2 * Copyright (C) 2014 The Android Open Source Project
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
8 * http://www.apache.org/licenses/LICENSE-2.0
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.
17 #include "RouteController.h"
20 #include "UidRanges.h"
22 #define LOG_TAG "Netd"
24 #include "logwrap/logwrap.h"
25 #include "resolv_netid.h"
27 #include <arpa/inet.h>
29 #include <linux/fib_rules.h>
36 // BEGIN CONSTANTS --------------------------------------------------------------------------------
38 const uint32_t RULE_PRIORITY_VPN_OVERRIDE_SYSTEM = 10000;
39 // const uint32_t RULE_PRIORITY_VPN_OVERRIDE_LOCAL = 11000;
40 const uint32_t RULE_PRIORITY_SECURE_VPN = 12000;
41 const uint32_t RULE_PRIORITY_EXPLICIT_NETWORK = 13000;
42 const uint32_t RULE_PRIORITY_OUTPUT_INTERFACE = 14000;
43 const uint32_t RULE_PRIORITY_LEGACY_SYSTEM = 15000;
44 const uint32_t RULE_PRIORITY_LEGACY_NETWORK = 16000;
45 const uint32_t RULE_PRIORITY_LOCAL_NETWORK = 17000;
46 const uint32_t RULE_PRIORITY_TETHERING = 18000;
47 const uint32_t RULE_PRIORITY_IMPLICIT_NETWORK = 19000;
48 // const uint32_t RULE_PRIORITY_BYPASSABLE_VPN = 20000;
49 // const uint32_t RULE_PRIORITY_VPN_FALLTHROUGH = 21000;
50 const uint32_t RULE_PRIORITY_DEFAULT_NETWORK = 22000;
51 const uint32_t RULE_PRIORITY_DIRECTLY_CONNECTED = 23000;
52 const uint32_t RULE_PRIORITY_UNREACHABLE = 24000;
54 const uint32_t ROUTE_TABLE_LOCAL_NETWORK = 97;
55 const uint32_t ROUTE_TABLE_LEGACY_NETWORK = 98;
56 const uint32_t ROUTE_TABLE_LEGACY_SYSTEM = 99;
58 const char* const ROUTE_TABLE_NAME_LOCAL_NETWORK = "local_network";
59 const char* const ROUTE_TABLE_NAME_LEGACY_NETWORK = "legacy_network";
60 const char* const ROUTE_TABLE_NAME_LEGACY_SYSTEM = "legacy_system";
62 const char* const ROUTE_TABLE_NAME_LOCAL = "local";
63 const char* const ROUTE_TABLE_NAME_MAIN = "main";
65 // TODO: These values aren't defined by the Linux kernel, because our UID routing changes are not
66 // upstream (yet?), so we can't just pick them up from kernel headers. When (if?) the changes make
67 // it upstream, we'll remove this and rely on the kernel header values. For now, add a static assert
68 // that will warn us if upstream has given these values some other meaning.
69 const uint16_t FRA_UID_START = 18;
70 const uint16_t FRA_UID_END = 19;
71 static_assert(FRA_UID_START > FRA_MAX,
72 "Android-specific FRA_UID_{START,END} values also assigned in Linux uapi. "
73 "Check that these values match what the kernel does and then update this assertion.");
75 const uint16_t NETLINK_REQUEST_FLAGS = NLM_F_REQUEST | NLM_F_ACK;
76 const uint16_t NETLINK_CREATE_REQUEST_FLAGS = NETLINK_REQUEST_FLAGS | NLM_F_CREATE | NLM_F_EXCL;
78 const sockaddr_nl NETLINK_ADDRESS = {AF_NETLINK, 0, 0, 0};
80 const uint8_t AF_FAMILIES[] = {AF_INET, AF_INET6};
82 const char* const IP_VERSIONS[] = {"-4", "-6"};
84 const uid_t UID_ROOT = 0;
85 const char* const IIF_NONE = NULL;
86 const char* const OIF_NONE = NULL;
87 const bool ACTION_ADD = true;
88 const bool ACTION_DEL = false;
89 const bool MODIFY_NON_UID_BASED_RULES = true;
91 const char* const RT_TABLES_PATH = "/data/misc/net/rt_tables";
92 const int RT_TABLES_FLAGS = O_CREAT | O_TRUNC | O_WRONLY | O_NOFOLLOW | O_CLOEXEC;
93 const mode_t RT_TABLES_MODE = S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH; // mode 0644, rw-r--r--
95 // Avoids "non-constant-expression cannot be narrowed from type 'unsigned int' to 'unsigned short'"
96 // warnings when using RTA_LENGTH(x) inside static initializers (even when x is already uint16_t).
