blob: 5027ec377a813a228b483a458e02466f4bf89703 [file] [log] [blame]
Daniel Drown0da73fc2012-06-20 16:51:39 -05001/*
2 * Copyright (C) 2008 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
Bernie Innocenti51a0e0f2018-10-05 20:24:06 +090016
Lorenzo Colittiac7fefc2014-10-20 17:14:13 +090017#include <map>
18#include <string>
19
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +090020#include <arpa/inet.h>
Daniel Drown0da73fc2012-06-20 16:51:39 -050021#include <errno.h>
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -070022#include <linux/if_arp.h>
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -070023#include <linux/if_tun.h>
24#include <linux/ioctl.h>
Maciej Żenczykowskic8c38aa2019-03-29 01:24:51 -070025#include <net/if.h>
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +090026#include <netinet/in.h>
Luke Huang94c43a12019-02-14 19:51:38 +080027#include <spawn.h>
Daniel Drown0da73fc2012-06-20 16:51:39 -050028#include <sys/types.h>
29#include <sys/wait.h>
Luke Huang94c43a12019-02-14 19:51:38 +080030#include <unistd.h>
Daniel Drown0da73fc2012-06-20 16:51:39 -050031
32#define LOG_TAG "ClatdController"
Logan Chien3f461482018-04-23 14:31:32 +080033#include <log/log.h>
Daniel Drown0da73fc2012-06-20 16:51:39 -050034
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -070035#include "ClatdController.h"
Maciej Żenczykowskia1699952019-05-11 17:07:44 -070036#include "InterfaceController.h"
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -070037
38#include "android-base/properties.h"
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -070039#include "android-base/scopeguard.h"
Lorenzo Colitti91fd5802019-06-28 19:22:01 +090040#include "android-base/stringprintf.h"
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +090041#include "android-base/unique_fd.h"
Maciej Żenczykowski55262712019-03-29 23:44:56 -070042#include "bpf/BpfMap.h"
43#include "netdbpf/bpf_shared.h"
44#include "netdutils/DumpWriter.h"
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +090045
46extern "C" {
47#include "netutils/checksum.h"
48}
49
Maciej Żenczykowski55262712019-03-29 23:44:56 -070050#include "ClatUtils.h"
Lorenzo Colitti45d3dd02014-06-09 14:09:20 +090051#include "Fwmark.h"
Paul Jensen84c1d032014-05-30 13:29:41 -040052#include "NetdConstants.h"
53#include "NetworkController.h"
Bernie Innocenti189eb502018-10-01 23:10:18 +090054#include "netid_client.h"
Daniel Drown0da73fc2012-06-20 16:51:39 -050055
Lorenzo Colittiac7fefc2014-10-20 17:14:13 +090056static const char* kClatdPath = "/system/bin/clatd";
57
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +090058// For historical reasons, start with 192.0.0.4, and after that, use all subsequent addresses in
59// 192.0.0.0/29 (RFC 7335).
60static const char* kV4AddrString = "192.0.0.4";
61static const in_addr kV4Addr = {inet_addr(kV4AddrString)};
62static const int kV4AddrLen = 29;
63
Lorenzo Colitti91fd5802019-06-28 19:22:01 +090064using android::base::StringPrintf;
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +090065using android::base::unique_fd;
Maciej Żenczykowski55262712019-03-29 23:44:56 -070066using android::bpf::BpfMap;
67using android::netdutils::DumpWriter;
68using android::netdutils::ScopedIndent;
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +090069
Lorenzo Colitti7035f222017-02-13 18:29:00 +090070namespace android {
71namespace net {
72
Maciej Żenczykowski56280272019-03-30 03:32:51 -070073void ClatdController::init(void) {
74 std::lock_guard guard(mutex);
75
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -070076 // TODO: should refactor into separate function for testability
77 if (bpf::getBpfSupportLevel() == bpf::BpfLevel::NONE) {
78 ALOGI("Pre-4.9 kernel or pre-P api shipping level - disabling clat ebpf.");
79 mClatEbpfMode = ClatEbpfDisabled;
80 return;
81 }
82
83 // We know the device initially shipped with at least P...,
84 // but did it ship with at least Q?
85
86 uint64_t api_level = base::GetUintProperty<uint64_t>("ro.product.first_api_level", 0);
87 if (api_level == 0) {
88 ALOGE("Cannot determine initial API level of the device.");
89 api_level = base::GetUintProperty<uint64_t>("ro.build.version.sdk", 0);
90 }
91
92 // Note: MINIMUM_API_REQUIRED is for eBPF as a whole and is thus P
93 if (api_level > bpf::MINIMUM_API_REQUIRED) {
94 ALOGI("4.9+ kernel and device shipped with Q+ - clat ebpf should work.");
95 mClatEbpfMode = ClatEbpfEnabled;
96 } else {
97 // We cannot guarantee that 4.9-P kernels will include NET_CLS_BPF support.
