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
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
|
use anyhow::{Context, Result, ensure};
use nftnl::{Batch, Chain, ChainType, Hook, MsgType, Policy, ProtoFamily, Rule, Table, nft_expr};
use std::{
collections::HashSet,
ffi::CString,
io,
net::{Ipv4Addr, Ipv6Addr},
os::fd::AsRawFd,
};
pub const CAPTURE_MARK: u32 = 0x0200_0000;
const INCOMING_MARK: u32 = 0x0100_0000;
const TABLE: &std::ffi::CStr = c"roentgen";
fn socket() -> io::Result<mnl::Socket> {
let socket = mnl::Socket::new(mnl::Bus::Netfilter)?;
let timeout = libc::timeval {
tv_sec: 5,
tv_usec: 0,
};
// SAFETY: the live socket fd and timeval pointer/length match SO_RCVTIMEO.
let result = unsafe {
libc::setsockopt(
socket.as_raw_fd(),
libc::SOL_SOCKET,
libc::SO_RCVTIMEO,
(&timeout as *const libc::timeval).cast(),
std::mem::size_of_val(&timeout) as _,
)
};
if result != 0 {
return Err(io::Error::last_os_error());
}
Ok(socket)
}
fn tables() -> Result<HashSet<CString>> {
let socket = socket()?;
socket.send(&nftnl::table::get_tables_nlmsg(1))?;
let mut tables = HashSet::new();
let mut buffer = vec![0u8; nftnl::nft_nlmsg_maxsize() as usize];
loop {
for message in socket.recv(&mut buffer)? {
if matches!(
mnl::cb_run2(
message?,
1,
socket.portid(),
nftnl::table::get_tables_cb,
&mut tables
)?,
mnl::CbResult::Stop
) {
return Ok(tables);
}
}
}
}
fn send(batch: nftnl::FinalizedBatch) -> Result<()> {
let socket = socket()?;
socket.send_all(&batch)?;
let mut buffer = vec![0u8; nftnl::nft_nlmsg_maxsize() as usize];
let mut sequences = batch.sequence_numbers();
while !sequences.is_empty() {
for message in socket.recv(&mut buffer)? {
let sequence = sequences
.next()
.context("Unexpected firewall acknowledgement")?;
mnl::cb_run(message?, sequence, socket.portid())?;
}
}
Ok(())
}
pub fn preflight() -> Result<()> {
let tables = tables()?;
ensure!(
tables.iter().all(|name| name.as_c_str() == TABLE),
"Other nftables tables exist; this firewall arrangement is not supported. Roentgen did not alter them"
);
Ok(())
}
/// Replacing the table is one kernel transaction. Connection marks survive table replacement.
pub fn apply(enabled: bool, xray_uid: u32) -> Result<()> {
apply_policy(enabled, xray_uid, false)
}
pub fn forward(xray_uid: u32) -> Result<()> {
apply_policy(true, xray_uid, true)
}
fn apply_policy(enabled: bool, xray_uid: u32, allow_tun: bool) -> Result<()> {
let existing = tables()?;
let table = Table::new(TABLE, ProtoFamily::Inet);
let mut batch = Batch::new();
if existing.contains(TABLE) {
batch.add(&table, MsgType::Del);
}
batch.add(&table, MsgType::Add);
let mut incoming = Chain::new(c"incoming", &table);
incoming.set_type(ChainType::Filter);
incoming.set_hook(Hook::In, -150);
incoming.set_policy(Policy::Accept);
batch.add(&incoming, MsgType::Add);
let mut classify = Rule::new(&incoming);
classify.add_expr(&nft_expr!(meta iifname));
classify.add_expr(&nft_expr!(cmp != &b"lo\0"[..]));
classify.add_expr(&nft_expr!(ct state));
classify.add_expr(&nft_expr!(bitwise mask 8u32, xor 0u32));
classify.add_expr(&nft_expr!(cmp == 8u32));
classify.add_expr(&nft_expr!(ct mark));
classify.add_expr(&nft_expr!(bitwise mask !INCOMING_MARK, xor INCOMING_MARK));
classify.add_expr(&nft_expr!(ct mark set));
batch.add(&classify, MsgType::Add);
if enabled {
let mut routing = Chain::new(c"route", &table);
routing.set_type(ChainType::Route);
routing.set_hook(Hook::Out, -150);
routing.set_policy(Policy::Accept);
batch.add(&routing, MsgType::Add);
exceptions(&mut batch, &routing, xray_uid);
let mut mark = Rule::new(&routing);
mark.add_expr(&nft_expr!(meta mark));
mark.add_expr(&nft_expr!(bitwise mask !CAPTURE_MARK, xor CAPTURE_MARK));
mark.add_expr(&nft_expr!(meta mark set));
batch.add(&mark, MsgType::Add);
let mut guard = Chain::new(c"guard", &table);
guard.set_type(ChainType::Filter);
// OUTPUT's hook state still names the original interface after the
// route chain changes skb's route. POSTROUTING sees the final egress,
// so the guard can allow TUN traffic without allowing direct fallback.