97 constexpr uint16_t U16_RTA_LENGTH(uint16_t x) {
101 // These are practically const, but can't be declared so, because they are used to initialize
102 // non-const pointers ("void* iov_base") in iovec arrays.
103 rtattr FRATTR_PRIORITY = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_PRIORITY };
104 rtattr FRATTR_TABLE = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_TABLE };
105 rtattr FRATTR_FWMARK = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_FWMARK };
106 rtattr FRATTR_FWMASK = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_FWMASK };
107 rtattr FRATTR_UID_START = { U16_RTA_LENGTH(sizeof(uid_t)), FRA_UID_START };
108 rtattr FRATTR_UID_END = { U16_RTA_LENGTH(sizeof(uid_t)), FRA_UID_END };
110 rtattr RTATTR_TABLE = { U16_RTA_LENGTH(sizeof(uint32_t)), RTA_TABLE };
111 rtattr RTATTR_OIF = { U16_RTA_LENGTH(sizeof(uint32_t)), RTA_OIF };
113 uint8_t PADDING_BUFFER[RTA_ALIGNTO] = {0, 0, 0, 0};
115 // END CONSTANTS ----------------------------------------------------------------------------------
117 // No locks needed because RouteController is accessed only from one thread (in CommandListener).
118 std::map<std::string, uint32_t> interfaceToTable;
120 uint32_t getRouteTableForInterface(const char* interface) {
121 uint32_t index = if_nametoindex(interface);
123 index += RouteController::ROUTE_TABLE_OFFSET_FROM_INDEX;
124 interfaceToTable[interface] = index;
127 // If the interface goes away if_nametoindex() will return 0 but we still need to know
128 // the index so we can remove the rules and routes.
129 auto iter = interfaceToTable.find(interface);
130 if (iter == interfaceToTable.end()) {
131 ALOGE("cannot find interface %s", interface);
132 return RT_TABLE_UNSPEC;
137 void addTableName(uint32_t table, const std::string& name, std::string* contents) {
138 char tableString[UINT32_STRLEN];
139 snprintf(tableString, sizeof(tableString), "%u", table);
140 *contents += tableString;
146 // Doesn't return success/failure as the file is optional; it's okay if we fail to update it.
147 void updateTableNamesFile() {
148 std::string contents;
150 addTableName(RT_TABLE_LOCAL, ROUTE_TABLE_NAME_LOCAL, &contents);
151 addTableName(RT_TABLE_MAIN, ROUTE_TABLE_NAME_MAIN, &contents);
153 addTableName(ROUTE_TABLE_LOCAL_NETWORK, ROUTE_TABLE_NAME_LOCAL_NETWORK, &contents);
154 addTableName(ROUTE_TABLE_LEGACY_NETWORK, ROUTE_TABLE_NAME_LEGACY_NETWORK, &contents);
155 addTableName(ROUTE_TABLE_LEGACY_SYSTEM, ROUTE_TABLE_NAME_LEGACY_SYSTEM, &contents);
157 for (const auto& entry : interfaceToTable) {
158 addTableName(entry.second, entry.first, &contents);
161 int fd = open(RT_TABLES_PATH, RT_TABLES_FLAGS, RT_TABLES_MODE);
163 ALOGE("failed to create %s (%s)", RT_TABLES_PATH, strerror(errno));
166 // File creation is affected by umask, so make sure the right mode bits are set.
167 if (fchmod(fd, RT_TABLES_MODE) == -1) {
168 ALOGE("failed to set mode 0%o on %s (%s)", RT_TABLES_MODE, RT_TABLES_PATH, strerror(errno));
170 ssize_t bytesWritten = write(fd, contents.data(), contents.size());
171 if (bytesWritten != static_cast<ssize_t>(contents.size())) {
172 ALOGE("failed to write to %s (%zd vs %zu bytes) (%s)", RT_TABLES_PATH, bytesWritten,
173 contents.size(), strerror(errno));
178 // Sends a netlink request and expects an ack.
179 // |iov| is an array of struct iovec that contains the netlink message payload.
180 // The netlink header is generated by this function based on |action| and |flags|.
181 // Returns -errno if there was an error or if the kernel reported an error.
