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"
19 #include <arpa/inet.h>
22 #include <linux/fib_rules.h>
26 #include <private/android_filesystem_config.h>
31 #include "UidRanges.h"
33 #include "base/file.h"
34 #define LOG_TAG "Netd"
36 #include "logwrap/logwrap.h"
37 #include "resolv_netid.h"
39 using android::base::WriteStringToFile;
43 // BEGIN CONSTANTS --------------------------------------------------------------------------------
45 const uint32_t RULE_PRIORITY_VPN_OVERRIDE_SYSTEM = 10000;
46 const uint32_t RULE_PRIORITY_VPN_OUTPUT_TO_LOCAL = 11000;
47 const uint32_t RULE_PRIORITY_SECURE_VPN = 12000;
48 const uint32_t RULE_PRIORITY_EXPLICIT_NETWORK = 13000;
49 const uint32_t RULE_PRIORITY_OUTPUT_INTERFACE = 14000;
50 const uint32_t RULE_PRIORITY_LEGACY_SYSTEM = 15000;
51 const uint32_t RULE_PRIORITY_LEGACY_NETWORK = 16000;
52 const uint32_t RULE_PRIORITY_LOCAL_NETWORK = 17000;
53 const uint32_t RULE_PRIORITY_TETHERING = 18000;
54 const uint32_t RULE_PRIORITY_IMPLICIT_NETWORK = 19000;
55 const uint32_t RULE_PRIORITY_BYPASSABLE_VPN = 20000;
56 const uint32_t RULE_PRIORITY_VPN_FALLTHROUGH = 21000;
57 const uint32_t RULE_PRIORITY_DEFAULT_NETWORK = 22000;
58 const uint32_t RULE_PRIORITY_DIRECTLY_CONNECTED = 23000;
59 const uint32_t RULE_PRIORITY_UNREACHABLE = 32000;
61 const uint32_t ROUTE_TABLE_LOCAL_NETWORK = 97;
62 const uint32_t ROUTE_TABLE_LEGACY_NETWORK = 98;
63 const uint32_t ROUTE_TABLE_LEGACY_SYSTEM = 99;
65 const char* const ROUTE_TABLE_NAME_LOCAL_NETWORK = "local_network";
66 const char* const ROUTE_TABLE_NAME_LEGACY_NETWORK = "legacy_network";
67 const char* const ROUTE_TABLE_NAME_LEGACY_SYSTEM = "legacy_system";
69 const char* const ROUTE_TABLE_NAME_LOCAL = "local";
70 const char* const ROUTE_TABLE_NAME_MAIN = "main";
72 // TODO: These values aren't defined by the Linux kernel, because our UID routing changes are not
73 // upstream (yet?), so we can't just pick them up from kernel headers. When (if?) the changes make
74 // it upstream, we'll remove this and rely on the kernel header values. For now, add a static assert
75 // that will warn us if upstream has given these values some other meaning.
76 const uint16_t FRA_UID_START = 18;
77 const uint16_t FRA_UID_END = 19;
78 static_assert(FRA_UID_START > FRA_MAX,
79 "Android-specific FRA_UID_{START,END} values also assigned in Linux uapi. "
80 "Check that these values match what the kernel does and then update this assertion.");
82 const uint16_t NETLINK_REQUEST_FLAGS = NLM_F_REQUEST | NLM_F_ACK;
83 const uint16_t NETLINK_CREATE_REQUEST_FLAGS = NETLINK_REQUEST_FLAGS | NLM_F_CREATE | NLM_F_EXCL;
85 const sockaddr_nl NETLINK_ADDRESS = {AF_NETLINK, 0, 0, 0};
87 const uint8_t AF_FAMILIES[] = {AF_INET, AF_INET6};
89 const char* const IP_VERSIONS[] = {"-4", "-6"};
91 const uid_t UID_ROOT = 0;
92 const char* const IIF_NONE = NULL;
93 const char* const OIF_NONE = NULL;
94 const bool ACTION_ADD = true;
95 const bool ACTION_DEL = false;
96 const bool MODIFY_NON_UID_BASED_RULES = true;
98 const char* const RT_TABLES_PATH = "/data/misc/net/rt_tables";
99 const mode_t RT_TABLES_MODE = S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH; // mode 0644, rw-r--r--
101 const unsigned ROUTE_FLUSH_ATTEMPTS = 2;
103 // Avoids "non-constant-expression cannot be narrowed from type 'unsigned int' to 'unsigned short'"
104 // warnings when using RTA_LENGTH(x) inside static initializers (even when x is already uint16_t).
105 constexpr uint16_t U16_RTA_LENGTH(uint16_t x) {
106 return RTA_LENGTH(x);
109 // These are practically const, but can't be declared so, because they are used to initialize
110 // non-const pointers ("void* iov_base") in iovec arrays.
