excl(&ns_tree_lock)) |
| 20 | |
| 21 | static struct ns_tree_root ns_unified_root = { /* protected by ns_tree_lock */ |
| 22 | .ns_rb = RB_ROOT, |
| 23 | .ns_list_head = LIST_HEAD_INIT(ns_unified_root.ns_list_head), |
| 24 | }; |
| 25 | |
| 26 | struct ns_tree_root mnt_ns_tree = { |
| 27 | .ns_rb = RB_ROOT, |
| 28 | .ns_list_head = LIST_HEAD_INIT(mnt_ns_tree.ns_list_head), |
| 29 | }; |
| 30 | |
| 31 | struct ns_tree_root net_ns_tree = { |
| 32 | .ns_rb = RB_ROOT, |
| 33 | .ns_list_head = LIST_HEAD_INIT(net_ns_tree.ns_list_head), |
| 34 | }; |
| 35 | EXPORT_SYMBOL_GPL(net_ns_tree); |
| 36 | |
| 37 | struct ns_tree_root uts_ns_tree = { |
| 38 | .ns_rb = RB_ROOT, |
| 39 | .ns_list_head = LIST_HEAD_INIT(uts_ns_tree.ns_list_head), |
| 40 | }; |
| 41 | |
| 42 | struct ns_tree_root user_ns_tree = { |
| 43 | .ns_rb = RB_ROOT, |
| 44 | .ns_list_head = LIST_HEAD_INIT(user_ns_tree.ns_list_head), |
| 45 | }; |
| 46 | |
| 47 | struct ns_tree_root ipc_ns_tree = { |
| 48 | .ns_rb = RB_ROOT, |
| 49 | .ns_list_head = LIST_HEAD_INIT(ipc_ns_tree.ns_list_head), |
| 50 | }; |
| 51 | |
| 52 | struct ns_tree_root pid_ns_tree = { |
| 53 | .ns_rb = RB_ROOT, |
| 54 | .ns_list_head = LIST_HEAD_INIT(pid_ns_tree.ns_list_head), |
| 55 | }; |
| 56 | |
| 57 | struct ns_tree_root cgroup_ns_tree = { |
| 58 | .ns_rb = RB_ROOT, |
| 59 | .ns_list_head = LIST_HEAD_INIT(cgroup_ns_tree.ns_list_head), |
| 60 | }; |
| 61 | |
| 62 | struct ns_tree_root time_ns_tree = { |
| 63 | .ns_rb = RB_ROOT, |
| 64 | .ns_list_head = LIST_HEAD_INIT(time_ns_tree.ns_list_head), |
| 65 | }; |
| 66 | |
| 67 | /** |
| 68 | * ns_tree_node_init - Initialize a namespace tree node |
| 69 | * @node: The node to initialize |
| 70 | * |
| 71 | * Initializes both the rbtree node and list entry. |
| 72 | */ |
| 73 | void ns_tree_node_init(struct ns_tree_node *node) |
| 74 | { |
| 75 | RB_CLEAR_NODE(&node->ns_node); |
| 76 | INIT_LIST_HEAD(list: &node->ns_list_entry); |
| 77 | } |
| 78 | |
| 79 | /** |
| 80 | * ns_tree_root_init - Initialize a namespace tree root |
| 81 | * @root: The root to initialize |
| 82 | * |
| 83 | * Initializes both the rbtree root and list head. |
| 84 | */ |
| 85 | void ns_tree_root_init(struct ns_tree_root *root) |
| 86 | { |
| 87 | root->ns_rb = RB_ROOT; |
| 88 | INIT_LIST_HEAD(list: &root->ns_list_head); |
| 89 | } |
| 90 | |
| 91 | /** |
| 92 | * ns_tree_node_empty - Check if a namespace tree node is empty |
| 93 | * @node: The node to check |
| 94 | * |
| 95 | * Returns true if the node is not in any tree. |
| 96 | */ |
| 97 | bool ns_tree_node_empty(const struct ns_tree_node *node) |
| 98 | { |
| 99 | return RB_EMPTY_NODE(&node->ns_node); |
| 100 | } |
| 101 | |
| 102 | /** |
| 103 | * ns_tree_node_add - Add a node to a namespace tree |
| 104 | * @node: The node to add |
| 105 | * @root: The tree root to add to |
| 106 | * @cmp: Comparison function for rbtree insertion |
| 107 | * |
| 108 | * Adds the node to both the rbtree and the list, maintaining sorted order. |
| 109 | * The list is maintained in the same order as the rbtree to enable efficient |
| 110 | * iteration. |
| 111 | * |
| 112 | * Returns: NULL if insertion succeeded, existing node if duplicate found |
| 113 | */ |
| 114 | struct rb_node *ns_tree_node_add(struct ns_tree_node *node, |