182 WARN_UNUSED_RESULT int sendNetlinkRequest(uint16_t action, uint16_t flags, iovec* iov, int iovlen) {
184 .nlmsg_type = action,
185 .nlmsg_flags = flags,
187 iov[0].iov_base = &nlmsg;
188 iov[0].iov_len = sizeof(nlmsg);
189 for (int i = 0; i < iovlen; ++i) {
190 nlmsg.nlmsg_len += iov[i].iov_len;
199 int sock = socket(AF_NETLINK, SOCK_DGRAM, NETLINK_ROUTE);
201 connect(sock, reinterpret_cast<const sockaddr*>(&NETLINK_ADDRESS),
202 sizeof(NETLINK_ADDRESS)) != -1 &&
203 writev(sock, iov, iovlen) != -1 &&
204 (ret = recv(sock, &response, sizeof(response), 0)) != -1) {
205 if (ret == sizeof(response)) {
206 ret = response.err.error; // Netlink errors are negative errno.
208 ALOGE("netlink response contains error (%s)", strerror(-ret));
211 ALOGE("bad netlink response message size (%d != %zu)", ret, sizeof(response));
215 ALOGE("netlink socket/connect/writev/recv failed (%s)", strerror(errno));
226 // Returns 0 on success or negative errno on failure.
227 int padInterfaceName(const char* input, char* name, size_t* length, uint16_t* padding) {
233 *length = strlcpy(name, input, IFNAMSIZ) + 1;
234 if (*length > IFNAMSIZ) {
235 ALOGE("interface name too long (%zu > %u)", *length, IFNAMSIZ);
236 return -ENAMETOOLONG;
238 *padding = RTA_SPACE(*length) - RTA_LENGTH(*length);
242 // Adds or removes a routing rule for IPv4 and IPv6.
244 // + If |table| is non-zero, the rule points at the specified routing table. Otherwise, the rule
245 // returns ENETUNREACH.
246 // + If |mask| is non-zero, the rule matches the specified fwmark and mask. Otherwise, |fwmark| is
248 // + If |iif| is non-NULL, the rule matches the specified incoming interface.
249 // + If |oif| is non-NULL, the rule matches the specified outgoing interface.
250 // + If |uidStart| and |uidEnd| are not INVALID_UID, the rule matches packets from UIDs in that
251 // range (inclusive). Otherwise, the rule matches packets from all UIDs.
253 // Returns 0 on success or negative errno on failure.
254 WARN_UNUSED_RESULT int modifyIpRule(uint16_t action, uint32_t priority, uint32_t table,
255 uint32_t fwmark, uint32_t mask, const char* iif,
256 const char* oif, uid_t uidStart, uid_t uidEnd) {
257 // Ensure that if you set a bit in the fwmark, it's not being ignored by the mask.
258 if (fwmark & ~mask) {
259 ALOGE("mask 0x%x does not select all the bits set in fwmark 0x%x", mask, fwmark);
263 // Interface names must include exactly one terminating NULL and be properly padded, or older
264 // kernels will refuse to delete rules.
265 char iifName[IFNAMSIZ], oifName[IFNAMSIZ];
266 size_t iifLength, oifLength;
267 uint16_t iifPadding, oifPadding;
268 if (int ret = padInterfaceName(iif, iifName, &iifLength, &iifPadding)) {
271 if (int ret = padInterfaceName(oif, oifName, &oifLength, &oifPadding)) {
275 // Either both start and end UID must be specified, or neither.
276 if ((uidStart == INVALID_UID) != (uidEnd == INVALID_UID)) {
277 ALOGE("incompatible start and end UIDs (%u vs %u)", uidStart, uidEnd);
280 bool isUidRule = (uidStart != INVALID_UID);
282 // Assemble a rule request and put it in an array of iovec structures.
