1 /*
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 */
16
17 #include <map>
18 #include <string>
19
20 #include <arpa/inet.h>
21 #include <errno.h>
22 #include <linux/if_tun.h>
23 #include <linux/ioctl.h>
24 #include <net/if.h>
25 #include <netinet/in.h>
26 #include <spawn.h>
27 #include <sys/types.h>
28 #include <sys/wait.h>
29 #include <unistd.h>
30
31 #define LOG_TAG "ClatdController"
32 #include <log/log.h>
33
34 #include "ClatdController.h"
35 #include "InterfaceController.h"
36
37 #include "android-base/properties.h"
38 #include "android-base/scopeguard.h"
39 #include "android-base/stringprintf.h"
40 #include "android-base/unique_fd.h"
41 #include "bpf/BpfMap.h"
42 #include "netdbpf/bpf_shared.h"
43 #include "netdutils/DumpWriter.h"
44
45 extern "C" {
46 #include "netutils/checksum.h"
47 }
48
49 #include "Fwmark.h"
50 #include "NetdConstants.h"
51 #include "NetworkController.h"
52 #include "OffloadUtils.h"
53 #include "netid_client.h"
54
55 static const char* kClatdPath = "/system/bin/clatd";
56
57 // For historical reasons, start with 192.0.0.4, and after that, use all subsequent addresses in
58 // 192.0.0.0/29 (RFC 7335).
59 static const char* kV4AddrString = "192.0.0.4";
60 static const in_addr kV4Addr = {inet_addr(kV4AddrString)};
61 static const int kV4AddrLen = 29;
62
63 using android::base::Result;
64 using android::base::StringPrintf;
65 using android::base::unique_fd;
66 using android::bpf::BpfMap;
67 using android::netdutils::DumpWriter;
68 using android::netdutils::ScopedIndent;
69
70 namespace android {
71 namespace net {
72
init(void)73 void ClatdController::init(void) {
74 std::lock_guard guard(mutex);
75
76 int rv = getClatEgress4MapFd();
77 if (rv < 0) {
78 ALOGE("getClatEgress4MapFd() failure: %s", strerror(-rv));
79 return;
80 }
81 mClatEgress4Map.reset(rv);
82
83 rv = getClatIngress6MapFd();
84 if (rv < 0) {
85 ALOGE("getClatIngress6MapFd() failure: %s", strerror(-rv));
86 mClatEgress4Map.reset(-1);
87 return;
88 }
89 mClatIngress6Map.reset(rv);
90
91 mClatEgress4Map.clear();
92 mClatIngress6Map.clear();
93 }
94
isIpv4AddressFree(in_addr_t addr)95 bool ClatdController::isIpv4AddressFree(in_addr_t addr) {
96 int s = socket(AF_INET, SOCK_DGRAM | SOCK_CLOEXEC, 0);
97 if (s == -1) {
98 return 0;
99 }
100
101 // Attempt to connect to the address. If the connection succeeds and getsockname returns the
102 // same then the address is already assigned to the system and we can't use it.
103 struct sockaddr_in sin = {
104 .sin_family = AF_INET,
105 .sin_port = htons(53),
106 .sin_addr = {addr},
107 };
108 socklen_t len = sizeof(sin);
109 bool inuse = connect(s, (struct sockaddr*)&sin, sizeof(sin)) == 0 &&
110 getsockname(s, (struct sockaddr*)&sin, &len) == 0 && (size_t)len >= sizeof(sin) &&
111 sin.sin_addr.s_addr == addr;
112
113 close(s);
114 return !inuse;
115 }
116
117 // Picks a free IPv4 address, starting from ip and trying all addresses in the prefix in order.
118 // ip - the IP address from the configuration file
119 // prefixlen - the length of the prefix from which addresses may be selected.