98 ALOGI("4.9+ kernel and device shipped with P - clat ebpf might work.");
99 mClatEbpfMode = ClatEbpfMaybe;
100 }
101
102 int rv = openNetlinkSocket();
103 if (rv < 0) {
104 ALOGE("openNetlinkSocket() failure: %s", strerror(-rv));
105 mClatEbpfMode = ClatEbpfDisabled;
106 return;
107 }
108 mNetlinkFd.reset(rv);
109
Maciej Żenczykowski0bd8da72019-12-16 15:16:28 -0800110 rv = getClatEgressMapFd();
111 if (rv < 0) {
112 ALOGE("getClatEgressMapFd() failure: %s", strerror(-rv));
113 mClatEbpfMode = ClatEbpfDisabled;
114 mNetlinkFd.reset(-1);
115 return;
116 }
117 mClatEgressMap.reset(rv);
118
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700119 rv = getClatIngressMapFd();
120 if (rv < 0) {
121 ALOGE("getClatIngressMapFd() failure: %s", strerror(-rv));
122 mClatEbpfMode = ClatEbpfDisabled;
Maciej Żenczykowski0bd8da72019-12-16 15:16:28 -0800123 mClatEgressMap.reset(-1);
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700124 mNetlinkFd.reset(-1);
125 return;
126 }
127 mClatIngressMap.reset(rv);
128
129 int netlinkFd = mNetlinkFd.get();
130
131 // TODO: perhaps this initial cleanup should be in its own function?
132 const auto del = [&netlinkFd](const ClatIngressKey& key,
133 const BpfMap<ClatIngressKey, ClatIngressValue>&) {
134 ALOGW("Removing stale clat config on interface %d.", key.iif);
135 int rv = tcQdiscDelDevClsact(netlinkFd, key.iif);
136 if (rv < 0) ALOGE("tcQdiscDelDevClsact() failure: %s", strerror(-rv));
137 return netdutils::status::ok; // keep on going regardless
138 };
139 auto ret = mClatIngressMap.iterate(del);
140 if (!isOk(ret)) ALOGE("mClatIngressMap.iterate() failure: %s", strerror(ret.code()));
141 ret = mClatIngressMap.clear();
142 if (!isOk(ret)) ALOGE("mClatIngressMap.clear() failure: %s", strerror(ret.code()));
143}
144
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900145bool ClatdController::isIpv4AddressFree(in_addr_t addr) {
146 int s = socket(AF_INET, SOCK_DGRAM | SOCK_CLOEXEC, 0);
147 if (s == -1) {
148 return 0;
149 }
150
151 // Attempt to connect to the address. If the connection succeeds and getsockname returns the
152 // same then the address is already assigned to the system and we can't use it.
Nick Desaulniers6b357502019-10-11 09:26:44 -0700153 struct sockaddr_in sin = {
154 .sin_family = AF_INET,
Maciej Żenczykowski268190e2019-10-31 23:47:53 -0700155 .sin_port = htons(53),
Nick Desaulniers6b357502019-10-11 09:26:44 -0700156 .sin_addr = {addr},
157 };
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900158 socklen_t len = sizeof(sin);
159 bool inuse = connect(s, (struct sockaddr*)&sin, sizeof(sin)) == 0 &&
160 getsockname(s, (struct sockaddr*)&sin, &len) == 0 && (size_t)len >= sizeof(sin) &&
161 sin.sin_addr.s_addr == addr;
162
163 close(s);
164 return !inuse;
Lorenzo Colittiac7fefc2014-10-20 17:14:13 +0900165}
Daniel Drown0da73fc2012-06-20 16:51:39 -0500166
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900167// Picks a free IPv4 address, starting from ip and trying all addresses in the prefix in order.
168// ip - the IP address from the configuration file
169// prefixlen - the length of the prefix from which addresses may be selected.
170// returns: the IPv4 address, or INADDR_NONE if no addresses were available
171in_addr_t ClatdController::selectIpv4Address(const in_addr ip, int16_t prefixlen) {
172 // Don't accept prefixes that are too large because we scan addresses one by one.
173 if (prefixlen < 16 || prefixlen > 32) {
174 return INADDR_NONE;
175 }
Lorenzo Colittiac7fefc2014-10-20 17:14:13 +0900176
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900177 // All these are in host byte order.
178 in_addr_t mask = 0xffffffff >> (32 - prefixlen) << (32 - prefixlen);
179 in_addr_t ipv4 = ntohl(ip.s_addr);
180 in_addr_t first_ipv4 = ipv4;
181 in_addr_t prefix = ipv4 & mask;
182
183 // Pick the first IPv4 address in the pool, wrapping around if necessary.