283 fib_rule_hdr rule = {
284 .action = static_cast<uint8_t>(table != RT_TABLE_UNSPEC ? FR_ACT_TO_TBL :
288 rtattr fraIifName = { U16_RTA_LENGTH(iifLength), FRA_IIFNAME };
289 rtattr fraOifName = { U16_RTA_LENGTH(oifLength), FRA_OIFNAME };
293 { &rule, sizeof(rule) },
294 { &FRATTR_PRIORITY, sizeof(FRATTR_PRIORITY) },
295 { &priority, sizeof(priority) },
296 { &FRATTR_TABLE, table != RT_TABLE_UNSPEC ? sizeof(FRATTR_TABLE) : 0 },
297 { &table, table != RT_TABLE_UNSPEC ? sizeof(table) : 0 },
298 { &FRATTR_FWMARK, mask ? sizeof(FRATTR_FWMARK) : 0 },
299 { &fwmark, mask ? sizeof(fwmark) : 0 },
300 { &FRATTR_FWMASK, mask ? sizeof(FRATTR_FWMASK) : 0 },
301 { &mask, mask ? sizeof(mask) : 0 },
302 { &FRATTR_UID_START, isUidRule ? sizeof(FRATTR_UID_START) : 0 },
303 { &uidStart, isUidRule ? sizeof(uidStart) : 0 },
304 { &FRATTR_UID_END, isUidRule ? sizeof(FRATTR_UID_END) : 0 },
305 { &uidEnd, isUidRule ? sizeof(uidEnd) : 0 },
306 { &fraIifName, iif != IIF_NONE ? sizeof(fraIifName) : 0 },
307 { iifName, iifLength },
308 { PADDING_BUFFER, iifPadding },
309 { &fraOifName, oif != OIF_NONE ? sizeof(fraOifName) : 0 },
310 { oifName, oifLength },
311 { PADDING_BUFFER, oifPadding },
314 uint16_t flags = (action == RTM_NEWRULE) ? NETLINK_CREATE_REQUEST_FLAGS : NETLINK_REQUEST_FLAGS;
315 for (size_t i = 0; i < ARRAY_SIZE(AF_FAMILIES); ++i) {
316 rule.family = AF_FAMILIES[i];
317 if (int ret = sendNetlinkRequest(action, flags, iov, ARRAY_SIZE(iov))) {
325 WARN_UNUSED_RESULT int modifyIpRule(uint16_t action, uint32_t priority, uint32_t table,
326 uint32_t fwmark, uint32_t mask) {
327 return modifyIpRule(action, priority, table, fwmark, mask, IIF_NONE, OIF_NONE, INVALID_UID,
331 // Adds or deletes an IPv4 or IPv6 route.
332 // Returns 0 on success or negative errno on failure.
333 WARN_UNUSED_RESULT int modifyIpRoute(uint16_t action, uint32_t table, const char* interface,
334 const char* destination, const char* nexthop) {
335 // At least the destination must be non-null.
337 ALOGE("null destination");
342 uint8_t rawAddress[sizeof(in6_addr)];
344 uint8_t prefixLength;
345 int rawLength = parsePrefix(destination, &family, rawAddress, sizeof(rawAddress),
348 ALOGE("parsePrefix failed for destination %s (%s)", destination, strerror(-rawLength));
352 if (static_cast<size_t>(rawLength) > sizeof(rawAddress)) {
353 ALOGE("impossible! address too long (%d vs %zu)", rawLength, sizeof(rawAddress));
354 return -ENOBUFS; // Cannot happen; parsePrefix only supports IPv4 and IPv6.
357 // If an interface was specified, find the ifindex.
359 if (interface != OIF_NONE) {
360 ifindex = if_nametoindex(interface);
362 ALOGE("cannot find interface %s", interface);
367 // If a nexthop was specified, parse it as the same family as the prefix.
368 uint8_t rawNexthop[sizeof(in6_addr)];
369 if (nexthop && inet_pton(family, nexthop, rawNexthop) <= 0) {
370 ALOGE("inet_pton failed for nexthop %s", nexthop);
374 // Assemble a rtmsg and put it in an array of iovec structures.
376 .rtm_protocol = RTPROT_STATIC,
377 .rtm_type = RTN_UNICAST,
378 .rtm_family = family,
379 .rtm_dst_len = prefixLength,
382 rtattr rtaDst = { U16_RTA_LENGTH(rawLength), RTA_DST };
383 rtattr rtaGateway = { U16_RTA_LENGTH(rawLength), RTA_GATEWAY };
387 { &route, sizeof(route) },
388 { &RTATTR_TABLE, sizeof(RTATTR_TABLE) },
389 { &table, sizeof(table) },
390 { &rtaDst, sizeof(rtaDst) },
391 { rawAddress, static_cast<size_t>(rawLength) },
392 { &RTATTR_OIF, interface != OIF_NONE ? sizeof(RTATTR_OIF) : 0 },
393 { &ifindex, interface != OIF_NONE ? sizeof(ifindex) : 0 },
394 { &rtaGateway, nexthop ? sizeof(rtaGateway) : 0 },
395 { rawNexthop, nexthop ? static_cast<size_t>(rawLength) : 0 },
398 uint16_t flags = (action == RTM_NEWROUTE) ? NETLINK_CREATE_REQUEST_FLAGS :
399 NETLINK_REQUEST_FLAGS;
400 return sendNetlinkRequest(action, flags, iov, ARRAY_SIZE(iov));
403 // Add rules to allow legacy routes added through the requestRouteToHost() API.