120 // returns: the IPv4 address, or INADDR_NONE if no addresses were available
selectIpv4Address(const in_addr ip,int16_t prefixlen)121 in_addr_t ClatdController::selectIpv4Address(const in_addr ip, int16_t prefixlen) {
122 // Don't accept prefixes that are too large because we scan addresses one by one.
123 if (prefixlen < 16 || prefixlen > 32) {
124 return INADDR_NONE;
125 }
126
127 // All these are in host byte order.
128 in_addr_t mask = 0xffffffff >> (32 - prefixlen) << (32 - prefixlen);
129 in_addr_t ipv4 = ntohl(ip.s_addr);
130 in_addr_t first_ipv4 = ipv4;
131 in_addr_t prefix = ipv4 & mask;
132
133 // Pick the first IPv4 address in the pool, wrapping around if necessary.
134 // So, for example, 192.0.0.4 -> 192.0.0.5 -> 192.0.0.6 -> 192.0.0.7 -> 192.0.0.0.
135 do {
136 if (isIpv4AddressFreeFunc(htonl(ipv4))) {
137 return htonl(ipv4);
138 }
139 ipv4 = prefix | ((ipv4 + 1) & ~mask);
140 } while (ipv4 != first_ipv4);
141
142 return INADDR_NONE;
143 }
144
145 // Alters the bits in the IPv6 address to make them checksum neutral with v4 and nat64Prefix.
makeChecksumNeutral(in6_addr * v6,const in_addr v4,const in6_addr & nat64Prefix)146 void ClatdController::makeChecksumNeutral(in6_addr* v6, const in_addr v4,
147 const in6_addr& nat64Prefix) {
148 // Fill last 8 bytes of IPv6 address with random bits.
149 arc4random_buf(&v6->s6_addr[8], 8);
150
151 // Make the IID checksum-neutral. That is, make it so that:
152 // checksum(Local IPv4 | Remote IPv4) = checksum(Local IPv6 | Remote IPv6)
153 // in other words (because remote IPv6 = NAT64 prefix | Remote IPv4):
154 // checksum(Local IPv4) = checksum(Local IPv6 | NAT64 prefix)
155 // Do this by adjusting the two bytes in the middle of the IID.
156
157 uint16_t middlebytes = (v6->s6_addr[11] << 8) + v6->s6_addr[12];
158
159 uint32_t c1 = ip_checksum_add(0, &v4, sizeof(v4));
160 uint32_t c2 = ip_checksum_add(0, &nat64Prefix, sizeof(nat64Prefix)) +
161 ip_checksum_add(0, v6, sizeof(*v6));
162
163 uint16_t delta = ip_checksum_adjust(middlebytes, c1, c2);
164 v6->s6_addr[11] = delta >> 8;
165 v6->s6_addr[12] = delta & 0xff;
166 }
167
168 // Picks a random interface ID that is checksum neutral with the IPv4 address and the NAT64 prefix.
generateIpv6Address(const char * iface,const in_addr v4,const in6_addr & nat64Prefix,in6_addr * v6)169 int ClatdController::generateIpv6Address(const char* iface, const in_addr v4,
170 const in6_addr& nat64Prefix, in6_addr* v6) {
171 unique_fd s(socket(AF_INET6, SOCK_DGRAM | SOCK_CLOEXEC, 0));
172 if (s == -1) return -errno;
173
174 if (setsockopt(s, SOL_SOCKET, SO_BINDTODEVICE, iface, strlen(iface) + 1) == -1) {
175 return -errno;
176 }
177
178 sockaddr_in6 sin6 = {.sin6_family = AF_INET6, .sin6_addr = nat64Prefix};
179 if (connect(s, reinterpret_cast<struct sockaddr*>(&sin6), sizeof(sin6)) == -1) {
180 return -errno;
181 }
182
183 socklen_t len = sizeof(sin6);
184 if (getsockname(s, reinterpret_cast<struct sockaddr*>(&sin6), &len) == -1) {
185 return -errno;
186 }
187
188 *v6 = sin6.sin6_addr;
189
190 if (IN6_IS_ADDR_UNSPECIFIED(v6) || IN6_IS_ADDR_LOOPBACK(v6) || IN6_IS_ADDR_LINKLOCAL(v6) ||
191 IN6_IS_ADDR_SITELOCAL(v6) || IN6_IS_ADDR_ULA(v6)) {
192 return -ENETUNREACH;
193 }
194
195 makeChecksumNeutral(v6, v4, nat64Prefix);
196
197 return 0;
198 }
199
maybeStartBpf(const ClatdTracker & tracker)200 void ClatdController::maybeStartBpf(const ClatdTracker& tracker) {
201 auto isEthernet = android::net::isEthernet(tracker.iface);
202 if (!isEthernet.ok()) {
203 ALOGE("isEthernet(%s[%d]) failure: %s", tracker.iface, tracker.ifIndex,
204 isEthernet.error().message().c_str());
205 return;
206 }
207
208 // This program will be attached to the v4-* interface which is a TUN and thus always rawip.
