1 // SPDX-License-Identifier: GPL-2.0
3 * linux/ipc/namespace.c
4 * Copyright (C) 2006 Pavel Emelyanov <xemul@openvz.org> OpenVZ, SWsoft Inc.
9 #include <linux/ipc_namespace.h>
10 #include <linux/rcupdate.h>
11 #include <linux/nsproxy.h>
12 #include <linux/slab.h>
13 #include <linux/cred.h>
15 #include <linux/mount.h>
16 #include <linux/user_namespace.h>
17 #include <linux/proc_ns.h>
18 #include <linux/sched/task.h>
23 * The work queue is used to avoid the cost of synchronize_rcu in kern_unmount.
25 static void free_ipc(struct work_struct *unused);
26 static DECLARE_WORK(free_ipc_work, free_ipc);
28 static struct ucounts *inc_ipc_namespaces(struct user_namespace *ns)
30 return inc_ucount(ns, current_euid(), UCOUNT_IPC_NAMESPACES);
33 static void dec_ipc_namespaces(struct ucounts *ucounts)
35 dec_ucount(ucounts, UCOUNT_IPC_NAMESPACES);
38 static struct ipc_namespace *create_ipc_ns(struct user_namespace *user_ns,
39 struct ipc_namespace *old_ns)
41 struct ipc_namespace *ns;
42 struct ucounts *ucounts;
47 ucounts = inc_ipc_namespaces(user_ns);
50 * IPC namespaces are freed asynchronously, by free_ipc_work.
51 * If frees were pending, flush_work will wait, and
52 * return true. Fail the allocation if no frees are pending.
54 if (flush_work(&free_ipc_work))
60 ns = kzalloc(sizeof(struct ipc_namespace), GFP_KERNEL_ACCOUNT);
64 err = ns_alloc_inum(&ns->ns);
67 ns->ns.ops = &ipcns_operations;
69 refcount_set(&ns->ns.count, 1);
70 ns->user_ns = get_user_ns(user_ns);
71 ns->ucounts = ucounts;
78 if (!setup_mq_sysctls(ns))
81 if (!setup_ipc_sysctls(ns))
84 err = msg_init_ns(ns);
94 retire_mq_sysctls(ns);
97 put_user_ns(ns->user_ns);
98 ns_free_inum(&ns->ns);
102 dec_ipc_namespaces(ucounts);
107 struct ipc_namespace *copy_ipcs(unsigned long flags,
108 struct user_namespace *user_ns, struct ipc_namespace *ns)
110 if (!(flags & CLONE_NEWIPC))
111 return get_ipc_ns(ns);
112 return create_ipc_ns(user_ns, ns);
116 * free_ipcs - free all ipcs of one type
117 * @ns: the namespace to remove the ipcs from
118 * @ids: the table of ipcs to free
119 * @free: the function called to free each individual ipc
121 * Called for each kind of ipc when an ipc_namespace exits.
123 void free_ipcs(struct ipc_namespace *ns, struct ipc_ids *ids,
124 void (*free)(struct ipc_namespace *, struct kern_ipc_perm *))
126 struct kern_ipc_perm *perm;
130 down_write(&ids->rwsem);
132 in_use = ids->in_use;
134 for (total = 0, next_id = 0; total < in_use; next_id++) {
135 perm = idr_find(&ids->ipcs_idr, next_id);
139 ipc_lock_object(perm);
143 up_write(&ids->rwsem);
146 static void free_ipc_ns(struct ipc_namespace *ns)
149 * Caller needs to wait for an RCU grace period to have passed
150 * after making the mount point inaccessible to new accesses.
157 retire_mq_sysctls(ns);
158 retire_ipc_sysctls(ns);
160 dec_ipc_namespaces(ns->ucounts);
161 put_user_ns(ns->user_ns);
162 ns_free_inum(&ns->ns);
166 static LLIST_HEAD(free_ipc_list);
167 static void free_ipc(struct work_struct *unused)
169 struct llist_node *node = llist_del_all(&free_ipc_list);
170 struct ipc_namespace *n, *t;
172 llist_for_each_entry_safe(n, t, node, mnt_llist)
173 mnt_make_shortterm(n->mq_mnt);
175 /* Wait for any last users to have gone away. */
178 llist_for_each_entry_safe(n, t, node, mnt_llist)
183 * put_ipc_ns - drop a reference to an ipc namespace.
184 * @ns: the namespace to put
186 * If this is the last task in the namespace exiting, and
187 * it is dropping the refcount to 0, then it can race with
188 * a task in another ipc namespace but in a mounts namespace
189 * which has this ipcns's mqueuefs mounted, doing some action
190 * with one of the mqueuefs files. That can raise the refcount.
191 * So dropping the refcount, and raising the refcount when
192 * accessing it through the VFS, are protected with mq_lock.
194 * (Clearly, a task raising the refcount on its own ipc_ns
195 * needn't take mq_lock since it can't race with the last task
196 * in the ipcns exiting).
198 void put_ipc_ns(struct ipc_namespace *ns)
200 if (refcount_dec_and_lock(&ns->ns.count, &mq_lock)) {
202 spin_unlock(&mq_lock);
204 if (llist_add(&ns->mnt_llist, &free_ipc_list))
205 schedule_work(&free_ipc_work);
209 static inline struct ipc_namespace *to_ipc_ns(struct ns_common *ns)
211 return container_of(ns, struct ipc_namespace, ns);
214 static struct ns_common *ipcns_get(struct task_struct *task)
216 struct ipc_namespace *ns = NULL;
217 struct nsproxy *nsproxy;
220 nsproxy = task->nsproxy;
222 ns = get_ipc_ns(nsproxy->ipc_ns);
225 return ns ? &ns->ns : NULL;
228 static void ipcns_put(struct ns_common *ns)
230 return put_ipc_ns(to_ipc_ns(ns));
233 static int ipcns_install(struct nsset *nsset, struct ns_common *new)
235 struct nsproxy *nsproxy = nsset->nsproxy;
236 struct ipc_namespace *ns = to_ipc_ns(new);
237 if (!ns_capable(ns->user_ns, CAP_SYS_ADMIN) ||
238 !ns_capable(nsset->cred->user_ns, CAP_SYS_ADMIN))
241 put_ipc_ns(nsproxy->ipc_ns);
242 nsproxy->ipc_ns = get_ipc_ns(ns);
246 static struct user_namespace *ipcns_owner(struct ns_common *ns)
248 return to_ipc_ns(ns)->user_ns;
251 const struct proc_ns_operations ipcns_operations = {
253 .type = CLONE_NEWIPC,
256 .install = ipcns_install,
257 .owner = ipcns_owner,