404 WARN_UNUSED_RESULT int addLegacyRouteRules() {
408 fwmark.explicitlySelected = false;
409 mask.explicitlySelected = true;
411 // Rules to allow legacy routes to override the default network.
412 if (int ret = modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_LEGACY_SYSTEM, ROUTE_TABLE_LEGACY_SYSTEM,
413 fwmark.intValue, mask.intValue)) {
416 if (int ret = modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_LEGACY_NETWORK,
417 ROUTE_TABLE_LEGACY_NETWORK, fwmark.intValue, mask.intValue)) {
421 fwmark.permission = PERMISSION_SYSTEM;
422 mask.permission = PERMISSION_SYSTEM;
424 // A rule to allow legacy routes from system apps to override VPNs.
425 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_VPN_OVERRIDE_SYSTEM, ROUTE_TABLE_LEGACY_SYSTEM,
426 fwmark.intValue, mask.intValue);
429 // Add a new rule to look up the 'main' table, with the same selectors as the "default network"
430 // rule, but with a lower priority. Since the default network rule points to a table with a default
431 // route, the rule we're adding will never be used for normal routing lookups. However, the kernel
432 // may fall-through to it to find directly-connected routes when it validates that a nexthop (in a
433 // route being added) is reachable.
434 WARN_UNUSED_RESULT int addDirectlyConnectedRule() {
438 fwmark.netId = NETID_UNSET;
439 mask.netId = FWMARK_NET_ID_MASK;
441 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_DIRECTLY_CONNECTED, RT_TABLE_MAIN,
442 fwmark.intValue, mask.intValue, IIF_NONE, OIF_NONE, UID_ROOT, UID_ROOT);
445 // Add a rule to preempt the pre-defined "from all lookup main" rule. Packets that reach this rule
446 // will be null-routed, and won't fall-through to the main table.
447 WARN_UNUSED_RESULT int addUnreachableRule() {
448 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_UNREACHABLE, RT_TABLE_UNSPEC, MARK_UNSET,
452 // A rule to lookup the local network before looking up the default network.
453 WARN_UNUSED_RESULT int addImplicitLocalNetworkRule() {
457 fwmark.explicitlySelected = false;
458 mask.explicitlySelected = true;
460 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_LOCAL_NETWORK, ROUTE_TABLE_LOCAL_NETWORK,
461 fwmark.intValue, mask.intValue);
464 // An iptables rule to mark incoming packets on a network with the netId of the network.
466 // This is so that the kernel can:
467 // + Use the right fwmark for (and thus correctly route) replies (e.g.: TCP RST, ICMP errors, ping
468 // replies, SYN-ACKs, etc).
469 // + Mark sockets that accept connections from this interface so that the connection stays on the
471 WARN_UNUSED_RESULT int modifyIncomingPacketMark(unsigned netId, const char* interface,
472 Permission permission, bool add) {
475 fwmark.netId = netId;
476 fwmark.explicitlySelected = true;
477 fwmark.protectedFromVpn = true;
478 fwmark.permission = permission;
480 char markString[UINT32_HEX_STRLEN];
481 snprintf(markString, sizeof(markString), "0x%x", fwmark.intValue);
483 if (execIptables(V4V6, "-t", "mangle", add ? "-A" : "-D", "INPUT", "-i", interface, "-j",
484 "MARK", "--set-mark", markString, NULL)) {
485 ALOGE("failed to change iptables rule that sets incoming packet mark");
492 // A rule to route traffic based on an explicitly chosen network.
494 // Supports apps that use the multinetwork APIs to restrict their traffic to a network.
496 // Even though we check permissions at the time we set a netId into the fwmark of a socket, we need
497 // to check it again in the rules here, because a network's permissions may have been updated via
498 // modifyNetworkPermission().
499 WARN_UNUSED_RESULT int modifyExplicitNetworkRule(unsigned netId, uint32_t table,
500 Permission permission, uid_t uidStart,
501 uid_t uidEnd, bool add) {
505 fwmark.netId = netId;
506 mask.netId = FWMARK_NET_ID_MASK;
508 fwmark.explicitlySelected = true;
509 mask.explicitlySelected = true;
511 fwmark.permission = permission;
512 mask.permission = permission;
514 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_EXPLICIT_NETWORK, table,
515 fwmark.intValue, mask.intValue, IIF_NONE, OIF_NONE, uidStart, uidEnd);
518 // A rule to route traffic based on a chosen outgoing interface.