209 int rv = getClatEgress4ProgFd(RAWIP);
210 if (rv < 0) {
211 ALOGE("getClatEgress4ProgFd(RAWIP) failure: %s", strerror(-rv));
212 return;
213 }
214 unique_fd txRawIpProgFd(rv);
215
216 rv = getClatIngress6ProgFd(isEthernet.value());
217 if (rv < 0) {
218 ALOGE("getClatIngress6ProgFd(%d) failure: %s", isEthernet.value(), strerror(-rv));
219 return;
220 }
221 unique_fd rxProgFd(rv);
222
223 ClatEgress4Key txKey = {
224 .iif = tracker.v4ifIndex,
225 .local4 = tracker.v4,
226 };
227 ClatEgress4Value txValue = {
228 .oif = tracker.ifIndex,
229 .local6 = tracker.v6,
230 .pfx96 = tracker.pfx96,
231 .oifIsEthernet = isEthernet.value(),
232 };
233
234 auto ret = mClatEgress4Map.writeValue(txKey, txValue, BPF_ANY);
235 if (!ret.ok()) {
236 ALOGE("mClatEgress4Map.writeValue failure: %s", strerror(ret.error().code()));
237 return;
238 }
239
240 ClatIngress6Key rxKey = {
241 .iif = tracker.ifIndex,
242 .pfx96 = tracker.pfx96,
243 .local6 = tracker.v6,
244 };
245 ClatIngress6Value rxValue = {
246 // TODO: move all the clat code to eBPF and remove the tun interface entirely.
247 .oif = tracker.v4ifIndex,
248 .local4 = tracker.v4,
249 };
250
251 ret = mClatIngress6Map.writeValue(rxKey, rxValue, BPF_ANY);
252 if (!ret.ok()) {
253 ALOGE("mClatIngress6Map.writeValue failure: %s", strerror(ret.error().code()));
254 ret = mClatEgress4Map.deleteValue(txKey);
255 if (!ret.ok())
256 ALOGE("mClatEgress4Map.deleteValue failure: %s", strerror(ret.error().code()));
257 return;
258 }
259
260 // We do tc setup *after* populating the maps, so scanning through them
261 // can always be used to tell us what needs cleanup.
262
263 // Usually the clsact will be added in RouteController::addInterfaceToPhysicalNetwork.
264 // But clat is started before the v4- interface is added to the network. The clat startup have
265 // to add clsact of v4- tun interface first for adding bpf filter in maybeStartBpf.
266 // TODO: move "qdisc add clsact" of v4- tun interface out from ClatdController.
