summaryrefslogtreecommitdiff
path: root/pkg/netns/netns_linux.go
blob: ecefb65ffd16a927b16e2a207a2e5b0ff9713031 (plain)
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
// Copyright 2018 CNI authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

// This file was originally a part of the containernetworking/plugins
// repository.
// It was copied here and modified for local use by the libpod maintainers.

package netns

import (
	"crypto/rand"
	"fmt"
	"os"
	"path"
	"path/filepath"
	"runtime"
	"strings"
	"sync"

	"github.com/containernetworking/plugins/pkg/ns"
	"github.com/containers/podman/v3/pkg/rootless"
	"github.com/containers/podman/v3/pkg/util"
	"github.com/sirupsen/logrus"
	"golang.org/x/sys/unix"
)

// GetNSRunDir returns the dir of where to create the netNS. When running
// rootless, it needs to be at a location writable by user.
func GetNSRunDir() (string, error) {
	if rootless.IsRootless() {
		rootlessDir, err := util.GetRuntimeDir()
		if err != nil {
			return "", err
		}
		return filepath.Join(rootlessDir, "netns"), nil
	}
	return "/run/netns", nil
}

// NewNS creates a new persistent (bind-mounted) network namespace and returns
// an object representing that namespace, without switching to it.
func NewNS() (ns.NetNS, error) {
	b := make([]byte, 16)
	_, err := rand.Reader.Read(b)
	if err != nil {
		return nil, fmt.Errorf("failed to generate random netns name: %v", err)
	}
	nsName := fmt.Sprintf("cni-%x-%x-%x-%x-%x", b[0:4], b[4:6], b[6:8], b[8:10], b[10:])
	return NewNSWithName(nsName)
}

// NewNSWithName creates a new persistent (bind-mounted) network namespace and returns
// an object representing that namespace, without switching to it.
func NewNSWithName(name string) (ns.NetNS, error) {
	nsRunDir, err := GetNSRunDir()
	if err != nil {
		return nil, err
	}

	// Create the directory for mounting network namespaces
	// This needs to be a shared mountpoint in case it is mounted in to
	// other namespaces (containers)
	err = os.MkdirAll(nsRunDir, 0755)
	if err != nil {
		return nil, err
	}

	// Remount the namespace directory shared. This will fail if it is not
	// already a mountpoint, so bind-mount it on to itself to "upgrade" it
	// to a mountpoint.
	err = unix.Mount("", nsRunDir, "none", unix.MS_SHARED|unix.MS_REC, "")
	if err != nil {
		if err != unix.EINVAL {
			return nil, fmt.Errorf("mount --make-rshared %s failed: %q", nsRunDir, err)
		}

		// Recursively remount /run/netns on itself. The recursive flag is
		// so that any existing netns bindmounts are carried over.
		err = unix.Mount(nsRunDir, nsRunDir, "none", unix.MS_BIND|unix.MS_REC, "")
		if err != nil {
			return nil, fmt.Errorf("mount --rbind %s %s failed: %q", nsRunDir, nsRunDir, err)
		}

		// Now we can make it shared
		err = unix.Mount("", nsRunDir, "none", unix.MS_SHARED|unix.MS_REC, "")
		if err != nil {
			return nil, fmt.Errorf("mount --make-rshared %s failed: %q", nsRunDir, err)
		}
	}

	// create an empty file at the mount point
	nsPath := path.Join(nsRunDir, name)
	mountPointFd, err := os.Create(nsPath)
	if err != nil {
		return nil, err
	}
	if err := mountPointFd.Close(); err != nil {
		return nil, err
	}

	// Ensure the mount point is cleaned up on errors; if the namespace
	// was successfully mounted this will have no effect because the file
	// is in-use
	defer func() {
		_ = os.RemoveAll(nsPath)
	}()

	var wg sync.WaitGroup
	wg.Add(1)

	// do namespace work in a dedicated goroutine, so that we can safely
	// Lock/Unlock OSThread without upsetting the lock/unlock state of
	// the caller of this function
	go (func() {
		defer wg.Done()
		runtime.LockOSThread()
		// Don't unlock. By not unlocking, golang will kill the OS thread when the
		// goroutine is done (for go1.10+)

		threadNsPath := getCurrentThreadNetNSPath()

		var origNS ns.NetNS
		origNS, err = ns.GetNS(threadNsPath)
		if err != nil {
			logrus.Warnf("cannot open current network namespace %s: %q", threadNsPath, err)
			return
		}
		defer func() {
			if err := origNS.Close(); err != nil {
				logrus.Errorf("unable to close namespace: %q", err)
			}
		}()

		// create a new netns on the current thread
		err = unix.Unshare(unix.CLONE_NEWNET)
		if err != nil {
			logrus.Warnf("cannot create a new network namespace: %q", err)
			return
		}

		// Put this thread back to the orig ns, since it might get reused (pre go1.10)
		defer func() {
			if err := origNS.Set(); err != nil {
				if rootless.IsRootless() && strings.Contains(err.Error(), "operation not permitted") {
					// When running in rootless mode it will fail to re-join
					// the network namespace owned by root on the host.
					return
				}
				logrus.Warnf("unable to reset namespace: %q", err)
			}
		}()

		// bind mount the netns from the current thread (from /proc) onto the
		// mount point. This causes the namespace to persist, even when there
		// are no threads in the ns. Make this a shared mount; it needs to be
		// back-propagated to the host
		err = unix.Mount(threadNsPath, nsPath, "none", unix.MS_BIND|unix.MS_SHARED|unix.MS_REC, "")
		if err != nil {
			err = fmt.Errorf("failed to bind mount ns at %s: %v", nsPath, err)
		}
	})()
	wg.Wait()

	if err != nil {
		return nil, fmt.Errorf("failed to create namespace: %v", err)
	}

	return ns.GetNS(nsPath)
}

// UnmountNS unmounts the NS held by the netns object
func UnmountNS(ns ns.NetNS) error {
	nsRunDir, err := GetNSRunDir()
	if err != nil {
		return err
	}

	nsPath := ns.Path()
	// Only unmount if it's been bind-mounted (don't touch namespaces in /proc...)
	if strings.HasPrefix(nsPath, nsRunDir) {
		if err := unix.Unmount(nsPath, unix.MNT_DETACH); err != nil {
			return fmt.Errorf("failed to unmount NS: at %s: %v", nsPath, err)
		}

		if err := os.Remove(nsPath); err != nil {
			return fmt.Errorf("failed to remove ns path %s: %v", nsPath, err)
		}
	}

	return nil
}

// getCurrentThreadNetNSPath copied from pkg/ns
func getCurrentThreadNetNSPath() string {
	// /proc/self/ns/net returns the namespace of the main thread, not
	// of whatever thread this goroutine is running on.  Make sure we
	// use the thread's net namespace since the thread is switching around
	return fmt.Sprintf("/proc/%d/task/%d/ns/net", os.Getpid(), unix.Gettid())
}