111 rtattr FRATTR_PRIORITY = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_PRIORITY };
112 rtattr FRATTR_TABLE = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_TABLE };
113 rtattr FRATTR_FWMARK = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_FWMARK };
114 rtattr FRATTR_FWMASK = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_FWMASK };
115 rtattr FRATTR_UID_START = { U16_RTA_LENGTH(sizeof(uid_t)), FRA_UID_START };
116 rtattr FRATTR_UID_END = { U16_RTA_LENGTH(sizeof(uid_t)), FRA_UID_END };
118 rtattr RTATTR_TABLE = { U16_RTA_LENGTH(sizeof(uint32_t)), RTA_TABLE };
119 rtattr RTATTR_OIF = { U16_RTA_LENGTH(sizeof(uint32_t)), RTA_OIF };
121 uint8_t PADDING_BUFFER[RTA_ALIGNTO] = {0, 0, 0, 0};
123 // END CONSTANTS ----------------------------------------------------------------------------------
125 // No locks needed because RouteController is accessed only from one thread (in CommandListener).
126 std::map<std::string, uint32_t> interfaceToTable;
128 uint32_t getRouteTableForInterface(const char* interface) {
129 uint32_t index = if_nametoindex(interface);
131 index += RouteController::ROUTE_TABLE_OFFSET_FROM_INDEX;
132 interfaceToTable[interface] = index;
135 // If the interface goes away if_nametoindex() will return 0 but we still need to know
136 // the index so we can remove the rules and routes.
137 auto iter = interfaceToTable.find(interface);
138 if (iter == interfaceToTable.end()) {
139 ALOGE("cannot find interface %s", interface);
140 return RT_TABLE_UNSPEC;
145 void addTableName(uint32_t table, const std::string& name, std::string* contents) {
146 char tableString[UINT32_STRLEN];
147 snprintf(tableString, sizeof(tableString), "%u", table);
148 *contents += tableString;
154 // Doesn't return success/failure as the file is optional; it's okay if we fail to update it.
155 void updateTableNamesFile() {
156 std::string contents;
158 addTableName(RT_TABLE_LOCAL, ROUTE_TABLE_NAME_LOCAL, &contents);
159 addTableName(RT_TABLE_MAIN, ROUTE_TABLE_NAME_MAIN, &contents);
161 addTableName(ROUTE_TABLE_LOCAL_NETWORK, ROUTE_TABLE_NAME_LOCAL_NETWORK, &contents);
162 addTableName(ROUTE_TABLE_LEGACY_NETWORK, ROUTE_TABLE_NAME_LEGACY_NETWORK, &contents);
163 addTableName(ROUTE_TABLE_LEGACY_SYSTEM, ROUTE_TABLE_NAME_LEGACY_SYSTEM, &contents);
165 for (const auto& entry : interfaceToTable) {
166 addTableName(entry.second, entry.first, &contents);
169 if (!WriteStringToFile(contents, RT_TABLES_PATH, RT_TABLES_MODE, AID_SYSTEM, AID_WIFI)) {
170 ALOGE("failed to write to %s (%s)", RT_TABLES_PATH, strerror(errno));
175 // Sends a netlink request and expects an ack.
176 // |iov| is an array of struct iovec that contains the netlink message payload.
177 // The netlink header is generated by this function based on |action| and |flags|.
178 // Returns -errno if there was an error or if the kernel reported an error.
179 WARN_UNUSED_RESULT int sendNetlinkRequest(uint16_t action, uint16_t flags, iovec* iov, int iovlen) {
181 .nlmsg_type = action,
182 .nlmsg_flags = flags,
184 iov[0].iov_base = &nlmsg;
185 iov[0].iov_len = sizeof(nlmsg);
186 for (int i = 0; i < iovlen; ++i) {
187 nlmsg.nlmsg_len += iov[i].iov_len;
196 int sock = socket(AF_NETLINK, SOCK_DGRAM | SOCK_CLOEXEC, NETLINK_ROUTE);
198 connect(sock, reinterpret_cast<const sockaddr*>(&NETLINK_ADDRESS),
199 sizeof(NETLINK_ADDRESS)) != -1 &&
200 writev(sock, iov, iovlen) != -1 &&
201 (ret = recv(sock, &response, sizeof(response), 0)) != -1) {
202 if (ret == sizeof(response)) {
203 ret = response.err.error; // Netlink errors are negative errno.
205 ALOGE("netlink response contains error (%s)", strerror(-ret));
208 ALOGE("bad netlink response message size (%d != %zu)", ret, sizeof(response));
212 ALOGE("netlink socket/connect/writev/recv failed (%s)", strerror(errno));
223 // Returns 0 on success or negative errno on failure.