267 rv = tcQdiscAddDevClsact(tracker.v4ifIndex);
268 if (rv) {
269 ALOGE("tcQdiscAddDevClsact(%d[%s]) failure: %s", tracker.v4ifIndex, tracker.v4iface,
270 strerror(-rv));
271 ret = mClatEgress4Map.deleteValue(txKey);
272 if (!ret.ok())
273 ALOGE("mClatEgress4Map.deleteValue failure: %s", strerror(ret.error().code()));
274 ret = mClatIngress6Map.deleteValue(rxKey);
275 if (!ret.ok())
276 ALOGE("mClatIngress6Map.deleteValue failure: %s", strerror(ret.error().code()));
277 return;
278 }
279
280 rv = tcFilterAddDevEgressClatIpv4(tracker.v4ifIndex, txRawIpProgFd, RAWIP);
281 if (rv) {
282 ALOGE("tcFilterAddDevEgressClatIpv4(%d[%s], RAWIP) failure: %s", tracker.v4ifIndex,
283 tracker.v4iface, strerror(-rv));
284
285 // The v4- interface clsact is not deleted for unwinding error because once it is created
286 // with interface addition, the lifetime is till interface deletion. Moreover, the clsact
287 // has no clat filter now. It should not break anything.
288
289 ret = mClatEgress4Map.deleteValue(txKey);
290 if (!ret.ok())
291 ALOGE("mClatEgress4Map.deleteValue failure: %s", strerror(ret.error().code()));
292 ret = mClatIngress6Map.deleteValue(rxKey);
293 if (!ret.ok())
294 ALOGE("mClatIngress6Map.deleteValue failure: %s", strerror(ret.error().code()));
295 return;
296 }
297
298 rv = tcFilterAddDevIngressClatIpv6(tracker.ifIndex, rxProgFd, isEthernet.value());
299 if (rv) {
300 ALOGE("tcFilterAddDevIngressClatIpv6(%d[%s], %d) failure: %s", tracker.ifIndex,
301 tracker.iface, isEthernet.value(), strerror(-rv));
302 rv = tcFilterDelDevEgressClatIpv4(tracker.v4ifIndex);
303 if (rv) {
304 ALOGE("tcFilterDelDevEgressClatIpv4(%d[%s]) failure: %s", tracker.v4ifIndex,
305 tracker.v4iface, strerror(-rv));
306 }
307
308 // The v4- interface clsact is not deleted. See the reason in the error unwinding code of
309 // the egress filter attaching of v4- tun interface.
310
311 ret = mClatEgress4Map.deleteValue(txKey);
312 if (!ret.ok())
313 ALOGE("mClatEgress4Map.deleteValue failure: %s", strerror(ret.error().code()));
314 ret = mClatIngress6Map.deleteValue(rxKey);
315 if (!ret.ok())
316 ALOGE("mClatIngress6Map.deleteValue failure: %s", strerror(ret.error().code()));
317 return;
318 }
319
320 // success
321 }
322
setIptablesDropRule(bool add,const char * iface,const char * pfx96Str,const char * v6Str)323 void ClatdController::setIptablesDropRule(bool add, const char* iface, const char* pfx96Str,
324 const char* v6Str) {
325 std::string cmd = StringPrintf(
326 "*raw\n"
327 "%s %s -i %s -s %s/96 -d %s -j DROP\n"
328 "COMMIT\n",
329 (add ? "-A" : "-D"), LOCAL_RAW_PREROUTING, iface, pfx96Str, v6Str);
330
331 iptablesRestoreFunction(V6, cmd);
332 }
333
maybeStopBpf(const ClatdTracker & tracker)334 void ClatdController::maybeStopBpf(const ClatdTracker& tracker) {
335 int rv = tcFilterDelDevIngressClatIpv6(tracker.ifIndex);
336 if (rv < 0) {
337 ALOGE("tcFilterDelDevIngressClatIpv6(%d[%s]) failure: %s", tracker.ifIndex, tracker.iface,
338 strerror(-rv));
339 }
340
341 rv = tcFilterDelDevEgressClatIpv4(tracker.v4ifIndex);
342 if (rv < 0) {
343 ALOGE("tcFilterDelDevEgressClatIpv4(%d[%s]) failure: %s", tracker.v4ifIndex,
344 tracker.v4iface, strerror(-rv));
345 }
346
347 // We cleanup the maps last, so scanning through them can be used to
348 // determine what still needs cleanup.