| 115 | struct ns_tree_root *root, |
| 116 | int (*cmp)(struct rb_node *, const struct rb_node *)) |
| 117 | { |
| 118 | struct rb_node *ret, *prev; |
| 119 | |
| 120 | /* Add to rbtree */ |
| 121 | ret = rb_find_add_rcu(node: &node->ns_node, tree: &root->ns_rb, cmp); |
| 122 | |
| 123 | /* Add to list in sorted order */ |
| 124 | prev = rb_prev(&node->ns_node); |
| 125 | if (!prev) { |
| 126 | /* No previous node, add at head */ |
| 127 | list_add_rcu(new: &node->ns_list_entry, head: &root->ns_list_head); |
| 128 | } else { |
| 129 | /* Add after previous node */ |
| 130 | struct ns_tree_node *prev_node; |
| 131 | prev_node = rb_entry(prev, struct ns_tree_node, ns_node); |
| 132 | list_add_rcu(new: &node->ns_list_entry, head: &prev_node->ns_list_entry); |
| 133 | } |
| 134 | |
| 135 | return ret; |
| 136 | } |
| 137 | |
| 138 | /** |
| 139 | * ns_tree_node_del - Remove a node from a namespace tree |
| 140 | * @node: The node to remove |
| 141 | * @root: The tree root to remove from |
| 142 | * |
| 143 | * Removes the node from both the rbtree and the list atomically. |
| 144 | */ |
| 145 | void ns_tree_node_del(struct ns_tree_node *node, struct ns_tree_root *root) |
| 146 | { |
| 147 | rb_erase(&node->ns_node, &root->ns_rb); |
| 148 | RB_CLEAR_NODE(&node->ns_node); |
| 149 | list_bidir_del_rcu(entry: &node->ns_list_entry); |
| 150 | } |
| 151 | |
| 152 | static inline struct ns_common *node_to_ns(const struct rb_node *node) |
| 153 | { |
| 154 | if (!node) |
| 155 | return NULL; |
| 156 | return rb_entry(node, struct ns_common, ns_tree_node.ns_node); |
| 157 | } |
| 158 | |
| 159 | static inline struct ns_common *node_to_ns_unified(const struct rb_node *node) |
| 160 | { |
| 161 | if (!node) |
| 162 | return NULL; |
| 163 | return rb_entry(node, struct ns_common, ns_unified_node.ns_node); |
| 164 | } |
| 165 | |
| 166 | static inline struct ns_common *node_to_ns_owner(const struct rb_node *node) |
| 167 | { |
| 168 | if (!node) |
| 169 | return NULL; |
| 170 | return rb_entry(node, struct ns_common, ns_owner_node.ns_node); |
| 171 | } |
| 172 | |
| 173 | static int ns_id_cmp(u64 id_a, u64 id_b) |
| 174 | { |
| 175 | if (id_a < id_b) |
| 176 | return -1; |
| 177 | if (id_a > id_b) |
| 178 | return 1; |
| 179 | return 0; |
| 180 | } |
| 181 | |
| 182 | static int ns_cmp(struct rb_node *a, const struct rb_node *b) |
| 183 | { |
| 184 | return ns_id_cmp(id_a: node_to_ns(node: a)->ns_id, id_b: node_to_ns(node: b)->ns_id); |
| 185 | } |
| 186 | |
| 187 | static int ns_cmp_unified(struct rb_node *a, const struct rb_node *b) |
| 188 | { |
| 189 | return ns_id_cmp(id_a: node_to_ns_unified(node: a)->ns_id, id_b: node_to_ns_unified(node: b)->ns_id); |
| 190 | } |
| 191 | |
| 192 | static int ns_cmp_owner(struct rb_node *a, const struct rb_node *b) |
| 193 | { |
| 194 | return ns_id_cmp(id_a: node_to_ns_owner(node: a)->ns_id, id_b: node_to_ns_owner(node: b)->ns_id); |
| 195 | } |
| 196 | |
| 197 | void __ns_tree_add_raw(struct ns_common *ns, struct ns_tree_root *ns_tree) |
| 198 | { |
| 199 | struct rb_node *node; |
| 200 | const struct proc_ns_operations *ops = ns->ops; |
| 201 | |
| 202 | VFS_WARN_ON_ONCE(!ns->ns_id); |
| 203 | |
| 204 | guard(ns_tree_writer)(); |
| 205 | |
