1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
mod cgroup;
mod ipc;
mod mnt;
mod net;
mod user;
mod uts;
use alloc::sync::Arc;
pub use ax_cgroup::{CgroupNamespace, CgroupNode};
pub use cgroup::{ROOT_CGROUP_NS, new_cgroup_namespace};
pub use ipc::{IpcNamespace, ROOT_IPC_NS};
pub use mnt::{MntNamespace, ROOT_MNT_NS};
pub use net::{NetNamespace, ROOT_NET_NS};
pub use user::{ROOT_USER_NS, UserNamespace};
pub use uts::{ROOT_UTS_NS, UtNamespace, build_utsname};
use crate::sync::IrqMutex;
pub(crate) use crate::task::{PidNamespace, PidNamespaceRef, ROOT_PID_NS};
/// Aggregates all namespace types for a process.
///
/// `ProcessData` holds a single `IrqMutex<NsProxy>` field. Clone and unshare
/// operations work through `NsProxy` methods so that syscall handlers do not
/// manipulate namespace internals directly.
pub(crate) struct NsProxy {
/// The UTS namespace (hostname, domainname).
pub(crate) uts_ns: Arc<IrqMutex<UtNamespace>>,
/// The IPC namespace (System V IPC objects).
pub(crate) ipc_ns: Arc<IrqMutex<IpcNamespace>>,
/// The mount namespace (filesystem mount points).
pub(crate) mnt_ns: Arc<IrqMutex<MntNamespace>>,
/// PID namespace used by future children.
///
/// Linux keeps this choice until another `unshare(CLONE_NEWPID)` or
/// `setns(CLONE_NEWPID)` replaces it. Forking a child never consumes it.
pub(crate) pid_ns_for_children: PidNamespaceRef,
/// The network namespace (interfaces, routing, sockets).
pub(crate) net_ns: Arc<IrqMutex<NetNamespace>>,
/// The user namespace (UID/GID mappings).
pub(crate) user_ns: Arc<IrqMutex<UserNamespace>>,
/// The cgroup namespace (cgroup hierarchy view).
pub(crate) cgroup_ns: Arc<IrqMutex<CgroupNamespace>>,
}
impl NsProxy {
/// Create a new [`NsProxy`] pointing to the root namespaces.
pub fn new_root() -> Self {
Self {
uts_ns: ROOT_UTS_NS.clone(),
ipc_ns: ROOT_IPC_NS.clone(),
mnt_ns: ROOT_MNT_NS.clone(),
pid_ns_for_children: ROOT_PID_NS.clone(),
net_ns: ROOT_NET_NS.clone(),
user_ns: ROOT_USER_NS.clone(),
cgroup_ns: ROOT_CGROUP_NS.clone(),
}
}
/// Clone all namespace references (shallow `Arc` clone).
///
/// Used by `fork` / `clone` (without `CLONE_NEW*` flags) so the child
/// shares the same namespaces as the parent. The future-child PID
/// namespace is persistent and therefore inherited unchanged.
pub fn clone_all(&self) -> Self {
Self {
uts_ns: self.uts_ns.clone(),
ipc_ns: self.ipc_ns.clone(),
mnt_ns: self.mnt_ns.clone(),
pid_ns_for_children: self.pid_ns_for_children.clone(),
net_ns: self.net_ns.clone(),
user_ns: self.user_ns.clone(),
cgroup_ns: self.cgroup_ns.clone(),
}
}
/// Clone namespace references for a transactional `unshare(2)` update.
///
/// Unlike [`Self::clone_all`], this preserves a PID namespace already
/// staged for the next child. Preparing an unrelated unshare operation
/// must not discard pending `unshare(CLONE_NEWPID)` or `setns` state.
pub fn clone_for_unshare(&self) -> Self {
Self {
uts_ns: self.uts_ns.clone(),
ipc_ns: self.ipc_ns.clone(),
mnt_ns: self.mnt_ns.clone(),
pid_ns_for_children: self.pid_ns_for_children.clone(),
net_ns: self.net_ns.clone(),
user_ns: self.user_ns.clone(),
cgroup_ns: self.cgroup_ns.clone(),
}
}
pub fn unshare_uts(&mut self) {
let new_inner = self.uts_ns.lock().clone_ns();
self.uts_ns = Arc::new(IrqMutex::new(new_inner));
}
pub fn unshare_ipc(&mut self) {
let new_inner = self.ipc_ns.lock().clone_ns();
self.ipc_ns = Arc::new(IrqMutex::new(new_inner));
}
pub fn unshare_mnt(&mut self) {
let new_inner = self.mnt_ns.lock().clone_ns();
self.mnt_ns = Arc::new(IrqMutex::new(new_inner));
}
/// Prepare a new PID namespace for the next child of this process.