349
350 ClatEgress4Key txKey = {
351 .iif = tracker.v4ifIndex,
352 .local4 = tracker.v4,
353 };
354
355 auto ret = mClatEgress4Map.deleteValue(txKey);
356 if (!ret.ok()) ALOGE("mClatEgress4Map.deleteValue failure: %s", strerror(ret.error().code()));
357
358 ClatIngress6Key rxKey = {
359 .iif = tracker.ifIndex,
360 .pfx96 = tracker.pfx96,
361 .local6 = tracker.v6,
362 };
363
364 ret = mClatIngress6Map.deleteValue(rxKey);
365 if (!ret.ok()) ALOGE("mClatIngress6Map.deleteValue failure: %s", strerror(ret.error().code()));
366 }
367
368 // Finds the tracker of the clatd running on interface |interface|, or nullptr if clatd has not been
369 // started on |interface|.
getClatdTracker(const std::string & interface)370 ClatdController::ClatdTracker* ClatdController::getClatdTracker(const std::string& interface) {
371 auto it = mClatdTrackers.find(interface);
372 return (it == mClatdTrackers.end() ? nullptr : &it->second);
373 }
374
375 // Initializes a ClatdTracker for the specified interface.
init(unsigned networkId,const std::string & interface,const std::string & v4interface,const std::string & nat64Prefix)376 int ClatdController::ClatdTracker::init(unsigned networkId, const std::string& interface,
377 const std::string& v4interface,
378 const std::string& nat64Prefix) {
379 fwmark.netId = networkId;
380 fwmark.explicitlySelected = true;
381 fwmark.protectedFromVpn = true;
382 fwmark.permission = PERMISSION_SYSTEM;
383
384 snprintf(fwmarkString, sizeof(fwmarkString), "0x%x", fwmark.intValue);
385 strlcpy(iface, interface.c_str(), sizeof(iface));
386 ifIndex = if_nametoindex(iface);
387 strlcpy(v4iface, v4interface.c_str(), sizeof(v4iface));
388 v4ifIndex = if_nametoindex(v4iface);
389
390 // Pass in everything that clatd needs: interface, a fwmark for outgoing packets, the NAT64
391 // prefix, and the IPv4 and IPv6 addresses.
392 // Validate the prefix and strip off the prefix length.
393 uint8_t family;
394 uint8_t prefixLen;
395 int res = parsePrefix(nat64Prefix.c_str(), &family, &pfx96, sizeof(pfx96), &prefixLen);
396 // clatd only supports /96 prefixes.
397 if (res != sizeof(pfx96)) return res;
398 if (family != AF_INET6) return -EAFNOSUPPORT;
399 if (prefixLen != 96) return -EINVAL;
400 if (!inet_ntop(AF_INET6, &pfx96, pfx96String, sizeof(pfx96String))) return -errno;
401
402 // Pick an IPv4 address.
403 // TODO: this picks the address based on other addresses that are assigned to interfaces, but
404 // the address is only actually assigned to an interface once clatd starts up. So we could end
405 // up with two clatd instances with the same IPv4 address.
406 // Stop doing this and instead pick a free one from the kV4Addr pool.
407 v4 = {selectIpv4Address(kV4Addr, kV4AddrLen)};
408 if (v4.s_addr == INADDR_NONE) {
409 ALOGE("No free IPv4 address in %s/%d", kV4AddrString, kV4AddrLen);
410 return -EADDRNOTAVAIL;
411 }
412 if (!inet_ntop(AF_INET, &v4, v4Str, sizeof(v4Str))) return -errno;
413
414 // Generate a checksum-neutral IID.