520 // Supports apps that use SO_BINDTODEVICE or IP_PKTINFO options and the kernel that already knows
521 // the outgoing interface (typically for link-local communications).
522 WARN_UNUSED_RESULT int modifyOutputInterfaceRule(const char* interface, uint32_t table,
523 Permission permission, uid_t uidStart,
524 uid_t uidEnd, bool add) {
528 fwmark.permission = permission;
529 mask.permission = permission;
531 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_OUTPUT_INTERFACE, table,
532 fwmark.intValue, mask.intValue, IIF_NONE, interface, uidStart, uidEnd);
535 // A rule to route traffic based on the chosen network.
537 // This is for sockets that have not explicitly requested a particular network, but have been
538 // bound to one when they called connect(). This ensures that sockets connected on a particular
539 // network stay on that network even if the default network changes.
540 WARN_UNUSED_RESULT int modifyImplicitNetworkRule(unsigned netId, uint32_t table,
541 Permission permission, bool add) {
545 fwmark.netId = netId;
546 mask.netId = FWMARK_NET_ID_MASK;
548 fwmark.explicitlySelected = false;
549 mask.explicitlySelected = true;
551 fwmark.permission = permission;
552 mask.permission = permission;
554 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_IMPLICIT_NETWORK, table,
555 fwmark.intValue, mask.intValue);
558 // A rule to route all traffic from a given set of UIDs to go over the VPN.
560 // Notice that this rule doesn't use the netId. I.e., no matter what netId the user's socket may
561 // have, if they are subject to this VPN, their traffic has to go through it. Allows the traffic to
562 // bypass the VPN if the protectedFromVpn bit is set.
563 WARN_UNUSED_RESULT int modifyVpnUidRangeRule(uint32_t table, uid_t uidStart, uid_t uidEnd,
568 fwmark.protectedFromVpn = false;
569 mask.protectedFromVpn = true;
571 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_SECURE_VPN, table,
572 fwmark.intValue, mask.intValue, IIF_NONE, OIF_NONE, uidStart, uidEnd);
575 // A rule to allow system apps to send traffic over this VPN even if they are not part of the target
578 // This is needed for DnsProxyListener to correctly resolve a request for a user who is in the
579 // target set, but where the DnsProxyListener itself is not.
580 WARN_UNUSED_RESULT int modifyVpnSystemPermissionRule(unsigned netId, uint32_t table, bool add) {
584 fwmark.netId = netId;
585 mask.netId = FWMARK_NET_ID_MASK;
587 fwmark.permission = PERMISSION_SYSTEM;
588 mask.permission = PERMISSION_SYSTEM;
590 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_SECURE_VPN, table,
591 fwmark.intValue, mask.intValue);
594 WARN_UNUSED_RESULT int modifyTetheringRule(uint16_t action, const char* inputInterface,
595 const char* outputInterface) {
596 uint32_t table = getRouteTableForInterface(outputInterface);
597 if (table == RT_TABLE_UNSPEC) {
601 return modifyIpRule(action, RULE_PRIORITY_TETHERING, table, MARK_UNSET, MARK_UNSET,
602 inputInterface, outputInterface, INVALID_UID, INVALID_UID);
605 WARN_UNUSED_RESULT int modifyLocalNetwork(unsigned netId, const char* interface, bool add) {
606 if (int ret = modifyIncomingPacketMark(netId, interface, PERMISSION_NONE, add)) {
609 return modifyOutputInterfaceRule(interface, ROUTE_TABLE_LOCAL_NETWORK, PERMISSION_NONE,
610 INVALID_UID, INVALID_UID, add);
613 WARN_UNUSED_RESULT int modifyPhysicalNetwork(unsigned netId, const char* interface,
614 Permission permission, bool add) {
615 uint32_t table = getRouteTableForInterface(interface);
616 if (table == RT_TABLE_UNSPEC) {
620 if (int ret = modifyIncomingPacketMark(netId, interface, permission, add)) {
623 if (int ret = modifyExplicitNetworkRule(netId, table, permission, INVALID_UID, INVALID_UID,
627 if (int ret = modifyOutputInterfaceRule(interface, table, permission, INVALID_UID, INVALID_UID,
631 return modifyImplicitNetworkRule(netId, table, permission, add);
634 WARN_UNUSED_RESULT int modifyVirtualNetwork(unsigned netId, const char* interface,
635 const UidRanges& uidRanges, bool add,
636 bool modifyNonUidBasedRules) {
637 uint32_t table = getRouteTableForInterface(interface);