224 int padInterfaceName(const char* input, char* name, size_t* length, uint16_t* padding) {
230 *length = strlcpy(name, input, IFNAMSIZ) + 1;
231 if (*length > IFNAMSIZ) {
232 ALOGE("interface name too long (%zu > %u)", *length, IFNAMSIZ);
233 return -ENAMETOOLONG;
235 *padding = RTA_SPACE(*length) - RTA_LENGTH(*length);
239 // Adds or removes a routing rule for IPv4 and IPv6.
241 // + If |table| is non-zero, the rule points at the specified routing table. Otherwise, the rule
242 // returns ENETUNREACH.
243 // + If |mask| is non-zero, the rule matches the specified fwmark and mask. Otherwise, |fwmark| is
245 // + If |iif| is non-NULL, the rule matches the specified incoming interface.
246 // + If |oif| is non-NULL, the rule matches the specified outgoing interface.
247 // + If |uidStart| and |uidEnd| are not INVALID_UID, the rule matches packets from UIDs in that
248 // range (inclusive). Otherwise, the rule matches packets from all UIDs.
250 // Returns 0 on success or negative errno on failure.
251 WARN_UNUSED_RESULT int modifyIpRule(uint16_t action, uint32_t priority, uint32_t table,
252 uint32_t fwmark, uint32_t mask, const char* iif,
253 const char* oif, uid_t uidStart, uid_t uidEnd) {
254 // Ensure that if you set a bit in the fwmark, it's not being ignored by the mask.
255 if (fwmark & ~mask) {
256 ALOGE("mask 0x%x does not select all the bits set in fwmark 0x%x", mask, fwmark);
260 // Interface names must include exactly one terminating NULL and be properly padded, or older
261 // kernels will refuse to delete rules.
262 char iifName[IFNAMSIZ], oifName[IFNAMSIZ];
263 size_t iifLength, oifLength;
264 uint16_t iifPadding, oifPadding;
265 if (int ret = padInterfaceName(iif, iifName, &iifLength, &iifPadding)) {
268 if (int ret = padInterfaceName(oif, oifName, &oifLength, &oifPadding)) {
272 // Either both start and end UID must be specified, or neither.
273 if ((uidStart == INVALID_UID) != (uidEnd == INVALID_UID)) {
274 ALOGE("incompatible start and end UIDs (%u vs %u)", uidStart, uidEnd);
277 bool isUidRule = (uidStart != INVALID_UID);
279 // Assemble a rule request and put it in an array of iovec structures.
280 fib_rule_hdr rule = {
281 .action = static_cast<uint8_t>(table != RT_TABLE_UNSPEC ? FR_ACT_TO_TBL :
285 rtattr fraIifName = { U16_RTA_LENGTH(iifLength), FRA_IIFNAME };
286 rtattr fraOifName = { U16_RTA_LENGTH(oifLength), FRA_OIFNAME };
290 { &rule, sizeof(rule) },
291 { &FRATTR_PRIORITY, sizeof(FRATTR_PRIORITY) },
292 { &priority, sizeof(priority) },
293 { &FRATTR_TABLE, table != RT_TABLE_UNSPEC ? sizeof(FRATTR_TABLE) : 0 },
294 { &table, table != RT_TABLE_UNSPEC ? sizeof(table) : 0 },
295 { &FRATTR_FWMARK, mask ? sizeof(FRATTR_FWMARK) : 0 },
296 { &fwmark, mask ? sizeof(fwmark) : 0 },
297 { &FRATTR_FWMASK, mask ? sizeof(FRATTR_FWMASK) : 0 },
298 { &mask, mask ? sizeof(mask) : 0 },
299 { &FRATTR_UID_START, isUidRule ? sizeof(FRATTR_UID_START) : 0 },
300 { &uidStart, isUidRule ? sizeof(uidStart) : 0 },
301 { &FRATTR_UID_END, isUidRule ? sizeof(FRATTR_UID_END) : 0 },
302 { &uidEnd, isUidRule ? sizeof(uidEnd) : 0 },
303 { &fraIifName, iif != IIF_NONE ? sizeof(fraIifName) : 0 },
304 { iifName, iifLength },
305 { PADDING_BUFFER, iifPadding },
306 { &fraOifName, oif != OIF_NONE ? sizeof(fraOifName) : 0 },
307 { oifName, oifLength },
308 { PADDING_BUFFER, oifPadding },
311 uint16_t flags = (action == RTM_NEWRULE) ? NETLINK_CREATE_REQUEST_FLAGS : NETLINK_REQUEST_FLAGS;
312 for (size_t i = 0; i < ARRAY_SIZE(AF_FAMILIES); ++i) {
313 rule.family = AF_FAMILIES[i];
314 if (int ret = sendNetlinkRequest(action, flags, iov, ARRAY_SIZE(iov))) {
322 WARN_UNUSED_RESULT int modifyIpRule(uint16_t action, uint32_t priority, uint32_t table,
323 uint32_t fwmark, uint32_t mask) {
324 return modifyIpRule(action, priority, table, fwmark, mask, IIF_NONE, OIF_NONE, INVALID_UID,
328 // Adds or deletes an IPv4 or IPv6 route.