guard.set_hook(Hook::PostRouting, 100);
guard.set_policy(Policy::Drop);
batch.add(&guard, MsgType::Add);
exceptions(&mut batch, &guard, xray_uid);
if allow_tun {
let mut tun = Rule::new(&guard);
tun.add_expr(&nft_expr!(meta oifname));
tun.add_expr(&nft_expr!(cmp == &b"roentgen0\0"[..]));
tun.add_expr(&nft_expr!(verdict accept));
batch.add(&tun, MsgType::Add);
}
}
send(batch.finalize()).context("Could not apply Roentgen firewall transaction")
}
fn exceptions(batch: &mut Batch, chain: &Chain<'_>, xray_uid: u32) {
let mut loopback = Rule::new(chain);
loopback.add_expr(&nft_expr!(meta oifname));
loopback.add_expr(&nft_expr!(cmp == &b"lo\0"[..]));
loopback.add_expr(&nft_expr!(verdict accept));
batch.add(&loopback, MsgType::Add);
let mut core = Rule::new(chain);
core.add_expr(&nft_expr!(meta skuid));
core.add_expr(&nft_expr!(cmp == xray_uid));
core.add_expr(&nft_expr!(verdict accept));
batch.add(&core, MsgType::Add);
let mut replies = Rule::new(chain);
replies.add_expr(&nft_expr!(ct mark));
replies.add_expr(&nft_expr!(bitwise mask INCOMING_MARK, xor 0u32));
replies.add_expr(&nft_expr!(cmp == INCOMING_MARK));
replies.add_expr(&nft_expr!(verdict accept));
batch.add(&replies, MsgType::Add);
for (network, mask) in [
(Ipv4Addr::new(10, 0, 0, 0), Ipv4Addr::new(255, 0, 0, 0)),
(Ipv4Addr::new(172, 16, 0, 0), Ipv4Addr::new(255, 240, 0, 0)),
(Ipv4Addr::new(192, 168, 0, 0), Ipv4Addr::new(255, 255, 0, 0)),
] {
let mut rule = Rule::new(chain);
rule.add_expr(&nft_expr!(meta nfproto));
rule.add_expr(&nft_expr!(cmp == libc::NFPROTO_IPV4 as u8));
rule.add_expr(&nft_expr!(payload ipv4 daddr));
rule.add_expr(&nft_expr!(bitwise mask mask, xor 0u32));
rule.add_expr(&nft_expr!(cmp == network));
rule.add_expr(&nft_expr!(verdict accept));
batch.add(&rule, MsgType::Add);
}
for (network, mask) in [(
Ipv6Addr::new(0xfc00, 0, 0, 0, 0, 0, 0, 0),
Ipv6Addr::new(0xfe00, 0, 0, 0, 0, 0, 0, 0),
)] {
let mut rule = Rule::new(chain);
rule.add_expr(&nft_expr!(meta nfproto));
rule.add_expr(&nft_expr!(cmp == libc::NFPROTO_IPV6 as u8));
rule.add_expr(&nft_expr!(payload ipv6 daddr));
rule.add_expr(&nft_expr!(bitwise mask mask, xor Ipv6Addr::UNSPECIFIED));
rule.add_expr(&nft_expr!(cmp == network));
rule.add_expr(&nft_expr!(verdict accept));
batch.add(&rule, MsgType::Add);
}
// DHCPv4 broadcast is link control, not a general broadcast exception.