| 206 | /* Add to per-type tree and list */ |
| 207 | node = ns_tree_node_add(node: &ns->ns_tree_node, root: ns_tree, cmp: ns_cmp); |
| 208 | |
| 209 | /* Add to unified tree and list */ |
| 210 | ns_tree_node_add(node: &ns->ns_unified_node, root: &ns_unified_root, cmp: ns_cmp_unified); |
| 211 | |
| 212 | /* Add to owner's tree if applicable */ |
| 213 | if (ops) { |
| 214 | struct user_namespace *user_ns; |
| 215 | |
| 216 | VFS_WARN_ON_ONCE(!ops->owner); |
| 217 | user_ns = ops->owner(ns); |
| 218 | if (user_ns) { |
| 219 | struct ns_common *owner = &user_ns->ns; |
| 220 | VFS_WARN_ON_ONCE(owner->ns_type != CLONE_NEWUSER); |
| 221 | |
| 222 | /* Insert into owner's tree and list */ |
| 223 | ns_tree_node_add(node: &ns->ns_owner_node, root: &owner->ns_owner_root, cmp: ns_cmp_owner); |
| 224 | } else { |
| 225 | /* Only the initial user namespace doesn't have an owner. */ |
| 226 | VFS_WARN_ON_ONCE(ns != to_ns_common(&init_user_ns)); |
| 227 | } |
| 228 | } |
| 229 | |
| 230 | VFS_WARN_ON_ONCE(node); |
| 231 | } |
| 232 | |
| 233 | void __ns_tree_remove(struct ns_common *ns, struct ns_tree_root *ns_tree) |
| 234 | { |
| 235 | const struct proc_ns_operations *ops = ns->ops; |
| 236 | struct user_namespace *user_ns; |
| 237 | |
| 238 | VFS_WARN_ON_ONCE(ns_tree_node_empty(&ns->ns_tree_node)); |
| 239 | VFS_WARN_ON_ONCE(list_empty(&ns->ns_tree_node.ns_list_entry)); |
| 240 | |
| 241 | write_seqlock(sl: &ns_tree_lock); |
| 242 | |
| 243 | /* Remove from per-type tree and list */ |
| 244 | ns_tree_node_del(node: &ns->ns_tree_node, root: ns_tree); |
| 245 | |
| 246 | /* Remove from unified tree and list */ |
| 247 | ns_tree_node_del(node: &ns->ns_unified_node, root: &ns_unified_root); |
| 248 | |
| 249 | /* Remove from owner's tree if applicable */ |
| 250 | if (ops) { |
| 251 | user_ns = ops->owner(ns); |
| 252 | if (user_ns) { |
| 253 | struct ns_common *owner = &user_ns->ns; |
| 254 | ns_tree_node_del(node: &ns->ns_owner_node, root: &owner->ns_owner_root); |
| 255 | } |
| 256 | } |
| 257 | |
| 258 | write_sequnlock(sl: &ns_tree_lock); |
| 259 | } |
| 260 | EXPORT_SYMBOL_GPL(__ns_tree_remove); |
| 261 | |
| 262 | static int ns_find(const void *key, const struct rb_node *node) |
| 263 | { |
| 264 | const u64 ns_id = *(u64 *)key; |
| 265 | const struct ns_common *ns = node_to_ns(node); |
| 266 | |
| 267 | if (ns_id < ns->ns_id) |
| 268 | return -1; |
| 269 | if (ns_id > ns->ns_id) |
| 270 | return 1; |
| 271 | return 0; |
| 272 | } |
| 273 | |
| 274 | static int ns_find_unified(const void *key, const struct rb_node *node) |
| 275 | { |
| 276 | const u64 ns_id = *(u64 *)key; |
| 277 | const struct ns_common *ns = node_to_ns_unified(node); |
| 278 | |
| 279 | if (ns_id < ns->ns_id) |
| 280 | return -1; |
| 281 | if (ns_id > ns->ns_id) |
| 282 | return 1; |
| 283 | return 0; |
| 284 | } |
| 285 | |
| 286 | static struct ns_tree_root *ns_tree_from_type(int ns_type) |
| 287 | { |
| 288 | switch (ns_type) { |
| 289 | case CLONE_NEWCGROUP: |
| 290 | return &cgroup_ns_tree; |
| 291 | case CLONE_NEWIPC: |
| 292 | return &ipc_ns_tree; |
| 293 | case CLONE_NEWNS: |
| 294 | return &mnt_ns_tree; |
| 295 | case CLONE_NEWNET: |
| 296 | return &net_ns_tree; |
| 297 | case CLONE_NEWPID: |
| 298 | return &pid_ns_tree; |
| 299 | case CLONE_NEWUSER: |