///
/// Called by `unshare(CLONE_NEWPID)`. The calling process stays in
/// its current PID namespace; the new namespace is consumed by the
/// next `fork` / `clone` child, which becomes PID 1 in that namespace.
pub fn prepare_pid_ns_for_children(&mut self, active: PidNamespaceRef) {
self.pid_ns_for_children = PidNamespace::new_child(active);
}
pub fn unshare_net(&mut self) {
let new_inner = self.net_ns.lock().clone_ns();
self.net_ns = Arc::new(IrqMutex::new(new_inner));
}
pub fn unshare_user(&mut self) {
let new_inner = self.user_ns.lock().clone_ns();
self.user_ns = Arc::new(IrqMutex::new(new_inner));
}
pub fn unshare_cgroup(&mut self, root: Arc<CgroupNode>) {
self.cgroup_ns = new_cgroup_namespace(root);
}
/// Replace the UTS namespace with an existing one (used by `setns(2)`).
pub fn set_ns_uts(&mut self, ns: Arc<IrqMutex<UtNamespace>>) {
self.uts_ns = ns;
}
/// Replace the IPC namespace with an existing one (used by `setns(2)`).
pub fn set_ns_ipc(&mut self, ns: Arc<IrqMutex<IpcNamespace>>) {
self.ipc_ns = ns;
}
/// Replace the mount namespace with an existing one (used by `setns(2)`).
pub fn set_ns_mnt(&mut self, ns: Arc<IrqMutex<MntNamespace>>) {
self.mnt_ns = ns;
}
/// Stage a PID namespace for the next child (used by `setns(2)`).
///
/// Linux `setns(CLONE_NEWPID)` never moves the calling process into the
/// target PID namespace. Instead the next `fork` / `clone` (without
/// `CLONE_NEWPID`) child enters it and becomes PID 1 there. This mirrors
/// `unshare(CLONE_NEWPID)` — both paths replace the persistent
/// `pid_ns_for_children` selection. The caller must be single-threaded.
pub fn set_ns_pid(&mut self, ns: PidNamespaceRef) {
self.pid_ns_for_children = ns;
}
/// Replace the network namespace with an existing one (used by `setns(2)`).
pub fn set_ns_net(&mut self, ns: Arc<IrqMutex<NetNamespace>>) {
self.net_ns = ns;
}
/// Replace the user namespace with an existing one (used by `setns(2)`).
pub fn set_ns_user(&mut self, ns: Arc<IrqMutex<UserNamespace>>) {
self.user_ns = ns;
}
/// Replace the cgroup namespace with an existing one (used by `setns(2)`).
pub fn set_ns_cgroup(&mut self, ns: Arc<IrqMutex<CgroupNamespace>>) {
self.cgroup_ns = ns;
}
/// Release the process-owned cgroup namespace after its final thread exits.
///
/// Exited scheduler tasks may retain `ProcessData` after userspace has
/// reaped the process, so cgroup root ownership cannot rely on `NsProxy`
/// destruction being synchronous with process exit.
pub fn release_cgroup_namespace(&mut self) {
self.cgroup_ns = ROOT_CGROUP_NS.clone();
}
}
#[cfg(test)]
mod tests {
extern crate std;
use super::*;
fn init_cgroup() {
static INIT: std::sync::Once = std::sync::Once::new();
INIT.call_once(ax_cgroup::init);
}
#[test]
fn clone_for_unshare_preserves_cgroup_namespace() {
init_cgroup();
let nsproxy = NsProxy::new_root();
let cloned = nsproxy.clone_for_unshare();
assert!(Arc::ptr_eq(&nsproxy.cgroup_ns, &cloned.cgroup_ns));
}
#[test]
fn final_process_exit_releases_cgroup_namespace_root() {
init_cgroup();
let mut nsproxy = NsProxy::new_root();
nsproxy.unshare_cgroup(ax_cgroup::root());
let exiting_namespace = nsproxy.cgroup_ns.clone();
assert!(!Arc::ptr_eq(&exiting_namespace, &ROOT_CGROUP_NS));
assert_eq!(Arc::strong_count(&exiting_namespace), 2);
nsproxy.release_cgroup_namespace();
assert!(Arc::ptr_eq(&nsproxy.cgroup_ns, &ROOT_CGROUP_NS));
assert_eq!(Arc::strong_count(&exiting_namespace), 1);
}
}