summaryrefslogtreecommitdiffhomepage
path: root/runsc/cmd/gofer.go
blob: f28e02798cb42cc51f1d7e9ed72bfa1a21454060 (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
// Copyright 2018 Google Inc.
//
// 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.

package cmd

import (
	"os"
	"path"
	"sync"
	"syscall"

	"context"
	"flag"
	"github.com/google/subcommands"
	specs "github.com/opencontainers/runtime-spec/specs-go"
	"gvisor.googlesource.com/gvisor/pkg/log"
	"gvisor.googlesource.com/gvisor/pkg/p9"
	"gvisor.googlesource.com/gvisor/pkg/unet"
	"gvisor.googlesource.com/gvisor/runsc/fsgofer"
	"gvisor.googlesource.com/gvisor/runsc/fsgofer/filter"
	"gvisor.googlesource.com/gvisor/runsc/specutils"
)

// Gofer implements subcommands.Command for the "gofer" command, which starts a
// filesystem gofer.  This command should not be called directly.
type Gofer struct {
	bundleDir string
	ioFDs     intFlags
	applyCaps bool

	panicOnWrite bool
}

// Name implements subcommands.Command.
func (*Gofer) Name() string {
	return "gofer"
}

// Synopsis implements subcommands.Command.
func (*Gofer) Synopsis() string {
	return "launch a gofer process that serves files over 9P protocol (internal use only)"
}

// Usage implements subcommands.Command.
func (*Gofer) Usage() string {
	return `gofer [flags]`
}

// SetFlags implements subcommands.Command.
func (g *Gofer) SetFlags(f *flag.FlagSet) {
	f.StringVar(&g.bundleDir, "bundle", "", "path to the root of the bundle directory, defaults to the current directory")
	f.Var(&g.ioFDs, "io-fds", "list of FDs to connect 9P servers. They must follow this order: root first, then mounts as defined in the spec")
	f.BoolVar(&g.applyCaps, "apply-caps", true, "if true, apply capabilities to restrict what the Gofer process can do")
	f.BoolVar(&g.panicOnWrite, "panic-on-write", false, "if true, panics on attempts to write to RO mounts. RW mounts are unnaffected")
}

// Execute implements subcommands.Command.
func (g *Gofer) Execute(_ context.Context, f *flag.FlagSet, args ...interface{}) subcommands.ExitStatus {
	if g.bundleDir == "" || len(g.ioFDs) < 1 {
		f.Usage()
		return subcommands.ExitUsageError
	}

	if g.applyCaps {
		// Minimal set of capabilities needed by the Gofer to operate on files.
		caps := []string{
			"CAP_CHOWN",
			"CAP_DAC_OVERRIDE",
			"CAP_DAC_READ_SEARCH",
			"CAP_FOWNER",
			"CAP_FSETID",
		}
		lc := &specs.LinuxCapabilities{
			Bounding:  caps,
			Effective: caps,
			Permitted: caps,
		}

		// Disable caps when calling myself again.
		// Note: minimal argument handling for the default case to keep it simple.
		args := os.Args
		args = append(args, "--apply-caps=false")
		if err := setCapsAndCallSelf(args, lc); err != nil {
			Fatalf("Unable to apply caps: %v", err)
		}
		panic("unreachable")
	}

	spec, err := specutils.ReadSpec(g.bundleDir)
	if err != nil {
		Fatalf("error reading spec: %v", err)
	}
	specutils.LogSpec(spec)

	// fsgofer should run with a umask of 0, because we want to preserve file
	// modes exactly as sent by the sandbox, which will have applied its own umask.
	syscall.Umask(0)

	// Find what path is going to be served by this gofer.
	root := absPath(g.bundleDir, spec.Root.Path)
	if err := syscall.Chroot(root); err != nil {
		Fatalf("failed to chroot to %q: %v", root, err)
	}
	if err := syscall.Chdir("/"); err != nil {
		Fatalf("failed to change working dir: %v", err)
	}
	log.Infof("Process chroot'd to %q", root)

	// Start with root mount, then add any other additional mount as needed.
	ats := make([]p9.Attacher, 0, len(spec.Mounts)+1)
	ats = append(ats, fsgofer.NewAttachPoint("/", fsgofer.Config{
		ROMount:      spec.Root.Readonly,
		PanicOnWrite: g.panicOnWrite,
		// Docker uses overlay2 by default for the root mount, and overlay2 does a copy-up when
		// each file is opened as writable. Thus, we open files lazily to avoid copy-up.
		LazyOpenForWrite: true,
	}))
	log.Infof("Serving %q mapped to %q on FD %d (ro: %t)", "/", root, g.ioFDs[0], spec.Root.Readonly)

	mountIdx := 1 // first one is the root
	for _, m := range spec.Mounts {
		if specutils.Is9PMount(m) {
			if !path.IsAbs(m.Destination) {
				Fatalf("destination must be absolute path: %v", m.Destination)
			}
			cfg := fsgofer.Config{
				ROMount:          isReadonlyMount(m.Options),
				PanicOnWrite:     g.panicOnWrite,
				LazyOpenForWrite: false,
			}
			ats = append(ats, fsgofer.NewAttachPoint(m.Destination, cfg))

			if mountIdx >= len(g.ioFDs) {
				Fatalf("No FD found for mount. Did you forget --io-fd? mount: %d, %v", len(g.ioFDs), m)
			}
			log.Infof("Serving %q mapped on FD %d (ro: %t)", m.Destination, g.ioFDs[mountIdx], cfg.ROMount)
			mountIdx++
		}
	}
	if mountIdx != len(g.ioFDs) {
		Fatalf("too many FDs passed for mounts. mounts: %d, FDs: %d", mountIdx, len(g.ioFDs))
	}

	if err := filter.Install(); err != nil {
		Fatalf("Failed to install seccomp filters: %v", err)
	}

	runServers(ats, g.ioFDs)
	return subcommands.ExitSuccess
}

func runServers(ats []p9.Attacher, ioFDs []int) {
	// Run the loops and wait for all to exit.
	var wg sync.WaitGroup
	for i, ioFD := range ioFDs {
		wg.Add(1)
		go func(ioFD int, at p9.Attacher) {
			socket, err := unet.NewSocket(ioFD)
			if err != nil {
				Fatalf("err creating server on FD %d: %v", ioFD, err)
			}
			s := p9.NewServer(at)
			if err := s.Handle(socket); err != nil {
				Fatalf("P9 server returned error. Gofer is shutting down. FD: %d, err: %v", ioFD, err)
			}
			wg.Done()
		}(ioFD, ats[i])
	}
	wg.Wait()
	log.Infof("All 9P servers exited.")
}

func isReadonlyMount(opts []string) bool {
	for _, o := range opts {
		if o == "ro" {
			return true
		}
	}
	return false
}