184 // So, for example, 192.0.0.4 -> 192.0.0.5 -> 192.0.0.6 -> 192.0.0.7 -> 192.0.0.0.
185 do {
186 if (isIpv4AddressFreeFunc(htonl(ipv4))) {
187 return htonl(ipv4);
188 }
189 ipv4 = prefix | ((ipv4 + 1) & ~mask);
190 } while (ipv4 != first_ipv4);
191
192 return INADDR_NONE;
193}
194
195// Alters the bits in the IPv6 address to make them checksum neutral with v4 and nat64Prefix.
196void ClatdController::makeChecksumNeutral(in6_addr* v6, const in_addr v4,
197 const in6_addr& nat64Prefix) {
198 // Fill last 8 bytes of IPv6 address with random bits.
199 arc4random_buf(&v6->s6_addr[8], 8);
200
201 // Make the IID checksum-neutral. That is, make it so that:
202 // checksum(Local IPv4 | Remote IPv4) = checksum(Local IPv6 | Remote IPv6)
203 // in other words (because remote IPv6 = NAT64 prefix | Remote IPv4):
204 // checksum(Local IPv4) = checksum(Local IPv6 | NAT64 prefix)
205 // Do this by adjusting the two bytes in the middle of the IID.
206
207 uint16_t middlebytes = (v6->s6_addr[11] << 8) + v6->s6_addr[12];
208
209 uint32_t c1 = ip_checksum_add(0, &v4, sizeof(v4));
210 uint32_t c2 = ip_checksum_add(0, &nat64Prefix, sizeof(nat64Prefix)) +
211 ip_checksum_add(0, v6, sizeof(*v6));
212
213 uint16_t delta = ip_checksum_adjust(middlebytes, c1, c2);
214 v6->s6_addr[11] = delta >> 8;
215 v6->s6_addr[12] = delta & 0xff;
216}
217
218// Picks a random interface ID that is checksum neutral with the IPv4 address and the NAT64 prefix.
219int ClatdController::generateIpv6Address(const char* iface, const in_addr v4,
220 const in6_addr& nat64Prefix, in6_addr* v6) {
221 unique_fd s(socket(AF_INET6, SOCK_DGRAM | SOCK_CLOEXEC, 0));
222 if (s == -1) return -errno;
223
224 if (setsockopt(s, SOL_SOCKET, SO_BINDTODEVICE, iface, strlen(iface) + 1) == -1) {
Luke Huang6d301232018-08-01 14:05:18 +0800225 return -errno;
Daniel Drown0da73fc2012-06-20 16:51:39 -0500226 }
227
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900228 sockaddr_in6 sin6 = {.sin6_family = AF_INET6, .sin6_addr = nat64Prefix};
229 if (connect(s, reinterpret_cast<struct sockaddr*>(&sin6), sizeof(sin6)) == -1) {
230 return -errno;
231 }
232
233 socklen_t len = sizeof(sin6);
234 if (getsockname(s, reinterpret_cast<struct sockaddr*>(&sin6), &len) == -1) {
235 return -errno;
236 }
237
238 *v6 = sin6.sin6_addr;
239
240 if (IN6_IS_ADDR_UNSPECIFIED(v6) || IN6_IS_ADDR_LOOPBACK(v6) || IN6_IS_ADDR_LINKLOCAL(v6) ||
241 IN6_IS_ADDR_SITELOCAL(v6) || IN6_IS_ADDR_ULA(v6)) {
242 return -ENETUNREACH;
243 }
244
245 makeChecksumNeutral(v6, v4, nat64Prefix);
246
247 return 0;
248}
249
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700250void ClatdController::maybeStartBpf(const ClatdTracker& tracker) {
251 if (mClatEbpfMode == ClatEbpfDisabled) return;
252
253 int rv = hardwareAddressType(tracker.iface);
254 if (rv < 0) {
255 ALOGE("hardwareAddressType(%s[%d]) failure: %s", tracker.iface, tracker.ifIndex,
256 strerror(-rv));
257 return;
258 }
259
260 bool isEthernet;
261 switch (rv) {
262 case ARPHRD_ETHER:
263 isEthernet = true;
264 break;
265 case ARPHRD_RAWIP: // in Linux 4.14+ rmnet support was upstreamed and this is 519
266 case 530: // this is ARPHRD_RAWIP on some Android 4.9 kernels with rmnet
267 isEthernet = false;
268 break;
269 default:
270 ALOGE("hardwareAddressType(%s[%d]) returned unknown type %d.", tracker.iface,
271 tracker.ifIndex, rv);
272 return;
273 }
274
275 rv = getClatIngressProgFd(isEthernet);
276 if (rv < 0) {
277 ALOGE("getClatIngressProgFd(%d) failure: %s", isEthernet, strerror(-rv));
278 return;
279 }
280 unique_fd progFd(rv);
281
282 ClatIngressKey key = {
283 .iif = tracker.ifIndex,
284 .pfx96 = tracker.pfx96,
285 .local6 = tracker.v6,
286 };
287 ClatIngressValue value = {
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700288 // TODO: move all the clat code to eBPF and remove the tun interface entirely.