415 if (generateIpv6Address(iface, v4, pfx96, &v6)) {
416 ALOGE("Unable to find global source address on %s for %s", iface, pfx96String);
417 return -EADDRNOTAVAIL;
418 }
419 if (!inet_ntop(AF_INET6, &v6, v6Str, sizeof(v6Str))) return -errno;
420
421 ALOGD("starting clatd on %s v4=%s v6=%s pfx96=%s", iface, v4Str, v6Str, pfx96String);
422 return 0;
423 }
424
startClatd(const std::string & interface,const std::string & nat64Prefix,std::string * v6Str)425 int ClatdController::startClatd(const std::string& interface, const std::string& nat64Prefix,
426 std::string* v6Str) {
427 std::lock_guard guard(mutex);
428
429 // 1. fail if pre-existing tracker already exists
430 ClatdTracker* existing = getClatdTracker(interface);
431 if (existing != nullptr) {
432 ALOGE("clatd pid=%d already started on %s", existing->pid, interface.c_str());
433 return -EBUSY;
434 }
435
436 // 2. get network id associated with this external interface
437 unsigned networkId = mNetCtrl->getNetworkForInterface(interface.c_str());
438 if (networkId == NETID_UNSET) {
439 ALOGE("Interface %s not assigned to any netId", interface.c_str());
440 return -ENODEV;
441 }
442
443 // 3. open the tun device in non blocking mode as required by clatd
444 int res = open("/dev/net/tun", O_RDWR | O_NONBLOCK | O_CLOEXEC);
445 if (res == -1) {
446 res = errno;
447 ALOGE("open of tun device failed (%s)", strerror(res));
448 return -res;
449 }
450 unique_fd tmpTunFd(res);
451
452 // 4. create the v4-... tun interface
453 std::string v4interface("v4-");
454 v4interface += interface;
455
456 struct ifreq ifr = {
457 .ifr_flags = IFF_TUN,
458 };
459 strlcpy(ifr.ifr_name, v4interface.c_str(), sizeof(ifr.ifr_name));
460
461 res = ioctl(tmpTunFd, TUNSETIFF, &ifr, sizeof(ifr));
462 if (res == -1) {
463 res = errno;
464 ALOGE("ioctl(TUNSETIFF) failed (%s)", strerror(res));
465 return -res;
466 }
467
468 // disable IPv6 on it - failing to do so is not a critical error
469 res = InterfaceController::setEnableIPv6(v4interface.c_str(), 0);
470 if (res) ALOGE("setEnableIPv6 %s failed (%s)", v4interface.c_str(), strerror(res));
471
472 // 5. initialize tracker object
473 ClatdTracker tracker;
474 int ret = tracker.init(networkId, interface, v4interface, nat64Prefix);
475 if (ret) return ret;
476
477 // 6. create a throwaway socket to reserve a file descriptor number
478 res = socket(AF_INET6, SOCK_DGRAM | SOCK_CLOEXEC, 0);
479 if (res == -1) {
480 res = errno;
481 ALOGE("socket(ipv6/udp) failed (%s)", strerror(res));
482 return -res;
483 }
484 unique_fd passedTunFd(res);
485
486 // 7. this is the FD we'll pass to clatd on the cli, so need it as a string
487 char passedTunFdStr[INT32_STRLEN];
488 snprintf(passedTunFdStr, sizeof(passedTunFdStr), "%d", passedTunFd.get());
489
490 // 8. we're going to use this as argv[0] to clatd to make ps output more useful
491 std::string progname("clatd-");
492 progname += tracker.iface;
493
494 // clang-format off
495 const char* args[] = {progname.c_str(),
496 "-i", tracker.iface,
497 "-m", tracker.fwmarkString,
498 "-p", tracker.pfx96String,
499 "-4", tracker.v4Str,