| 300 | return &user_ns_tree; |
| 301 | case CLONE_NEWUTS: |
| 302 | return &uts_ns_tree; |
| 303 | case CLONE_NEWTIME: |
| 304 | return &time_ns_tree; |
| 305 | } |
| 306 | |
| 307 | return NULL; |
| 308 | } |
| 309 | |
| 310 | static struct ns_common *__ns_unified_tree_lookup_rcu(u64 ns_id) |
| 311 | { |
| 312 | struct rb_node *node; |
| 313 | unsigned int seq; |
| 314 | |
| 315 | do { |
| 316 | seq = read_seqbegin(sl: &ns_tree_lock); |
| 317 | node = rb_find_rcu(key: &ns_id, tree: &ns_unified_root.ns_rb, cmp: ns_find_unified); |
| 318 | if (node) |
| 319 | break; |
| 320 | } while (read_seqretry(sl: &ns_tree_lock, start: seq)); |
| 321 | |
| 322 | return node_to_ns_unified(node); |
| 323 | } |
| 324 | |
| 325 | static struct ns_common *__ns_tree_lookup_rcu(u64 ns_id, int ns_type) |
| 326 | { |
| 327 | struct ns_tree_root *ns_tree; |
| 328 | struct rb_node *node; |
| 329 | unsigned int seq; |
| 330 | |
| 331 | ns_tree = ns_tree_from_type(ns_type); |
| 332 | if (!ns_tree) |
| 333 | return NULL; |
| 334 | |
| 335 | do { |
| 336 | seq = read_seqbegin(sl: &ns_tree_lock); |
| 337 | node = rb_find_rcu(key: &ns_id, tree: &ns_tree->ns_rb, cmp: ns_find); |
| 338 | if (node) |
| 339 | break; |
| 340 | } while (read_seqretry(sl: &ns_tree_lock, start: seq)); |
| 341 | |
| 342 | return node_to_ns(node); |
| 343 | } |
| 344 | |
| 345 | struct ns_common *ns_tree_lookup_rcu(u64 ns_id, int ns_type) |
| 346 | { |
| 347 | RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "suspicious ns_tree_lookup_rcu() usage" ); |
| 348 | |
| 349 | if (ns_type) |
| 350 | return __ns_tree_lookup_rcu(ns_id, ns_type); |
| 351 | |
| 352 | return __ns_unified_tree_lookup_rcu(ns_id); |
| 353 | } |
| 354 | |
| 355 | /** |
| 356 | * __ns_tree_adjoined_rcu - find the next/previous namespace in the same |
| 357 | * tree |
| 358 | * @ns: namespace to start from |
| 359 | * @ns_tree: namespace tree to search in |
| 360 | * @previous: if true find the previous namespace, otherwise the next |
| 361 | * |
| 362 | * Find the next or previous namespace in the same tree as @ns. If |
| 363 | * there is no next/previous namespace, -ENOENT is returned. |
| 364 | */ |
| 365 | struct ns_common *__ns_tree_adjoined_rcu(struct ns_common *ns, |
| 366 | struct ns_tree_root *ns_tree, bool previous) |
| 367 | { |
| 368 | struct list_head *list; |
| 369 | |
| 370 | RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "suspicious ns_tree_adjoined_rcu() usage" ); |
| 371 | |
| 372 | if (previous) |
| 373 | list = rcu_dereference(list_bidir_prev_rcu(&ns->ns_tree_node.ns_list_entry)); |
| 374 | else |
| 375 | list = rcu_dereference(list_next_rcu(&ns->ns_tree_node.ns_list_entry)); |
| 376 | if (list_is_head(list, head: &ns_tree->ns_list_head)) |
| 377 | return ERR_PTR(error: -ENOENT); |
| 378 | |
| 379 | return list_entry_rcu(list, struct ns_common, ns_tree_node.ns_list_entry); |
| 380 | } |
| 381 | |
| 382 | /** |
| 383 | * __ns_tree_gen_id - generate a new namespace id |
| 384 | * @ns: namespace to generate id for |
| 385 | * @id: if non-zero, this is the initial namespace and this is a fixed id |
| 386 | * |
| 387 | * Generates a new namespace id and assigns it to the namespace. All |
| 388 | * namespaces types share the same id space and thus can be compared |