638 if (table == RT_TABLE_UNSPEC) {
642 for (const UidRanges::Range& range : uidRanges.getRanges()) {
643 if (int ret = modifyExplicitNetworkRule(netId, table, PERMISSION_NONE, range.first,
644 range.second, add)) {
647 if (int ret = modifyOutputInterfaceRule(interface, table, PERMISSION_NONE, range.first,
648 range.second, add)) {
651 if (int ret = modifyVpnUidRangeRule(table, range.first, range.second, add)) {
656 if (modifyNonUidBasedRules) {
657 if (int ret = modifyIncomingPacketMark(netId, interface, PERMISSION_NONE, add)) {
660 if (int ret = modifyExplicitNetworkRule(netId, table, PERMISSION_NONE, UID_ROOT, UID_ROOT,
664 return modifyVpnSystemPermissionRule(netId, table, add);
670 WARN_UNUSED_RESULT int modifyDefaultNetwork(uint16_t action, const char* interface,
671 Permission permission) {
672 uint32_t table = getRouteTableForInterface(interface);
673 if (table == RT_TABLE_UNSPEC) {
680 fwmark.netId = NETID_UNSET;
681 mask.netId = FWMARK_NET_ID_MASK;
683 fwmark.permission = permission;
684 mask.permission = permission;
686 return modifyIpRule(action, RULE_PRIORITY_DEFAULT_NETWORK, table, fwmark.intValue,
690 // Returns 0 on success or negative errno on failure.
691 WARN_UNUSED_RESULT int flushRules() {
692 for (size_t i = 0; i < ARRAY_SIZE(IP_VERSIONS); ++i) {
693 const char* argv[] = {
699 if (android_fork_execvp(ARRAY_SIZE(argv), const_cast<char**>(argv), NULL, false, false)) {
700 ALOGE("failed to flush rules");
707 // Adds or removes an IPv4 or IPv6 route to the specified table and, if it's a directly-connected
708 // route, to the main table as well.
709 // Returns 0 on success or negative errno on failure.
710 WARN_UNUSED_RESULT int modifyRoute(uint16_t action, const char* interface, const char* destination,
711 const char* nexthop, RouteController::TableType tableType) {
714 case RouteController::INTERFACE: {
715 table = getRouteTableForInterface(interface);
716 if (table == RT_TABLE_UNSPEC) {
721 case RouteController::LOCAL_NETWORK: {
722 table = ROUTE_TABLE_LOCAL_NETWORK;
725 case RouteController::LEGACY_NETWORK: {
726 table = ROUTE_TABLE_LEGACY_NETWORK;
729 case RouteController::LEGACY_SYSTEM: {
730 table = ROUTE_TABLE_LEGACY_SYSTEM;
735 int ret = modifyIpRoute(action, table, interface, destination, nexthop);
736 // We allow apps to call requestRouteToHost() multiple times with the same route, so ignore
737 // EEXIST failures when adding routes to legacy tables.
738 if (ret && !(action == RTM_NEWROUTE && ret == -EEXIST &&
739 (tableType == RouteController::LEGACY_NETWORK ||
740 tableType == RouteController::LEGACY_SYSTEM))) {
744 // If there's no nexthop, this is a directly connected route. Add it to the main table also, to
745 // let the kernel find it when validating nexthops when global routes are added.
747 ret = modifyIpRoute(action, RT_TABLE_MAIN, interface, destination, NULL);
748 // A failure with action == ADD && errno == EEXIST means that the route already exists in
749 // the main table, perhaps because the kernel added it automatically as part of adding the
750 // IP address to the interface. Ignore this, but complain about everything else.
751 if (ret && !(action == RTM_NEWROUTE && ret == -EEXIST)) {
759 // Returns 0 on success or negative errno on failure.
760 WARN_UNUSED_RESULT int flushRoutes(const char* interface) {
761 uint32_t table = getRouteTableForInterface(interface);
762 if (table == RT_TABLE_UNSPEC) {
766 char tableString[UINT32_STRLEN];
767 snprintf(tableString, sizeof(tableString), "%u", table);
769 for (size_t i = 0; i < ARRAY_SIZE(IP_VERSIONS); ++i) {
770 const char* argv[] = {
778 if (android_fork_execvp(ARRAY_SIZE(argv), const_cast<char**>(argv), NULL, false, false)) {
779 ALOGE("failed to flush routes");
784 interfaceToTable.erase(interface);
790 int RouteController::Init(unsigned localNetId) {
791 if (int ret = flushRules()) {
795 if (int ret = addDirectlyConnectedRule()) {
798 if (int ret = addLegacyRouteRules()) {
801 if (int ret = addUnreachableRule()) {
804 if (int ret = addImplicitLocalNetworkRule()) {
807 // Add a rule to lookup the local network if it has been explicitly selected.