500 "-6", tracker.v6Str,
501 "-t", passedTunFdStr,
502 nullptr};
503 // clang-format on
504
505 // 9. register vfork requirement
506 posix_spawnattr_t attr;
507 res = posix_spawnattr_init(&attr);
508 if (res) {
509 ALOGE("posix_spawnattr_init failed (%s)", strerror(res));
510 return -res;
511 }
512 const android::base::ScopeGuard attrGuard = [&] { posix_spawnattr_destroy(&attr); };
513 res = posix_spawnattr_setflags(&attr, POSIX_SPAWN_USEVFORK);
514 if (res) {
515 ALOGE("posix_spawnattr_setflags failed (%s)", strerror(res));
516 return -res;
517 }
518
519 // 10. register dup2() action: this is what 'clears' the CLOEXEC flag
520 // on the tun fd that we want the child clatd process to inherit
521 // (this will happen after the vfork, and before the execve)
522 posix_spawn_file_actions_t fa;
523 res = posix_spawn_file_actions_init(&fa);
524 if (res) {
525 ALOGE("posix_spawn_file_actions_init failed (%s)", strerror(res));
526 return -res;
527 }
528 const android::base::ScopeGuard faGuard = [&] { posix_spawn_file_actions_destroy(&fa); };
529 res = posix_spawn_file_actions_adddup2(&fa, tmpTunFd, passedTunFd);
530 if (res) {
531 ALOGE("posix_spawn_file_actions_adddup2 failed (%s)", strerror(res));
532 return -res;
533 }
534
535 // 11. add the drop rule for iptables.
536 setIptablesDropRule(true, tracker.iface, tracker.pfx96String, tracker.v6Str);
537
538 // 12. actually perform vfork/dup2/execve
539 res = posix_spawn(&tracker.pid, kClatdPath, &fa, &attr, (char* const*)args, nullptr);
540 if (res) {
541 ALOGE("posix_spawn failed (%s)", strerror(res));
542 return -res;
543 }
544
545 // 13. configure eBPF offload - if possible
546 maybeStartBpf(tracker);
547
548 mClatdTrackers[interface] = tracker;
549 ALOGD("clatd started on %s", interface.c_str());
550
551 *v6Str = tracker.v6Str;
552 return 0;
553 }
554
stopClatd(const std::string & interface)555 int ClatdController::stopClatd(const std::string& interface) {
556 std::lock_guard guard(mutex);
557 ClatdTracker* tracker = getClatdTracker(interface);
558
559 if (tracker == nullptr) {
560 ALOGE("clatd already stopped");
561 return -ENODEV;
562 }
563
564 ALOGD("Stopping clatd pid=%d on %s", tracker->pid, interface.c_str());
565
566 maybeStopBpf(*tracker);
567
568 kill(tracker->pid, SIGTERM);
569 waitpid(tracker->pid, nullptr, 0);
570
571 setIptablesDropRule(false, tracker->iface, tracker->pfx96String, tracker->v6Str);
572 mClatdTrackers.erase(interface);
573
574 ALOGD("clatd on %s stopped", interface.c_str());
575
576 return 0;
577 }
578
dumpEgress(DumpWriter & dw)579 void ClatdController::dumpEgress(DumpWriter& dw) {
580 if (!mClatEgress4Map.isValid()) return; // if unsupported just don't dump anything
581
582 ScopedIndent bpfIndent(dw);
583 dw.println("BPF egress map: iif(iface) v4Addr -> v6Addr nat64Prefix oif(iface)");
584
585 ScopedIndent bpfDetailIndent(dw);
586 const auto printClatMap = [&dw](const ClatEgress4Key& key, const ClatEgress4Value& value,
587 const BpfMap<ClatEgress4Key, ClatEgress4Value>&) {
588 char iifStr[IFNAMSIZ] = "?";
589 char local4Str[INET_ADDRSTRLEN] = "?";