| 389 | * directly. IOW, when two ids of two namespace are equal, they are |
| 390 | * identical. |
| 391 | */ |
| 392 | u64 __ns_tree_gen_id(struct ns_common *ns, u64 id) |
| 393 | { |
| 394 | static atomic64_t namespace_cookie = ATOMIC64_INIT(NS_LAST_INIT_ID + 1); |
| 395 | |
| 396 | if (id) |
| 397 | ns->ns_id = id; |
| 398 | else |
| 399 | ns->ns_id = atomic64_inc_return(v: &namespace_cookie); |
| 400 | return ns->ns_id; |
| 401 | } |
| 402 | |
| 403 | struct klistns { |
| 404 | u64 __user *uns_ids; |
| 405 | u32 nr_ns_ids; |
| 406 | u64 last_ns_id; |
| 407 | u64 user_ns_id; |
| 408 | u32 ns_type; |
| 409 | struct user_namespace *user_ns; |
| 410 | bool userns_capable; |
| 411 | struct ns_common *first_ns; |
| 412 | }; |
| 413 | |
| 414 | static void __free_klistns_free(const struct klistns *kls) |
| 415 | { |
| 416 | if (kls->user_ns_id != LISTNS_CURRENT_USER) |
| 417 | put_user_ns(ns: kls->user_ns); |
| 418 | if (kls->first_ns && kls->first_ns->ops) |
| 419 | kls->first_ns->ops->put(kls->first_ns); |
| 420 | } |
| 421 | |
| 422 | #define NS_ALL (PID_NS | USER_NS | MNT_NS | UTS_NS | IPC_NS | NET_NS | CGROUP_NS | TIME_NS) |
| 423 | |
| 424 | static int copy_ns_id_req(const struct ns_id_req __user *req, |
| 425 | struct ns_id_req *kreq) |
| 426 | { |
| 427 | int ret; |
| 428 | size_t usize; |
| 429 | |
| 430 | BUILD_BUG_ON(sizeof(struct ns_id_req) != NS_ID_REQ_SIZE_VER0); |
| 431 | |
| 432 | ret = get_user(usize, &req->size); |
| 433 | if (ret) |
| 434 | return -EFAULT; |
| 435 | if (unlikely(usize > PAGE_SIZE)) |
| 436 | return -E2BIG; |
| 437 | if (unlikely(usize < NS_ID_REQ_SIZE_VER0)) |
| 438 | return -EINVAL; |
| 439 | memset(kreq, 0, sizeof(*kreq)); |
| 440 | ret = copy_struct_from_user(dst: kreq, ksize: sizeof(*kreq), src: req, usize); |
| 441 | if (ret) |
| 442 | return ret; |
| 443 | if (kreq->spare != 0) |
| 444 | return -EINVAL; |
| 445 | if (kreq->ns_type & ~NS_ALL) |
| 446 | return -EOPNOTSUPP; |
| 447 | return 0; |
| 448 | } |
| 449 | |
| 450 | static inline int prepare_klistns(struct klistns *kls, struct ns_id_req *kreq, |
| 451 | u64 __user *ns_ids, size_t nr_ns_ids) |
| 452 | { |
| 453 | kls->last_ns_id = kreq->ns_id; |
| 454 | kls->user_ns_id = kreq->user_ns_id; |
| 455 | kls->nr_ns_ids = nr_ns_ids; |
| 456 | kls->ns_type = kreq->ns_type; |
| 457 | kls->uns_ids = ns_ids; |
| 458 | return 0; |
| 459 | } |
| 460 | |
| 461 | /* |
| 462 | * Lookup a namespace owned by owner with id >= ns_id. |
| 463 | * Returns the namespace with the smallest id that is >= ns_id. |
| 464 | */ |
| 465 | static struct ns_common *lookup_ns_owner_at(u64 ns_id, struct ns_common *owner) |
| 466 | { |
| 467 | struct ns_common *ret = NULL; |
| 468 | struct rb_node *node; |
| 469 | |
| 470 | VFS_WARN_ON_ONCE(owner->ns_type != CLONE_NEWUSER); |
| 471 | |
| 472 | guard(ns_tree_locked_reader)(); |
| 473 | |
| 474 | node = owner->ns_owner_root.ns_rb.rb_node; |
| 475 | while (node) { |
| 476 | struct ns_common *ns; |
| 477 | |
| 478 | ns = node_to_ns_owner(node); |
| 479 | if (ns_id <= ns->ns_id) { |
| 480 | ret = ns; |
| 481 | if (ns_id == ns->ns_id) |
| 482 | break; |
| 483 | node = node->rb_left; |
| 484 | } else { |
| 485 | node = node->rb_right; |
| 486 | } |
| 487 | } |
| 488 | |
| 489 | if (ret) |