329 // Returns 0 on success or negative errno on failure.
330 WARN_UNUSED_RESULT int modifyIpRoute(uint16_t action, uint32_t table, const char* interface,
331 const char* destination, const char* nexthop) {
332 // At least the destination must be non-null.
334 ALOGE("null destination");
339 uint8_t rawAddress[sizeof(in6_addr)];
341 uint8_t prefixLength;
342 int rawLength = parsePrefix(destination, &family, rawAddress, sizeof(rawAddress),
345 ALOGE("parsePrefix failed for destination %s (%s)", destination, strerror(-rawLength));
349 if (static_cast<size_t>(rawLength) > sizeof(rawAddress)) {
350 ALOGE("impossible! address too long (%d vs %zu)", rawLength, sizeof(rawAddress));
351 return -ENOBUFS; // Cannot happen; parsePrefix only supports IPv4 and IPv6.
354 uint8_t type = RTN_UNICAST;
356 uint8_t rawNexthop[sizeof(in6_addr)];
358 if (nexthop && !strcmp(nexthop, "unreachable")) {
359 type = RTN_UNREACHABLE;
360 // 'interface' is likely non-NULL, as the caller (modifyRoute()) likely used it to lookup
361 // the table number. But it's an error to specify an interface ("dev ...") or a nexthop for
362 // unreachable routes, so nuke them. (IPv6 allows them to be specified; IPv4 doesn't.)
363 interface = OIF_NONE;
365 } else if (nexthop && !strcmp(nexthop, "throw")) {
367 interface = OIF_NONE;
370 // If an interface was specified, find the ifindex.
371 if (interface != OIF_NONE) {
372 ifindex = if_nametoindex(interface);
374 ALOGE("cannot find interface %s", interface);
379 // If a nexthop was specified, parse it as the same family as the prefix.
380 if (nexthop && inet_pton(family, nexthop, rawNexthop) <= 0) {
381 ALOGE("inet_pton failed for nexthop %s", nexthop);
386 // Assemble a rtmsg and put it in an array of iovec structures.
388 .rtm_protocol = RTPROT_STATIC,
390 .rtm_family = family,
391 .rtm_dst_len = prefixLength,
392 .rtm_scope = static_cast<uint8_t>(nexthop ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK),
395 rtattr rtaDst = { U16_RTA_LENGTH(rawLength), RTA_DST };
396 rtattr rtaGateway = { U16_RTA_LENGTH(rawLength), RTA_GATEWAY };
400 { &route, sizeof(route) },
401 { &RTATTR_TABLE, sizeof(RTATTR_TABLE) },
402 { &table, sizeof(table) },
403 { &rtaDst, sizeof(rtaDst) },
404 { rawAddress, static_cast<size_t>(rawLength) },
405 { &RTATTR_OIF, interface != OIF_NONE ? sizeof(RTATTR_OIF) : 0 },
406 { &ifindex, interface != OIF_NONE ? sizeof(ifindex) : 0 },
407 { &rtaGateway, nexthop ? sizeof(rtaGateway) : 0 },
408 { rawNexthop, nexthop ? static_cast<size_t>(rawLength) : 0 },
411 uint16_t flags = (action == RTM_NEWROUTE) ? NETLINK_CREATE_REQUEST_FLAGS :
412 NETLINK_REQUEST_FLAGS;
413 return sendNetlinkRequest(action, flags, iov, ARRAY_SIZE(iov));
416 // An iptables rule to mark incoming packets on a network with the netId of the network.
418 // This is so that the kernel can:
419 // + Use the right fwmark for (and thus correctly route) replies (e.g.: TCP RST, ICMP errors, ping
420 // replies, SYN-ACKs, etc).
421 // + Mark sockets that accept connections from this interface so that the connection stays on the
423 WARN_UNUSED_RESULT int modifyIncomingPacketMark(unsigned netId, const char* interface,
424 Permission permission, bool add) {
427 fwmark.netId = netId;
428 fwmark.explicitlySelected = true;
429 fwmark.protectedFromVpn = true;
430 fwmark.permission = permission;
432 char markString[UINT32_HEX_STRLEN];
433 snprintf(markString, sizeof(markString), "0x%x", fwmark.intValue);
435 if (execIptables(V4V6, "-t", "mangle", add ? "-A" : "-D", "INPUT", "-i", interface, "-j",
436 "MARK", "--set-mark", markString, NULL)) {
437 ALOGE("failed to change iptables rule that sets incoming packet mark");
444 // A rule to route responses to the local network forwarded via the VPN.
446 // When a VPN is in effect, packets from the local network to upstream networks are forwarded into
447 // the VPN's tunnel interface. When the VPN forwards the responses, they emerge out of the tunnel.