590 char local6Str[INET6_ADDRSTRLEN] = "?";
591 char pfx96Str[INET6_ADDRSTRLEN] = "?";
592 char oifStr[IFNAMSIZ] = "?";
593
594 if_indextoname(key.iif, iifStr);
595 inet_ntop(AF_INET, &key.local4, local4Str, sizeof(local4Str));
596 inet_ntop(AF_INET6, &value.local6, local6Str, sizeof(local6Str));
597 inet_ntop(AF_INET6, &value.pfx96, pfx96Str, sizeof(pfx96Str));
598 if_indextoname(value.oif, oifStr);
599
600 dw.println("%u(%s) %s -> %s %s/96 %u(%s) %s", key.iif, iifStr, local4Str, local6Str,
601 pfx96Str, value.oif, oifStr, value.oifIsEthernet ? "ether" : "rawip");
602 return Result<void>();
603 };
604 auto res = mClatEgress4Map.iterateWithValue(printClatMap);
605 if (!res.ok()) {
606 dw.println("Error printing BPF map: %s", res.error().message().c_str());
607 }
608 }
609
dumpIngress(DumpWriter & dw)610 void ClatdController::dumpIngress(DumpWriter& dw) {
611 if (!mClatIngress6Map.isValid()) return; // if unsupported just don't dump anything
612
613 ScopedIndent bpfIndent(dw);
614 dw.println("BPF ingress map: iif(iface) nat64Prefix v6Addr -> v4Addr oif(iface)");
615
616 ScopedIndent bpfDetailIndent(dw);
617 const auto printClatMap = [&dw](const ClatIngress6Key& key, const ClatIngress6Value& value,
618 const BpfMap<ClatIngress6Key, ClatIngress6Value>&) {
619 char iifStr[IFNAMSIZ] = "?";
620 char pfx96Str[INET6_ADDRSTRLEN] = "?";
621 char local6Str[INET6_ADDRSTRLEN] = "?";
622 char local4Str[INET_ADDRSTRLEN] = "?";
623 char oifStr[IFNAMSIZ] = "?";
624
625 if_indextoname(key.iif, iifStr);
626 inet_ntop(AF_INET6, &key.pfx96, pfx96Str, sizeof(pfx96Str));
627 inet_ntop(AF_INET6, &key.local6, local6Str, sizeof(local6Str));
628 inet_ntop(AF_INET, &value.local4, local4Str, sizeof(local4Str));
629 if_indextoname(value.oif, oifStr);
630
631 dw.println("%u(%s) %s/96 %s -> %s %u(%s)", key.iif, iifStr, pfx96Str, local6Str, local4Str,
632 value.oif, oifStr);
633 return Result<void>();
634 };
635 auto res = mClatIngress6Map.iterateWithValue(printClatMap);
636 if (!res.ok()) {
637 dw.println("Error printing BPF map: %s", res.error().message().c_str());
638 }
639 }
640
dumpTrackers(DumpWriter & dw)641 void ClatdController::dumpTrackers(DumpWriter& dw) {
642 ScopedIndent trackerIndent(dw);
643 dw.println("Trackers: iif[iface] nat64Prefix v6Addr -> v4Addr v4iif[v4iface] [fwmark]");
644
645 ScopedIndent trackerDetailIndent(dw);
646 for (const auto& pair : mClatdTrackers) {
647 const ClatdTracker& tracker = pair.second;
648 dw.println("%u[%s] %s/96 %s -> %s %u[%s] [%s]", tracker.ifIndex, tracker.iface,
649 tracker.pfx96String, tracker.v6Str, tracker.v4Str, tracker.v4ifIndex,
650 tracker.v4iface, tracker.fwmarkString);
651 }
652 }
653
dump(DumpWriter & dw)654 void ClatdController::dump(DumpWriter& dw) {
655 std::lock_guard guard(mutex);
656
657 ScopedIndent clatdIndent(dw);
658 dw.println("ClatdController");
659
660 dumpTrackers(dw);
661 dumpIngress(dw);
662 dumpEgress(dw);
663 }
664
665 auto ClatdController::isIpv4AddressFreeFunc = isIpv4AddressFree;
666 auto ClatdController::iptablesRestoreFunction = execIptablesRestore;
667
668 } // namespace net
669 } // namespace android
670