Maciej Żenczykowski39b0b902019-05-08 00:36:30 -0700289 .oif = tracker.v4ifIndex,
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700290 .local4 = tracker.v4,
291 };
292
293 auto ret = mClatIngressMap.writeValue(key, value, BPF_ANY);
294 if (!isOk(ret)) {
295 ALOGE("mClatIngress.Map.writeValue failure: %s", strerror(ret.code()));
296 return;
297 }
298
299 // We do tc setup *after* populating map, so scanning through map
300 // can always be used to tell us what needs cleanup.
301
302 rv = tcQdiscAddDevClsact(mNetlinkFd, tracker.ifIndex);
303 if (rv) {
304 ALOGE("tcQdiscAddDevClsact(%d[%s]) failure: %s", tracker.ifIndex, tracker.iface,
305 strerror(-rv));
306 ret = mClatIngressMap.deleteValue(key);
307 if (!isOk(ret)) ALOGE("mClatIngressMap.deleteValue failure: %s", strerror(ret.code()));
308 return;
309 }
310
Maciej Żenczykowskib140a3a2019-12-15 11:53:46 -0800311 rv = tcFilterAddDevIngressBpf(mNetlinkFd, tracker.ifIndex, progFd, isEthernet);
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700312 if (rv) {
313 if ((rv == -ENOENT) && (mClatEbpfMode == ClatEbpfMaybe)) {
Maciej Żenczykowskib140a3a2019-12-15 11:53:46 -0800314 ALOGI("tcFilterAddDevIngressBpf(%d[%s], %d): %s", tracker.ifIndex, tracker.iface,
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700315 isEthernet, strerror(-rv));
Maciej Żenczykowskib140a3a2019-12-15 11:53:46 -0800316 } else {
317 ALOGE("tcFilterAddDevIngressBpf(%d[%s], %d) failure: %s", tracker.ifIndex,
318 tracker.iface, isEthernet, strerror(-rv));
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700319 }
320 rv = tcQdiscDelDevClsact(mNetlinkFd, tracker.ifIndex);
321 if (rv)
322 ALOGE("tcQdiscDelDevClsact(%d[%s]) failure: %s", tracker.ifIndex, tracker.iface,
323 strerror(-rv));
324 ret = mClatIngressMap.deleteValue(key);
325 if (!isOk(ret)) ALOGE("mClatIngressMap.deleteValue failure: %s", strerror(ret.code()));
326 return;
327 }
328
329 // success
330}
331
Lorenzo Colitti91fd5802019-06-28 19:22:01 +0900332void ClatdController::maybeSetIptablesDropRule(bool add, const char* pfx96Str, const char* v6Str) {
333 if (mClatEbpfMode == ClatEbpfDisabled) return;
334
335 std::string cmd = StringPrintf(
336 "*raw\n"
337 "%s %s -s %s/96 -d %s -j DROP\n"
338 "COMMIT\n",
339 (add ? "-A" : "-D"), LOCAL_RAW_PREROUTING, pfx96Str, v6Str);
340
341 iptablesRestoreFunction(V6, cmd);
342}
343
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700344void ClatdController::maybeStopBpf(const ClatdTracker& tracker) {
345 if (mClatEbpfMode == ClatEbpfDisabled) return;
346
347 // No need to remove filter, since we remove qdisc it is attached to,
348 // which automatically removes everything attached to the qdisc.
349 int rv = tcQdiscDelDevClsact(mNetlinkFd, tracker.ifIndex);
350 if (rv < 0)
351 ALOGE("tcQdiscDelDevClsact(%d[%s]) failure: %s", tracker.ifIndex, tracker.iface,
352 strerror(-rv));
353
354 // We cleanup map last, so scanning through map can be used to
355 // determine what still needs cleanup.
356
357 ClatIngressKey key = {
358 .iif = tracker.ifIndex,
359 .pfx96 = tracker.pfx96,
360 .local6 = tracker.v6,
361 };
362
363 auto ret = mClatIngressMap.deleteValue(key);
364 if (!isOk(ret)) ALOGE("mClatIngressMap.deleteValue failure: %s", strerror(ret.code()));
365}
366
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900367// Finds the tracker of the clatd running on interface |interface|, or nullptr if clatd has not been
368// started on |interface|.