448 WARN_UNUSED_RESULT int modifyVpnOutputToLocalRule(const char* vpnInterface, bool add) {
449 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_VPN_OUTPUT_TO_LOCAL,
450 ROUTE_TABLE_LOCAL_NETWORK, MARK_UNSET, MARK_UNSET, vpnInterface, OIF_NONE,
451 INVALID_UID, INVALID_UID);
454 // A rule to route all traffic from a given set of UIDs to go over the VPN.
456 // Notice that this rule doesn't use the netId. I.e., no matter what netId the user's socket may
457 // have, if they are subject to this VPN, their traffic has to go through it. Allows the traffic to
458 // bypass the VPN if the protectedFromVpn bit is set.
459 WARN_UNUSED_RESULT int modifyVpnUidRangeRule(uint32_t table, uid_t uidStart, uid_t uidEnd,
460 bool secure, bool add) {
464 fwmark.protectedFromVpn = false;
465 mask.protectedFromVpn = true;
470 priority = RULE_PRIORITY_SECURE_VPN;
472 priority = RULE_PRIORITY_BYPASSABLE_VPN;
474 fwmark.explicitlySelected = false;
475 mask.explicitlySelected = true;
478 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, priority, table, fwmark.intValue,
479 mask.intValue, IIF_NONE, OIF_NONE, uidStart, uidEnd);
482 // A rule to allow system apps to send traffic over this VPN even if they are not part of the target
485 // This is needed for DnsProxyListener to correctly resolve a request for a user who is in the
486 // target set, but where the DnsProxyListener itself is not.
487 WARN_UNUSED_RESULT int modifyVpnSystemPermissionRule(unsigned netId, uint32_t table, bool secure,
492 fwmark.netId = netId;
493 mask.netId = FWMARK_NET_ID_MASK;
495 fwmark.permission = PERMISSION_SYSTEM;
496 mask.permission = PERMISSION_SYSTEM;
498 uint32_t priority = secure ? RULE_PRIORITY_SECURE_VPN : RULE_PRIORITY_BYPASSABLE_VPN;
500 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, priority, table, fwmark.intValue,
504 // A rule to route traffic based on an explicitly chosen network.
506 // Supports apps that use the multinetwork APIs to restrict their traffic to a network.
508 // Even though we check permissions at the time we set a netId into the fwmark of a socket, we need
509 // to check it again in the rules here, because a network's permissions may have been updated via
510 // modifyNetworkPermission().
511 WARN_UNUSED_RESULT int modifyExplicitNetworkRule(unsigned netId, uint32_t table,
512 Permission permission, uid_t uidStart,
513 uid_t uidEnd, bool add) {
517 fwmark.netId = netId;
518 mask.netId = FWMARK_NET_ID_MASK;
520 fwmark.explicitlySelected = true;
521 mask.explicitlySelected = true;
523 fwmark.permission = permission;
524 mask.permission = permission;
526 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_EXPLICIT_NETWORK, table,
527 fwmark.intValue, mask.intValue, IIF_NONE, OIF_NONE, uidStart, uidEnd);
530 // A rule to route traffic based on a chosen outgoing interface.
532 // Supports apps that use SO_BINDTODEVICE or IP_PKTINFO options and the kernel that already knows
533 // the outgoing interface (typically for link-local communications).
534 WARN_UNUSED_RESULT int modifyOutputInterfaceRule(const char* interface, uint32_t table,
535 Permission permission, uid_t uidStart,
536 uid_t uidEnd, bool add) {
540 fwmark.permission = permission;
541 mask.permission = permission;
543 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_OUTPUT_INTERFACE, table,
544 fwmark.intValue, mask.intValue, IIF_NONE, interface, uidStart, uidEnd);
547 // A rule to route traffic based on the chosen network.
549 // This is for sockets that have not explicitly requested a particular network, but have been
550 // bound to one when they called connect(). This ensures that sockets connected on a particular
551 // network stay on that network even if the default network changes.
552 WARN_UNUSED_RESULT int modifyImplicitNetworkRule(unsigned netId, uint32_t table,
553 Permission permission, bool add) {
557 fwmark.netId = netId;
558 mask.netId = FWMARK_NET_ID_MASK;
560 fwmark.explicitlySelected = false;
561 mask.explicitlySelected = true;
563 fwmark.permission = permission;
564 mask.permission = permission;
566 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_IMPLICIT_NETWORK, table,
567 fwmark.intValue, mask.intValue);
570 // A rule to enable split tunnel VPNs.
572 // If a packet with a VPN's netId doesn't find a route in the VPN's routing table, it's allowed to
573 // go over the default network, provided it wasn't explicitly restricted to the VPN and has the
574 // permissions required by the default network.