808 if (int ret = modifyExplicitNetworkRule(localNetId, ROUTE_TABLE_LOCAL_NETWORK, PERMISSION_NONE,
809 INVALID_UID, INVALID_UID, ACTION_ADD)) {
812 updateTableNamesFile();
816 int RouteController::addInterfaceToLocalNetwork(unsigned netId, const char* interface) {
817 return modifyLocalNetwork(netId, interface, ACTION_ADD);
820 int RouteController::removeInterfaceFromLocalNetwork(unsigned netId, const char* interface) {
821 return modifyLocalNetwork(netId, interface, ACTION_DEL);
824 int RouteController::addInterfaceToPhysicalNetwork(unsigned netId, const char* interface,
825 Permission permission) {
826 if (int ret = modifyPhysicalNetwork(netId, interface, permission, ACTION_ADD)) {
829 updateTableNamesFile();
833 int RouteController::removeInterfaceFromPhysicalNetwork(unsigned netId, const char* interface,
834 Permission permission) {
835 if (int ret = modifyPhysicalNetwork(netId, interface, permission, ACTION_DEL)) {
838 if (int ret = flushRoutes(interface)) {
841 updateTableNamesFile();
845 int RouteController::addInterfaceToVirtualNetwork(unsigned netId, const char* interface,
846 const UidRanges& uidRanges) {
847 if (int ret = modifyVirtualNetwork(netId, interface, uidRanges, ACTION_ADD,
848 MODIFY_NON_UID_BASED_RULES)) {
851 updateTableNamesFile();
855 int RouteController::removeInterfaceFromVirtualNetwork(unsigned netId, const char* interface,
856 const UidRanges& uidRanges) {
857 if (int ret = modifyVirtualNetwork(netId, interface, uidRanges, ACTION_DEL,
858 MODIFY_NON_UID_BASED_RULES)) {
861 if (int ret = flushRoutes(interface)) {
864 updateTableNamesFile();
868 int RouteController::modifyPhysicalNetworkPermission(unsigned netId, const char* interface,
869 Permission oldPermission,
870 Permission newPermission) {
871 // Add the new rules before deleting the old ones, to avoid race conditions.
872 if (int ret = modifyPhysicalNetwork(netId, interface, newPermission, ACTION_ADD)) {
875 return modifyPhysicalNetwork(netId, interface, oldPermission, ACTION_DEL);
878 int RouteController::addUsersToVirtualNetwork(unsigned netId, const char* interface,
879 const UidRanges& uidRanges) {
880 return modifyVirtualNetwork(netId, interface, uidRanges, ACTION_ADD,
881 !MODIFY_NON_UID_BASED_RULES);
884 int RouteController::removeUsersFromVirtualNetwork(unsigned netId, const char* interface,
885 const UidRanges& uidRanges) {
886 return modifyVirtualNetwork(netId, interface, uidRanges, ACTION_DEL,
887 !MODIFY_NON_UID_BASED_RULES);
890 int RouteController::addInterfaceToDefaultNetwork(const char* interface, Permission permission) {
891 return modifyDefaultNetwork(RTM_NEWRULE, interface, permission);
894 int RouteController::removeInterfaceFromDefaultNetwork(const char* interface,
895 Permission permission) {
896 return modifyDefaultNetwork(RTM_DELRULE, interface, permission);
899 int RouteController::addRoute(const char* interface, const char* destination, const char* nexthop,
900 TableType tableType) {
901 return modifyRoute(RTM_NEWROUTE, interface, destination, nexthop, tableType);
904 int RouteController::removeRoute(const char* interface, const char* destination,
905 const char* nexthop, TableType tableType) {
906 return modifyRoute(RTM_DELROUTE, interface, destination, nexthop, tableType);
909 int RouteController::enableTethering(const char* inputInterface, const char* outputInterface) {
910 return modifyTetheringRule(RTM_NEWRULE, inputInterface, outputInterface);
913 int RouteController::disableTethering(const char* inputInterface, const char* outputInterface) {
914 return modifyTetheringRule(RTM_DELRULE, inputInterface, outputInterface);