369ClatdController::ClatdTracker* ClatdController::getClatdTracker(const std::string& interface) {
370 auto it = mClatdTrackers.find(interface);
371 return (it == mClatdTrackers.end() ? nullptr : &it->second);
372}
373
374// Initializes a ClatdTracker for the specified interface.
Maciej Żenczykowskia56b2e62019-04-24 13:17:18 -0700375int ClatdController::ClatdTracker::init(unsigned networkId, const std::string& interface,
Maciej Żenczykowski4657e512019-05-08 06:35:08 -0700376 const std::string& v4interface,
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900377 const std::string& nat64Prefix) {
Maciej Żenczykowskia56b2e62019-04-24 13:17:18 -0700378 netId = networkId;
Lorenzo Colitti32b2e792015-01-07 15:11:30 +0900379
Lorenzo Colitti32b2e792015-01-07 15:11:30 +0900380 fwmark.netId = netId;
381 fwmark.explicitlySelected = true;
382 fwmark.protectedFromVpn = true;
383 fwmark.permission = PERMISSION_SYSTEM;
384
Lorenzo Colitti32b2e792015-01-07 15:11:30 +0900385 snprintf(fwmarkString, sizeof(fwmarkString), "0x%x", fwmark.intValue);
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900386 snprintf(netIdString, sizeof(netIdString), "%u", netId);
387 strlcpy(iface, interface.c_str(), sizeof(iface));
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700388 ifIndex = if_nametoindex(iface);
Maciej Żenczykowski4657e512019-05-08 06:35:08 -0700389 strlcpy(v4iface, v4interface.c_str(), sizeof(v4iface));
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700390 v4ifIndex = if_nametoindex(v4iface);
Lorenzo Colitti32b2e792015-01-07 15:11:30 +0900391
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900392 // Pass in everything that clatd needs: interface, a netid to use for DNS lookups, a fwmark for
393 // outgoing packets, the NAT64 prefix, and the IPv4 and IPv6 addresses.
394 // Validate the prefix and strip off the prefix length.
395 uint8_t family;
396 uint8_t prefixLen;
Maciej Żenczykowski1c06f9c2019-03-29 23:19:19 -0700397 int res = parsePrefix(nat64Prefix.c_str(), &family, &pfx96, sizeof(pfx96), &prefixLen);
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900398 // clatd only supports /96 prefixes.
Maciej Żenczykowski1c06f9c2019-03-29 23:19:19 -0700399 if (res != sizeof(pfx96)) return res;
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900400 if (family != AF_INET6) return -EAFNOSUPPORT;
401 if (prefixLen != 96) return -EINVAL;
Maciej Żenczykowski1c06f9c2019-03-29 23:19:19 -0700402 if (!inet_ntop(AF_INET6, &pfx96, pfx96String, sizeof(pfx96String))) return -errno;
Luke Huang6d301232018-08-01 14:05:18 +0800403
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900404 // Pick an IPv4 address.
405 // TODO: this picks the address based on other addresses that are assigned to interfaces, but
406 // the address is only actually assigned to an interface once clatd starts up. So we could end
407 // up with two clatd instances with the same IPv4 address.
408 // Stop doing this and instead pick a free one from the kV4Addr pool.
Maciej Żenczykowski55cacfb2019-03-30 02:01:35 -0700409 v4 = {selectIpv4Address(kV4Addr, kV4AddrLen)};
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900410 if (v4.s_addr == INADDR_NONE) {
411 ALOGE("No free IPv4 address in %s/%d", kV4AddrString, kV4AddrLen);
412 return -EADDRNOTAVAIL;
413 }
414 if (!inet_ntop(AF_INET, &v4, v4Str, sizeof(v4Str))) return -errno;
415
416 // Generate a checksum-neutral IID.