575 WARN_UNUSED_RESULT int modifyVpnFallthroughRule(uint16_t action, unsigned vpnNetId,
576 const char* physicalInterface,
577 Permission permission) {
578 uint32_t table = getRouteTableForInterface(physicalInterface);
579 if (table == RT_TABLE_UNSPEC) {
586 fwmark.netId = vpnNetId;
587 mask.netId = FWMARK_NET_ID_MASK;
589 fwmark.explicitlySelected = false;
590 mask.explicitlySelected = true;
592 fwmark.permission = permission;
593 mask.permission = permission;
595 return modifyIpRule(action, RULE_PRIORITY_VPN_FALLTHROUGH, table, fwmark.intValue,
599 // Add rules to allow legacy routes added through the requestRouteToHost() API.
600 WARN_UNUSED_RESULT int addLegacyRouteRules() {
604 fwmark.explicitlySelected = false;
605 mask.explicitlySelected = true;
607 // Rules to allow legacy routes to override the default network.
608 if (int ret = modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_LEGACY_SYSTEM, ROUTE_TABLE_LEGACY_SYSTEM,
609 fwmark.intValue, mask.intValue)) {
612 if (int ret = modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_LEGACY_NETWORK,
613 ROUTE_TABLE_LEGACY_NETWORK, fwmark.intValue, mask.intValue)) {
617 fwmark.permission = PERMISSION_SYSTEM;
618 mask.permission = PERMISSION_SYSTEM;
620 // A rule to allow legacy routes from system apps to override VPNs.
621 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_VPN_OVERRIDE_SYSTEM, ROUTE_TABLE_LEGACY_SYSTEM,
622 fwmark.intValue, mask.intValue);
625 // Add rules to lookup the local network when specified explicitly or otherwise.
626 WARN_UNUSED_RESULT int addLocalNetworkRules(unsigned localNetId) {
627 if (int ret = modifyExplicitNetworkRule(localNetId, ROUTE_TABLE_LOCAL_NETWORK, PERMISSION_NONE,
628 INVALID_UID, INVALID_UID, ACTION_ADD)) {
635 fwmark.explicitlySelected = false;
636 mask.explicitlySelected = true;
638 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_LOCAL_NETWORK, ROUTE_TABLE_LOCAL_NETWORK,
639 fwmark.intValue, mask.intValue);
642 // Add a new rule to look up the 'main' table, with the same selectors as the "default network"
643 // rule, but with a lower priority. We will never create routes in the main table; it should only be
644 // used for directly-connected routes implicitly created by the kernel when adding IP addresses.
645 // This is necessary, for example, when adding a route through a directly-connected gateway: in
646 // order to add the route, there must already be a directly-connected route that covers the gateway.
647 WARN_UNUSED_RESULT int addDirectlyConnectedRule() {
651 fwmark.netId = NETID_UNSET;
652 mask.netId = FWMARK_NET_ID_MASK;
654 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_DIRECTLY_CONNECTED, RT_TABLE_MAIN,
655 fwmark.intValue, mask.intValue, IIF_NONE, OIF_NONE, UID_ROOT, UID_ROOT);
658 // Add an explicit unreachable rule close to the end of the prioriy list to make it clear that
659 // relying on the kernel-default "from all lookup main" rule at priority 32766 is not intended
660 // behaviour. We do flush the kernel-default rules at startup, but having an explicit unreachable
661 // rule will hopefully make things even clearer.
662 WARN_UNUSED_RESULT int addUnreachableRule() {
663 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_UNREACHABLE, RT_TABLE_UNSPEC, MARK_UNSET,
667 WARN_UNUSED_RESULT int modifyLocalNetwork(unsigned netId, const char* interface, bool add) {
668 if (int ret = modifyIncomingPacketMark(netId, interface, PERMISSION_NONE, add)) {
671 return modifyOutputInterfaceRule(interface, ROUTE_TABLE_LOCAL_NETWORK, PERMISSION_NONE,
672 INVALID_UID, INVALID_UID, add);
675 WARN_UNUSED_RESULT int modifyPhysicalNetwork(unsigned netId, const char* interface,
676 Permission permission, bool add) {
677 uint32_t table = getRouteTableForInterface(interface);
678 if (table == RT_TABLE_UNSPEC) {
682 if (int ret = modifyIncomingPacketMark(netId, interface, permission, add)) {
685 if (int ret = modifyExplicitNetworkRule(netId, table, permission, INVALID_UID, INVALID_UID,
689 if (int ret = modifyOutputInterfaceRule(interface, table, permission, INVALID_UID, INVALID_UID,
693 return modifyImplicitNetworkRule(netId, table, permission, add);
696 WARN_UNUSED_RESULT int modifyVirtualNetwork(unsigned netId, const char* interface,