| 490 | ret = ns_get_unless_inactive(ns: ret); |
| 491 | return ret; |
| 492 | } |
| 493 | |
| 494 | static struct ns_common *lookup_ns_id(u64 mnt_ns_id, int ns_type) |
| 495 | { |
| 496 | struct ns_common *ns; |
| 497 | |
| 498 | guard(rcu)(); |
| 499 | ns = ns_tree_lookup_rcu(ns_id: mnt_ns_id, ns_type); |
| 500 | if (!ns) |
| 501 | return NULL; |
| 502 | |
| 503 | if (!ns_get_unless_inactive(ns)) |
| 504 | return NULL; |
| 505 | |
| 506 | return ns; |
| 507 | } |
| 508 | |
| 509 | static inline bool __must_check ns_requested(const struct klistns *kls, |
| 510 | const struct ns_common *ns) |
| 511 | { |
| 512 | return !kls->ns_type || (kls->ns_type & ns->ns_type); |
| 513 | } |
| 514 | |
| 515 | static inline bool __must_check may_list_ns(const struct klistns *kls, |
| 516 | struct ns_common *ns) |
| 517 | { |
| 518 | if (kls->user_ns) { |
| 519 | if (kls->userns_capable) |
| 520 | return true; |
| 521 | } else { |
| 522 | struct ns_common *owner; |
| 523 | struct user_namespace *user_ns; |
| 524 | |
| 525 | owner = ns_owner(ns); |
| 526 | if (owner) |
| 527 | user_ns = to_user_ns(ns: owner); |
| 528 | else |
| 529 | user_ns = &init_user_ns; |
| 530 | if (ns_capable_noaudit(ns: user_ns, CAP_SYS_ADMIN)) |
| 531 | return true; |
| 532 | } |
| 533 | |
| 534 | if (is_current_namespace(ns)) |
| 535 | return true; |
| 536 | |
| 537 | if (ns->ns_type != CLONE_NEWUSER) |
| 538 | return false; |
| 539 | |
| 540 | if (ns_capable_noaudit(ns: to_user_ns(ns), CAP_SYS_ADMIN)) |
| 541 | return true; |
| 542 | |
| 543 | return false; |
| 544 | } |
| 545 | |
| 546 | static inline void ns_put(struct ns_common *ns) |
| 547 | { |
| 548 | if (ns && ns->ops) |
| 549 | ns->ops->put(ns); |
| 550 | } |
| 551 | |
| 552 | DEFINE_FREE(ns_put, struct ns_common *, if (!IS_ERR_OR_NULL(_T)) ns_put(_T)) |
| 553 | |
| 554 | static inline struct ns_common *__must_check legitimize_ns(const struct klistns *kls, |
| 555 | struct ns_common *candidate) |
| 556 | { |
| 557 | struct ns_common *ns __free(ns_put) = NULL; |
| 558 | |
| 559 | if (!ns_requested(kls, ns: candidate)) |
| 560 | return NULL; |
| 561 | |
| 562 | ns = ns_get_unless_inactive(ns: candidate); |
| 563 | if (!ns) |
| 564 | return NULL; |
| 565 | |
| 566 | if (!may_list_ns(kls, ns)) |
| 567 | return NULL; |
| 568 | |
| 569 | return no_free_ptr(ns); |
| 570 | } |
| 571 | |
| 572 | static ssize_t do_listns_userns(struct klistns *kls) |
| 573 | { |
| 574 | u64 __user *ns_ids = kls->uns_ids; |
| 575 | size_t nr_ns_ids = kls->nr_ns_ids; |
| 576 | struct ns_common *ns = NULL, *first_ns = NULL, *prev = NULL; |
| 577 | const struct list_head *head; |
| 578 | ssize_t ret; |
| 579 | |
| 580 | VFS_WARN_ON_ONCE(!kls->user_ns_id); |
| 581 | |
| 582 | if (kls->user_ns_id == LISTNS_CURRENT_USER) |
| 583 | ns = to_ns_common(current_user_ns()); |
| 584 | else if (kls->user_ns_id) |
| 585 | ns = lookup_ns_id(mnt_ns_id: kls->user_ns_id, CLONE_NEWUSER); |
| 586 | if (!ns) |
| 587 | return -EINVAL; |
| 588 | kls->user_ns = to_user_ns(ns); |
| 589 | |
| 590 | /* |
| 591 | * Use the rbtree to find the first namespace we care about and |
| 592 | * then use it's list entry to iterate from there. |
| 593 | */ |
| 594 | if (kls->last_ns_id) { |
| 595 | kls->first_ns = lookup_ns_owner_at(ns_id: kls->last_ns_id + 1, owner: ns); |