Maciej Żenczykowski1c06f9c2019-03-29 23:19:19 -0700417 if (generateIpv6Address(iface, v4, pfx96, &v6)) {
418 ALOGE("Unable to find global source address on %s for %s", iface, pfx96String);
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900419 return -EADDRNOTAVAIL;
420 }
421 if (!inet_ntop(AF_INET6, &v6, v6Str, sizeof(v6Str))) return -errno;
422
Maciej Żenczykowski1c06f9c2019-03-29 23:19:19 -0700423 ALOGD("starting clatd on %s v4=%s v6=%s pfx96=%s", iface, v4Str, v6Str, pfx96String);
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900424 return 0;
425}
426
427int ClatdController::startClatd(const std::string& interface, const std::string& nat64Prefix,
428 std::string* v6Str) {
Maciej Żenczykowski56280272019-03-30 03:32:51 -0700429 std::lock_guard guard(mutex);
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700430
431 // 1. fail if pre-existing tracker already exists
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900432 ClatdTracker* existing = getClatdTracker(interface);
433 if (existing != nullptr) {
434 ALOGE("clatd pid=%d already started on %s", existing->pid, interface.c_str());
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700435 return -EBUSY;
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900436 }
437
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700438 // 2. get network id associated with this external interface
Maciej Żenczykowskia56b2e62019-04-24 13:17:18 -0700439 unsigned networkId = mNetCtrl->getNetworkForInterface(interface.c_str());
440 if (networkId == NETID_UNSET) {
441 ALOGE("Interface %s not assigned to any netId", interface.c_str());
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700442 return -ENODEV;
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900443 }
JP Abgrall69261cb2014-06-19 18:35:24 -0700444
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700445 // 3. open the tun device in non blocking mode as required by clatd
446 int res = open("/dev/tun", O_RDWR | O_NONBLOCK | O_CLOEXEC);
447 if (res == -1) {
448 res = errno;
449 ALOGE("open of tun device failed (%s)", strerror(res));
450 return -res;
451 }
452 unique_fd tmpTunFd(res);
453
454 // 4. create the v4-... tun interface
Maciej Żenczykowski4657e512019-05-08 06:35:08 -0700455 std::string v4interface("v4-");
456 v4interface += interface;
457
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700458 struct ifreq ifr = {
459 .ifr_flags = IFF_TUN,
460 };
Maciej Żenczykowski4657e512019-05-08 06:35:08 -0700461 strlcpy(ifr.ifr_name, v4interface.c_str(), sizeof(ifr.ifr_name));
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700462
463 res = ioctl(tmpTunFd, TUNSETIFF, &ifr, sizeof(ifr));
464 if (res == -1) {
465 res = errno;
466 ALOGE("ioctl(TUNSETIFF) failed (%s)", strerror(res));
467 return -res;
468 }
469
Maciej Żenczykowskia1699952019-05-11 17:07:44 -0700470 // disable IPv6 on it - failing to do so is not a critical error
471 res = InterfaceController::setEnableIPv6(v4interface.c_str(), 0);
472 if (res) ALOGE("setEnableIPv6 %s failed (%s)", v4interface.c_str(), strerror(res));
473
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700474 // 5. initialize tracker object
Maciej Żenczykowskia56b2e62019-04-24 13:17:18 -0700475 ClatdTracker tracker;
Maciej Żenczykowski4657e512019-05-08 06:35:08 -0700476 int ret = tracker.init(networkId, interface, v4interface, nat64Prefix);
Maciej Żenczykowskia56b2e62019-04-24 13:17:18 -0700477 if (ret) return ret;
478
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700479 // 6. create a throwaway socket to reserve a file descriptor number
480 res = socket(AF_INET6, SOCK_DGRAM | SOCK_CLOEXEC, 0);
481 if (res == -1) {
482 res = errno;
483 ALOGE("socket(ipv6/udp) failed (%s)", strerror(res));
484 return -res;
485 }
486 unique_fd passedTunFd(res);
487
488 // 7. this is the FD we'll pass to clatd on the cli, so need it as a string
489 char passedTunFdStr[INT32_STRLEN];
490 snprintf(passedTunFdStr, sizeof(passedTunFdStr), "%d", passedTunFd.get());
491
492 // 8. we're going to use this as argv[0] to clatd to make ps output more useful
Lorenzo Colittiac7fefc2014-10-20 17:14:13 +0900493 std::string progname("clatd-");
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900494 progname += tracker.iface;
Daniel Drown0da73fc2012-06-20 16:51:39 -0500495
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900496 // clang-format off
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700497 const char* args[] = {progname.c_str(),
498 "-i", tracker.iface,
499 "-n", tracker.netIdString,
500 "-m", tracker.fwmarkString,
501 "-p", tracker.pfx96String,
502 "-4", tracker.v4Str,
503 "-6", tracker.v6Str,
504 "-t", passedTunFdStr,
505 nullptr};
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900506 // clang-format on
507
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700508 // 9. register vfork requirement
509 posix_spawnattr_t attr;
510 res = posix_spawnattr_init(&attr);
511 if (res) {
512 ALOGE("posix_spawnattr_init failed (%s)", strerror(res));
513 return -res;
514 }
515 const android::base::ScopeGuard attrGuard = [&] { posix_spawnattr_destroy(&attr); };
516 res = posix_spawnattr_setflags(&attr, POSIX_SPAWN_USEVFORK);
517 if (res) {
518 ALOGE("posix_spawnattr_setflags failed (%s)", strerror(res));
519 return -res;
520 }
521
522 // 10. register dup2() action: this is what 'clears' the CLOEXEC flag
523 // on the tun fd that we want the child clatd process to inherit
524 // (this will happen after the vfork, and before the execve)
525 posix_spawn_file_actions_t fa;
526 res = posix_spawn_file_actions_init(&fa);
527 if (res) {
528 ALOGE("posix_spawn_file_actions_init failed (%s)", strerror(res));
529 return -res;
530 }
531 const android::base::ScopeGuard faGuard = [&] { posix_spawn_file_actions_destroy(&fa); };
532 res = posix_spawn_file_actions_adddup2(&fa, tmpTunFd, passedTunFd);
533 if (res) {
534 ALOGE("posix_spawn_file_actions_adddup2 failed (%s)", strerror(res));
535 return -res;
536 }
537
Lorenzo Colitti91fd5802019-06-28 19:22:01 +0900538 // 11. If necessary, add the drop rule for iptables.