697 const UidRanges& uidRanges, bool secure, bool add,
698 bool modifyNonUidBasedRules) {
699 uint32_t table = getRouteTableForInterface(interface);
700 if (table == RT_TABLE_UNSPEC) {
704 for (const UidRanges::Range& range : uidRanges.getRanges()) {
705 if (int ret = modifyVpnUidRangeRule(table, range.first, range.second, secure, add)) {
708 if (int ret = modifyExplicitNetworkRule(netId, table, PERMISSION_NONE, range.first,
709 range.second, add)) {
712 if (int ret = modifyOutputInterfaceRule(interface, table, PERMISSION_NONE, range.first,
713 range.second, add)) {
718 if (modifyNonUidBasedRules) {
719 if (int ret = modifyIncomingPacketMark(netId, interface, PERMISSION_NONE, add)) {
722 if (int ret = modifyVpnOutputToLocalRule(interface, add)) {
725 if (int ret = modifyVpnSystemPermissionRule(netId, table, secure, add)) {
728 return modifyExplicitNetworkRule(netId, table, PERMISSION_NONE, UID_ROOT, UID_ROOT, add);
734 WARN_UNUSED_RESULT int modifyDefaultNetwork(uint16_t action, const char* interface,
735 Permission permission) {
736 uint32_t table = getRouteTableForInterface(interface);
737 if (table == RT_TABLE_UNSPEC) {
744 fwmark.netId = NETID_UNSET;
745 mask.netId = FWMARK_NET_ID_MASK;
747 fwmark.permission = permission;
748 mask.permission = permission;
750 return modifyIpRule(action, RULE_PRIORITY_DEFAULT_NETWORK, table, fwmark.intValue,
754 WARN_UNUSED_RESULT int modifyTetheredNetwork(uint16_t action, const char* inputInterface,
755 const char* outputInterface) {
756 uint32_t table = getRouteTableForInterface(outputInterface);
757 if (table == RT_TABLE_UNSPEC) {
761 return modifyIpRule(action, RULE_PRIORITY_TETHERING, table, MARK_UNSET, MARK_UNSET,
762 inputInterface, OIF_NONE, INVALID_UID, INVALID_UID);
765 // Returns 0 on success or negative errno on failure.
766 WARN_UNUSED_RESULT int flushRules() {
767 for (size_t i = 0; i < ARRAY_SIZE(IP_VERSIONS); ++i) {
768 const char* argv[] = {
774 if (android_fork_execvp(ARRAY_SIZE(argv), const_cast<char**>(argv), NULL, false, false)) {
775 ALOGE("failed to flush rules");
782 // Adds or removes an IPv4 or IPv6 route to the specified table and, if it's a directly-connected
783 // route, to the main table as well.
784 // Returns 0 on success or negative errno on failure.
785 WARN_UNUSED_RESULT int modifyRoute(uint16_t action, const char* interface, const char* destination,
786 const char* nexthop, RouteController::TableType tableType) {
789 case RouteController::INTERFACE: {
790 table = getRouteTableForInterface(interface);
791 if (table == RT_TABLE_UNSPEC) {
796 case RouteController::LOCAL_NETWORK: {
797 table = ROUTE_TABLE_LOCAL_NETWORK;
800 case RouteController::LEGACY_NETWORK: {
801 table = ROUTE_TABLE_LEGACY_NETWORK;
804 case RouteController::LEGACY_SYSTEM: {
805 table = ROUTE_TABLE_LEGACY_SYSTEM;
810 int ret = modifyIpRoute(action, table, interface, destination, nexthop);
811 // Trying to add a route that already exists shouldn't cause an error.
812 if (ret && !(action == RTM_NEWROUTE && ret == -EEXIST)) {
819 // Returns 0 on success or negative errno on failure.
820 WARN_UNUSED_RESULT int flushRoutes(const char* interface) {
821 uint32_t table = getRouteTableForInterface(interface);
822 if (table == RT_TABLE_UNSPEC) {
826 char tableString[UINT32_STRLEN];
827 snprintf(tableString, sizeof(tableString), "%u", table);
830 for (size_t i = 0; i < ARRAY_SIZE(IP_VERSIONS); ++i) {
831 const char* argv[] = {
840 // A flush works by dumping routes and deleting each route as it's returned, and it can
841 // fail if something else deletes the route between the dump and the delete. This can
842 // happen, for example, if an interface goes down while we're trying to flush its routes.
843 // So try multiple times and only return an error if the last attempt fails.
845 // TODO: replace this with our own netlink code.
846 unsigned attempts = 0;
849 err = android_fork_execvp(ARRAY_SIZE(argv), const_cast<char**>(argv),
852 } while (err != 0 && attempts < ROUTE_FLUSH_ATTEMPTS);
854 ALOGE("failed to flush %s routes in table %s after %d attempts",
855 IP_VERSIONS[i], tableString, attempts);
860 // If we failed to flush routes, the caller may elect to keep this interface around, so keep
861 // track of its name.