| 596 | if (!kls->first_ns) |
| 597 | return -ENOENT; |
| 598 | first_ns = kls->first_ns; |
| 599 | } |
| 600 | |
| 601 | ret = 0; |
| 602 | head = &to_ns_common(kls->user_ns)->ns_owner_root.ns_list_head; |
| 603 | kls->userns_capable = ns_capable_noaudit(ns: kls->user_ns, CAP_SYS_ADMIN); |
| 604 | |
| 605 | rcu_read_lock(); |
| 606 | |
| 607 | if (!first_ns) |
| 608 | first_ns = list_entry_rcu(head->next, typeof(*first_ns), ns_owner_node.ns_list_entry); |
| 609 | |
| 610 | ns = first_ns; |
| 611 | list_for_each_entry_from_rcu(ns, head, ns_owner_node.ns_list_entry) { |
| 612 | struct ns_common *valid; |
| 613 | |
| 614 | if (!nr_ns_ids) |
| 615 | break; |
| 616 | |
| 617 | valid = legitimize_ns(kls, candidate: ns); |
| 618 | if (!valid) |
| 619 | continue; |
| 620 | |
| 621 | rcu_read_unlock(); |
| 622 | |
| 623 | ns_put(ns: prev); |
| 624 | prev = valid; |
| 625 | |
| 626 | if (put_user(valid->ns_id, ns_ids + ret)) { |
| 627 | ns_put(ns: prev); |
| 628 | return -EFAULT; |
| 629 | } |
| 630 | |
| 631 | nr_ns_ids--; |
| 632 | ret++; |
| 633 | |
| 634 | rcu_read_lock(); |
| 635 | } |
| 636 | |
| 637 | rcu_read_unlock(); |
| 638 | ns_put(ns: prev); |
| 639 | return ret; |
| 640 | } |
| 641 | |
| 642 | /* |
| 643 | * Lookup a namespace with id >= ns_id in either the unified tree or a type-specific tree. |
| 644 | * Returns the namespace with the smallest id that is >= ns_id. |
| 645 | */ |
| 646 | static struct ns_common *lookup_ns_id_at(u64 ns_id, int ns_type) |
| 647 | { |
| 648 | struct ns_common *ret = NULL; |
| 649 | struct ns_tree_root *ns_tree = NULL; |
| 650 | struct rb_node *node; |
| 651 | |
| 652 | if (ns_type) { |
| 653 | ns_tree = ns_tree_from_type(ns_type); |
| 654 | if (!ns_tree) |
| 655 | return NULL; |
| 656 | } |
| 657 | |
| 658 | guard(ns_tree_locked_reader)(); |
| 659 | |
| 660 | if (ns_tree) |
| 661 | node = ns_tree->ns_rb.rb_node; |
| 662 | else |
| 663 | node = ns_unified_root.ns_rb.rb_node; |
| 664 | |
| 665 | while (node) { |
| 666 | struct ns_common *ns; |
| 667 | |
| 668 | if (ns_type) |
| 669 | ns = node_to_ns(node); |
| 670 | else |
| 671 | ns = node_to_ns_unified(node); |
| 672 | |
| 673 | if (ns_id <= ns->ns_id) { |
| 674 | if (ns_type) |
| 675 | ret = node_to_ns(node); |
| 676 | else |
| 677 | ret = node_to_ns_unified(node); |
| 678 | if (ns_id == ns->ns_id) |
| 679 | break; |
| 680 | node = node->rb_left; |
| 681 | } else { |
| 682 | node = node->rb_right; |
| 683 | } |
| 684 | } |
| 685 | |
| 686 | if (ret) |
| 687 | ret = ns_get_unless_inactive(ns: ret); |
| 688 | return ret; |
| 689 | } |
| 690 | |
| 691 | static inline struct ns_common *first_ns_common(const struct list_head *head, |
| 692 | struct ns_tree_root *ns_tree) |
| 693 | { |
| 694 | if (ns_tree) |
| 695 | return list_entry_rcu(head->next, struct ns_common, ns_tree_node.ns_list_entry); |
| 696 | return list_entry_rcu(head->next, struct ns_common, ns_unified_node.ns_list_entry); |
| 697 | } |
| 698 | |
| 699 | static inline struct ns_common *next_ns_common(struct ns_common *ns, |
| 700 | struct ns_tree_root *ns_tree) |
| 701 | { |
| 702 | if (ns_tree) |
| 703 | return list_entry_rcu(ns->ns_tree_node.ns_list_entry.next, struct ns_common, ns_tree_node.ns_list_entry); |