539 maybeSetIptablesDropRule(true, tracker.pfx96String, tracker.v6Str);
540
541 // 12. actually perform vfork/dup2/execve
Maciej Żenczykowskif4b44fe2019-04-08 16:18:50 -0700542 res = posix_spawn(&tracker.pid, kClatdPath, &fa, &attr, (char* const*)args, nullptr);
Luke Huang40962442019-02-26 11:46:10 +0800543 if (res) {
544 ALOGE("posix_spawn failed (%s)", strerror(res));
Luke Huang6d301232018-08-01 14:05:18 +0800545 return -res;
Daniel Drown0da73fc2012-06-20 16:51:39 -0500546 }
547
Lorenzo Colitti91fd5802019-06-28 19:22:01 +0900548 // 13. configure eBPF offload - if possible
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700549 maybeStartBpf(tracker);
550
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900551 mClatdTrackers[interface] = tracker;
552 ALOGD("clatd started on %s", interface.c_str());
Daniel Drown0da73fc2012-06-20 16:51:39 -0500553
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900554 *v6Str = tracker.v6Str;
Daniel Drown0da73fc2012-06-20 16:51:39 -0500555 return 0;
556}
557
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900558int ClatdController::stopClatd(const std::string& interface) {
Maciej Żenczykowski56280272019-03-30 03:32:51 -0700559 std::lock_guard guard(mutex);
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900560 ClatdTracker* tracker = getClatdTracker(interface);
Lorenzo Colittiac7fefc2014-10-20 17:14:13 +0900561
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900562 if (tracker == nullptr) {
Daniel Drown0da73fc2012-06-20 16:51:39 -0500563 ALOGE("clatd already stopped");
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900564 return -ENODEV;
Daniel Drown0da73fc2012-06-20 16:51:39 -0500565 }
566
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900567 ALOGD("Stopping clatd pid=%d on %s", tracker->pid, interface.c_str());
Daniel Drown0da73fc2012-06-20 16:51:39 -0500568
Maciej Żenczykowski1c086e52019-03-29 23:13:49 -0700569 maybeStopBpf(*tracker);
570
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900571 kill(tracker->pid, SIGTERM);
572 waitpid(tracker->pid, nullptr, 0);
Lorenzo Colitti91fd5802019-06-28 19:22:01 +0900573
574 maybeSetIptablesDropRule(false, tracker->pfx96String, tracker->v6Str);
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900575 mClatdTrackers.erase(interface);
Daniel Drown0da73fc2012-06-20 16:51:39 -0500576
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900577 ALOGD("clatd on %s stopped", interface.c_str());
Daniel Drown0da73fc2012-06-20 16:51:39 -0500578
579 return 0;
580}
581
Maciej Żenczykowski1afbd992019-12-16 11:44:14 -0800582void ClatdController::dumpEgress(DumpWriter& dw) {
583 int mapFd = getClatEgressMapFd();
584 if (mapFd < 0) return; // if unsupported just don't dump anything
585 BpfMap<ClatEgressKey, ClatEgressValue> configMap(mapFd);
586
587 ScopedIndent bpfIndent(dw);
588 dw.println("BPF egress map: iif(iface) v4Addr -> v6Addr nat64Prefix oif(iface)");
589
590 ScopedIndent bpfDetailIndent(dw);
591 const auto printClatMap = [&dw](const ClatEgressKey& key, const ClatEgressValue& value,
592 const BpfMap<ClatEgressKey, ClatEgressValue>&) {
593 char iifStr[IFNAMSIZ] = "?";
594 char local4Str[INET_ADDRSTRLEN] = "?";
595 char local6Str[INET6_ADDRSTRLEN] = "?";
596 char pfx96Str[INET6_ADDRSTRLEN] = "?";
597 char oifStr[IFNAMSIZ] = "?";
598
599 if_indextoname(key.iif, iifStr);
600 inet_ntop(AF_INET, &key.local4, local4Str, sizeof(local4Str));
601 inet_ntop(AF_INET6, &value.local6, local6Str, sizeof(local6Str));
