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
218
219
|
package app
import (
"bytes"
"context"
"encoding/gob"
"errors"
"fmt"
"io"
"maps"
"os"
"os/user"
"slices"
"strings"
"sync/atomic"
"syscall"
"hakurei.app/container"
"hakurei.app/hst"
"hakurei.app/internal/app/state"
"hakurei.app/system"
"hakurei.app/system/acl"
)
func newWithMessage(msg string) error { return newWithMessageError(msg, os.ErrInvalid) }
func newWithMessageError(msg string, err error) error {
return &hst.AppError{Step: "finalise", Err: err, Msg: msg}
}
// An outcome is the runnable state of a hakurei container via [hst.Config].
type outcome struct {
// initial [hst.Config] gob stream for state data;
// this is prepared ahead of time as config is clobbered during seal creation
ct io.WriterTo
sys *system.I
ctx context.Context
container container.Params
// Populated during outcome.finalise.
proc *finaliseProcess
// Whether the current process is in outcome.main.
active atomic.Bool
syscallDispatcher
}
func (k *outcome) finalise(ctx context.Context, msg container.Msg, id *state.ID, config *hst.Config) error {
const (
// only used for a nil configured env map
envAllocSize = 1 << 6
)
var kp finaliseProcess
if ctx == nil || id == nil {
// unreachable
panic("invalid call to finalise")
}
if k.ctx != nil || k.proc != nil {
// unreachable
panic("attempting to finalise twice")
}
k.ctx = ctx
if err := config.Validate(); err != nil {
return err
}
// TODO(ophestra): do not clobber during finalise
{
// encode initial configuration for state tracking
ct := new(bytes.Buffer)
if err := gob.NewEncoder(ct).Encode(config); err != nil {
return &hst.AppError{Step: "encode initial config", Err: err}
}
k.ct = ct
}
// this is checked again in hsu
if config.Identity < hst.IdentityMin || config.Identity > hst.IdentityMax {
return newWithMessage(fmt.Sprintf("identity %d out of range", config.Identity))
}
kp.supp = make([]string, len(config.Groups))
for i, name := range config.Groups {
if gid, err := k.lookupGroupId(name); err != nil {
var unknownGroupError user.UnknownGroupError
if errors.As(err, &unknownGroupError) {
return newWithMessageError(fmt.Sprintf("unknown group %q", name), unknownGroupError)
} else {
return &hst.AppError{Step: "look up group by name", Err: err}
}
} else {
kp.supp[i] = gid
}
}
// validation complete at this point
s := outcomeState{
ID: id,
Identity: config.Identity,
UserID: (&Hsu{k: k}).MustIDMsg(msg),
EnvPaths: copyPaths(k.syscallDispatcher),
Container: config.Container,
}
// enforce bounds and default early
if s.Container.WaitDelay <= 0 {
kp.waitDelay = hst.WaitDelayDefault
} else if s.Container.WaitDelay > hst.WaitDelayMax {
kp.waitDelay = hst.WaitDelayMax
} else {
kp.waitDelay = s.Container.WaitDelay
}
if s.Container.MapRealUID {
s.Mapuid, s.Mapgid = k.getuid(), k.getgid()
} else {
s.Mapuid, s.Mapgid = k.overflowUid(msg), k.overflowGid(msg)
}
// TODO(ophestra): duplicate in shim (params to shim)
if err := s.populateLocal(k.syscallDispatcher, msg); err != nil {
return err
}
kp.runDirPath, kp.identity, kp.id = s.sc.RunDirPath, s.identity, s.id
k.sys = system.New(k.ctx, msg, s.uid.unwrap())
{
ops := []outcomeOp{
// must run first
&spParamsOp{},
// TODO(ophestra): move this late for #8 and #9
spFilesystemOp{},
spRuntimeOp{},
spTmpdirOp{},
spAccountOp{},
}
et := config.Enablements.Unwrap()
if et&hst.EWayland != 0 {
ops = append(ops, &spWaylandOp{})
}
if et&hst.EX11 != 0 {
ops = append(ops, &spX11Op{})
}
if et&hst.EPulse != 0 {
ops = append(ops, &spPulseOp{})
}
if et&hst.EDBus != 0 {
ops = append(ops, &spDBusOp{})
}
stateSys := outcomeStateSys{sys: k.sys, outcomeState: &s}
for _, op := range ops {
if err := op.toSystem(&stateSys, config); err != nil {
return err
}
}
// TODO(ophestra): move to shim
stateParams := outcomeStateParams{params: &k.container, outcomeState: &s}
if s.Container.Env == nil {
stateParams.env = make(map[string]string, envAllocSize)
} else {
stateParams.env = maps.Clone(s.Container.Env)
}
for _, op := range ops {
if err := op.toContainer(&stateParams); err != nil {
return err
}
}
// flatten and sort env for deterministic behaviour
k.container.Env = make([]string, 0, len(stateParams.env))
for key, value := range stateParams.env {
if strings.IndexByte(key, '=') != -1 {
return &hst.AppError{Step: "flatten environment", Err: syscall.EINVAL,
Msg: fmt.Sprintf("invalid environment variable %s", key)}
}
k.container.Env = append(k.container.Env, key+"="+value)
}
slices.Sort(k.container.Env)
}
// mount root read-only as the final setup Op
// TODO(ophestra): move this to spFilesystemOp after #8 and #9
k.container.Remount(container.AbsFHSRoot, syscall.MS_RDONLY)
// append ExtraPerms last
for _, p := range config.ExtraPerms {
if p == nil || p.Path == nil {
continue
}
if p.Ensure {
k.sys.Ensure(p.Path, 0700)
}
perms := make(acl.Perms, 0, 3)
if p.Read {
perms = append(perms, acl.Read)
}
if p.Write {
perms = append(perms, acl.Write)
}
if p.Execute {
perms = append(perms, acl.Execute)
}
k.sys.UpdatePermType(system.User, p.Path, perms...)
}
k.proc = &kp
return nil
}
|