let mut dhcp = Rule::new(chain);
dhcp.add_expr(&nft_expr!(meta nfproto));
dhcp.add_expr(&nft_expr!(cmp == libc::NFPROTO_IPV4 as u8));
dhcp.add_expr(&nft_expr!(payload ipv4 daddr));
dhcp.add_expr(&nft_expr!(cmp == Ipv4Addr::BROADCAST));
dhcp.add_expr(&nft_expr!(meta l4proto));
dhcp.add_expr(&nft_expr!(cmp == libc::IPPROTO_UDP as u8));
dhcp.add_expr(&nft_expr!(payload udp sport));
dhcp.add_expr(&nft_expr!(cmp == 68u16.to_be()));
dhcp.add_expr(&nft_expr!(payload udp dport));
dhcp.add_expr(&nft_expr!(cmp == 67u16.to_be()));
dhcp.add_expr(&nft_expr!(verdict accept));
batch.add(&dhcp, MsgType::Add);
let mut dhcp6 = Rule::new(chain);
dhcp6.add_expr(&nft_expr!(meta nfproto));
dhcp6.add_expr(&nft_expr!(cmp == libc::NFPROTO_IPV6 as u8));
dhcp6.add_expr(&nft_expr!(payload ipv6 daddr));
dhcp6.add_expr(&nft_expr!(
cmp == Ipv6Addr::new(0xff02, 0, 0, 0, 0, 0, 1, 2)
));
dhcp6.add_expr(&nft_expr!(meta l4proto));
dhcp6.add_expr(&nft_expr!(cmp == libc::IPPROTO_UDP as u8));
dhcp6.add_expr(&nft_expr!(payload udp sport));
dhcp6.add_expr(&nft_expr!(cmp == 546u16.to_be()));
dhcp6.add_expr(&nft_expr!(payload udp dport));
dhcp6.add_expr(&nft_expr!(cmp == 547u16.to_be()));
dhcp6.add_expr(&nft_expr!(verdict accept));
batch.add(&dhcp6, MsgType::Add);
// Limit multicast ICMPv6 to router/neighbor discovery with hop limit 255.
for kind in [133u8, 134, 135, 136] {
let mut rule = Rule::new(chain);
rule.add_expr(&nft_expr!(meta nfproto));
rule.add_expr(&nft_expr!(cmp == libc::NFPROTO_IPV6 as u8));
rule.add_expr(&nft_expr!(meta l4proto));
rule.add_expr(&nft_expr!(cmp == libc::IPPROTO_ICMPV6 as u8));
rule.add_expr(&nft_expr!(payload ipv6 hoplimit));
rule.add_expr(&nft_expr!(cmp == 255u8));
rule.add_expr(&nftnl::expr::Payload::Transport(
nftnl::expr::TransportHeaderField::Icmpv6(nftnl::expr::Icmpv6HeaderField::Type),
));
rule.add_expr(&nft_expr!(cmp == kind));
rule.add_expr(&nft_expr!(verdict accept));
batch.add(&rule, MsgType::Add);
}
}
pub fn remove() -> Result<()> {
if tables()?.contains(TABLE) {
let mut batch = Batch::new();
batch.add(&Table::new(TABLE, ProtoFamily::Inet), MsgType::Del);
send(batch.finalize())?;
}
Ok(())
}
|