| 704 | return list_entry_rcu(ns->ns_unified_node.ns_list_entry.next, struct ns_common, ns_unified_node.ns_list_entry); |
| 705 | } |
| 706 | |
| 707 | static inline bool ns_common_is_head(struct ns_common *ns, |
| 708 | const struct list_head *head, |
| 709 | struct ns_tree_root *ns_tree) |
| 710 | { |
| 711 | if (ns_tree) |
| 712 | return &ns->ns_tree_node.ns_list_entry == head; |
| 713 | return &ns->ns_unified_node.ns_list_entry == head; |
| 714 | } |
| 715 | |
| 716 | static ssize_t do_listns(struct klistns *kls) |
| 717 | { |
| 718 | u64 __user *ns_ids = kls->uns_ids; |
| 719 | size_t nr_ns_ids = kls->nr_ns_ids; |
| 720 | struct ns_common *ns, *first_ns = NULL, *prev = NULL; |
| 721 | struct ns_tree_root *ns_tree = NULL; |
| 722 | const struct list_head *head; |
| 723 | u32 ns_type; |
| 724 | ssize_t ret; |
| 725 | |
| 726 | if (hweight32(kls->ns_type) == 1) |
| 727 | ns_type = kls->ns_type; |
| 728 | else |
| 729 | ns_type = 0; |
| 730 | |
| 731 | if (ns_type) { |
| 732 | ns_tree = ns_tree_from_type(ns_type); |
| 733 | if (!ns_tree) |
| 734 | return -EINVAL; |
| 735 | } |
| 736 | |
| 737 | if (kls->last_ns_id) { |
| 738 | kls->first_ns = lookup_ns_id_at(ns_id: kls->last_ns_id + 1, ns_type); |
| 739 | if (!kls->first_ns) |
| 740 | return -ENOENT; |
| 741 | first_ns = kls->first_ns; |
| 742 | } |
| 743 | |
| 744 | ret = 0; |
| 745 | if (ns_tree) |
| 746 | head = &ns_tree->ns_list_head; |
| 747 | else |
| 748 | head = &ns_unified_root.ns_list_head; |
| 749 | |
| 750 | rcu_read_lock(); |
| 751 | |
| 752 | if (!first_ns) |
| 753 | first_ns = first_ns_common(head, ns_tree); |
| 754 | |
| 755 | for (ns = first_ns; !ns_common_is_head(ns, head, ns_tree) && nr_ns_ids; |
| 756 | ns = next_ns_common(ns, ns_tree)) { |
| 757 | struct ns_common *valid; |
| 758 | |
| 759 | valid = legitimize_ns(kls, candidate: ns); |
| 760 | if (!valid) |
| 761 | continue; |
| 762 | |
| 763 | rcu_read_unlock(); |
| 764 | |
| 765 | ns_put(ns: prev); |
| 766 | prev = valid; |
| 767 | |
| 768 | if (put_user(valid->ns_id, ns_ids + ret)) { |
| 769 | ns_put(ns: prev); |
| 770 | return -EFAULT; |
| 771 | } |
| 772 | |
| 773 | nr_ns_ids--; |
| 774 | ret++; |
| 775 | |
| 776 | rcu_read_lock(); |
| 777 | } |
| 778 | |
| 779 | rcu_read_unlock(); |
| 780 | ns_put(ns: prev); |
| 781 | return ret; |
| 782 | } |
| 783 | |
| 784 | SYSCALL_DEFINE4(listns, const struct ns_id_req __user *, req, |
| 785 | u64 __user *, ns_ids, size_t, nr_ns_ids, unsigned int, flags) |
| 786 | { |
| 787 | struct klistns klns __free(klistns_free) = {}; |
| 788 | const size_t maxcount = 1000000; |
| 789 | struct ns_id_req kreq; |
| 790 | ssize_t ret; |
| 791 | |
| 792 | if (flags) |
| 793 | return -EINVAL; |
| 794 | |
| 795 | if (unlikely(nr_ns_ids > maxcount)) |
| 796 | return -EOVERFLOW; |
| 797 | |
| 798 | if (!access_ok(ns_ids, nr_ns_ids * sizeof(*ns_ids))) |
| 799 | return -EFAULT; |
| 800 | |
| 801 | ret = copy_ns_id_req(req, kreq: &kreq); |
| 802 | if (ret) |
| 803 | return ret; |
| 804 | |
| 805 | ret = prepare_klistns(kls: &klns, kreq: &kreq, ns_ids, nr_ns_ids); |
| 806 | if (ret) |
| 807 | return ret; |
| 808 | |
| 809 | if (kreq.user_ns_id) |
| 810 | return do_listns_userns(kls: &klns); |
| 811 | |
| 812 | return do_listns(kls: &klns); |
| 813 | } |
| 814 | |