602 inet_ntop(AF_INET6, &value.pfx96, pfx96Str, sizeof(pfx96Str));
603 if_indextoname(value.oif, oifStr);
604
605 dw.println("%u(%s) %s -> %s %s/96 %u(%s)", key.iif, iifStr, local4Str, local6Str, pfx96Str,
606 value.oif, oifStr);
607 return netdutils::status::ok;
608 };
609 auto res = configMap.iterateWithValue(printClatMap);
610 if (!isOk(res)) {
611 dw.println("Error printing BPF map: %s", res.msg().c_str());
612 }
613}
614
Maciej Żenczykowski7dffa6f2019-12-16 11:20:44 -0800615void ClatdController::dumpIngress(DumpWriter& dw) {
Maciej Żenczykowskiabb2cbf2019-04-01 01:29:29 -0700616 int mapFd = getClatIngressMapFd();
617 if (mapFd < 0) return; // if unsupported just don't dump anything
618 BpfMap<ClatIngressKey, ClatIngressValue> configMap(mapFd);
Maciej Żenczykowski55262712019-03-29 23:44:56 -0700619
620 ScopedIndent bpfIndent(dw);
621 dw.println("BPF ingress map: iif(iface) nat64Prefix v6Addr -> v4Addr oif(iface)");
622
623 ScopedIndent bpfDetailIndent(dw);
Maciej Żenczykowski55262712019-03-29 23:44:56 -0700624 const auto printClatMap = [&dw](const ClatIngressKey& key, const ClatIngressValue& value,
625 const BpfMap<ClatIngressKey, ClatIngressValue>&) {
626 char iifStr[IFNAMSIZ] = "?";
627 char pfx96Str[INET6_ADDRSTRLEN] = "?";
628 char local6Str[INET6_ADDRSTRLEN] = "?";
629 char local4Str[INET_ADDRSTRLEN] = "?";
630 char oifStr[IFNAMSIZ] = "?";
631
632 if_indextoname(key.iif, iifStr);
633 inet_ntop(AF_INET6, &key.pfx96, pfx96Str, sizeof(pfx96Str));
634 inet_ntop(AF_INET6, &key.local6, local6Str, sizeof(local6Str));
635 inet_ntop(AF_INET, &value.local4, local4Str, sizeof(local4Str));
636 if_indextoname(value.oif, oifStr);
637
638 dw.println("%u(%s) %s/96 %s -> %s %u(%s)", key.iif, iifStr, pfx96Str, local6Str, local4Str,
639 value.oif, oifStr);
640 return netdutils::status::ok;
641 };
642 auto res = configMap.iterateWithValue(printClatMap);
643 if (!isOk(res)) {
644 dw.println("Error printing BPF map: %s", res.msg().c_str());
645 }
646}
647
Maciej Żenczykowski4c262172019-12-16 11:31:24 -0800648void ClatdController::dumpTrackers(DumpWriter& dw) {
649 ScopedIndent trackerIndent(dw);
650 dw.println("Trackers: iif[iface] nat64Prefix v6Addr -> v4Addr v4iif[v4iface] [netId]");
651
652 ScopedIndent trackerDetailIndent(dw);
653 for (const auto& pair : mClatdTrackers) {
654 const ClatdTracker& tracker = pair.second;
655 dw.println("%u[%s] %s/96 %s -> %s %u[%s] [%u]", tracker.ifIndex, tracker.iface,
656 tracker.pfx96String, tracker.v6Str, tracker.v4Str, tracker.v4ifIndex,
657 tracker.v4iface, tracker.netId);
658 }
659}
660
Maciej Żenczykowski7dffa6f2019-12-16 11:20:44 -0800661void ClatdController::dump(DumpWriter& dw) {
662 std::lock_guard guard(mutex);
663
664 ScopedIndent clatdIndent(dw);
665 dw.println("ClatdController");
666
Maciej Żenczykowski4c262172019-12-16 11:31:24 -0800667 dumpTrackers(dw);
Maciej Żenczykowski7dffa6f2019-12-16 11:20:44 -0800668 dumpIngress(dw);
Maciej Żenczykowski1afbd992019-12-16 11:44:14 -0800669 dumpEgress(dw);
Maciej Żenczykowski7dffa6f2019-12-16 11:20:44 -0800670}
671
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900672auto ClatdController::isIpv4AddressFreeFunc = isIpv4AddressFree;
Lorenzo Colitti91fd5802019-06-28 19:22:01 +0900673auto ClatdController::iptablesRestoreFunction = execIptablesRestore;
Lorenzo Colitti7ef8c0f2019-01-11 22:34:58 +0900674
Lorenzo Colitti7035f222017-02-13 18:29:00 +0900675} // namespace net
676} // namespace android