863 interfaceToTable.erase(interface);
871 int RouteController::Init(unsigned localNetId) {
872 if (int ret = flushRules()) {
875 if (int ret = addLegacyRouteRules()) {
878 if (int ret = addLocalNetworkRules(localNetId)) {
881 if (int ret = addDirectlyConnectedRule()) {
884 if (int ret = addUnreachableRule()) {
887 updateTableNamesFile();
891 int RouteController::addInterfaceToLocalNetwork(unsigned netId, const char* interface) {
892 return modifyLocalNetwork(netId, interface, ACTION_ADD);
895 int RouteController::removeInterfaceFromLocalNetwork(unsigned netId, const char* interface) {
896 return modifyLocalNetwork(netId, interface, ACTION_DEL);
899 int RouteController::addInterfaceToPhysicalNetwork(unsigned netId, const char* interface,
900 Permission permission) {
901 if (int ret = modifyPhysicalNetwork(netId, interface, permission, ACTION_ADD)) {
904 updateTableNamesFile();
908 int RouteController::removeInterfaceFromPhysicalNetwork(unsigned netId, const char* interface,
909 Permission permission) {
910 if (int ret = modifyPhysicalNetwork(netId, interface, permission, ACTION_DEL)) {
913 if (int ret = flushRoutes(interface)) {
916 updateTableNamesFile();
920 int RouteController::addInterfaceToVirtualNetwork(unsigned netId, const char* interface,
921 bool secure, const UidRanges& uidRanges) {
922 if (int ret = modifyVirtualNetwork(netId, interface, uidRanges, secure, ACTION_ADD,
923 MODIFY_NON_UID_BASED_RULES)) {
926 updateTableNamesFile();
930 int RouteController::removeInterfaceFromVirtualNetwork(unsigned netId, const char* interface,
931 bool secure, const UidRanges& uidRanges) {
932 if (int ret = modifyVirtualNetwork(netId, interface, uidRanges, secure, ACTION_DEL,
933 MODIFY_NON_UID_BASED_RULES)) {
936 if (int ret = flushRoutes(interface)) {
939 updateTableNamesFile();
943 int RouteController::modifyPhysicalNetworkPermission(unsigned netId, const char* interface,
944 Permission oldPermission,
945 Permission newPermission) {
946 // Add the new rules before deleting the old ones, to avoid race conditions.
947 if (int ret = modifyPhysicalNetwork(netId, interface, newPermission, ACTION_ADD)) {
950 return modifyPhysicalNetwork(netId, interface, oldPermission, ACTION_DEL);
953 int RouteController::addUsersToVirtualNetwork(unsigned netId, const char* interface, bool secure,
954 const UidRanges& uidRanges) {
955 return modifyVirtualNetwork(netId, interface, uidRanges, secure, ACTION_ADD,
956 !MODIFY_NON_UID_BASED_RULES);
959 int RouteController::removeUsersFromVirtualNetwork(unsigned netId, const char* interface,
960 bool secure, const UidRanges& uidRanges) {
961 return modifyVirtualNetwork(netId, interface, uidRanges, secure, ACTION_DEL,
962 !MODIFY_NON_UID_BASED_RULES);
965 int RouteController::addInterfaceToDefaultNetwork(const char* interface, Permission permission) {
966 return modifyDefaultNetwork(RTM_NEWRULE, interface, permission);
969 int RouteController::removeInterfaceFromDefaultNetwork(const char* interface,
970 Permission permission) {
971 return modifyDefaultNetwork(RTM_DELRULE, interface, permission);
974 int RouteController::addRoute(const char* interface, const char* destination, const char* nexthop,
975 TableType tableType) {
976 return modifyRoute(RTM_NEWROUTE, interface, destination, nexthop, tableType);
979 int RouteController::removeRoute(const char* interface, const char* destination,
980 const char* nexthop, TableType tableType) {
981 return modifyRoute(RTM_DELROUTE, interface, destination, nexthop, tableType);
984 int RouteController::enableTethering(const char* inputInterface, const char* outputInterface) {
985 return modifyTetheredNetwork(RTM_NEWRULE, inputInterface, outputInterface);
988 int RouteController::disableTethering(const char* inputInterface, const char* outputInterface) {
989 return modifyTetheredNetwork(RTM_DELRULE, inputInterface, outputInterface);
992 int RouteController::addVirtualNetworkFallthrough(unsigned vpnNetId, const char* physicalInterface,
993 Permission permission) {
994 return modifyVpnFallthroughRule(RTM_NEWRULE, vpnNetId, physicalInterface, permission);
997 int RouteController::removeVirtualNetworkFallthrough(unsigned vpnNetId,
998 const char* physicalInterface,
999 Permission permission) {
1000 return modifyVpnFallthroughRule(RTM_DELRULE, vpnNetId, physicalInterface, permission);