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"
32 #include "DummyNetwork.h"
34 #include "android-base/file.h"
35 #define LOG_TAG "Netd"
37 #include "logwrap/logwrap.h"
38 #include "netutils/ifc.h"
39 #include "resolv_netid.h"
41 using android::base::WriteStringToFile;
45 // BEGIN CONSTANTS --------------------------------------------------------------------------------
47 const uint32_t RULE_PRIORITY_VPN_OVERRIDE_SYSTEM = 10000;
48 const uint32_t RULE_PRIORITY_VPN_OVERRIDE_OIF = 10500;
49 const uint32_t RULE_PRIORITY_VPN_OUTPUT_TO_LOCAL = 11000;
50 const uint32_t RULE_PRIORITY_SECURE_VPN = 12000;
51 const uint32_t RULE_PRIORITY_EXPLICIT_NETWORK = 13000;
52 const uint32_t RULE_PRIORITY_OUTPUT_INTERFACE = 14000;
53 const uint32_t RULE_PRIORITY_LEGACY_SYSTEM = 15000;
54 const uint32_t RULE_PRIORITY_LEGACY_NETWORK = 16000;
55 const uint32_t RULE_PRIORITY_LOCAL_NETWORK = 17000;
56 const uint32_t RULE_PRIORITY_TETHERING = 18000;
57 const uint32_t RULE_PRIORITY_IMPLICIT_NETWORK = 19000;
58 const uint32_t RULE_PRIORITY_BYPASSABLE_VPN = 20000;
59 const uint32_t RULE_PRIORITY_VPN_FALLTHROUGH = 21000;
60 const uint32_t RULE_PRIORITY_DEFAULT_NETWORK = 22000;
61 const uint32_t RULE_PRIORITY_DIRECTLY_CONNECTED = 23000;
62 const uint32_t RULE_PRIORITY_UNREACHABLE = 32000;
64 const uint32_t ROUTE_TABLE_LOCAL_NETWORK = 97;
65 const uint32_t ROUTE_TABLE_LEGACY_NETWORK = 98;
66 const uint32_t ROUTE_TABLE_LEGACY_SYSTEM = 99;
68 const char* const ROUTE_TABLE_NAME_LOCAL_NETWORK = "local_network";
69 const char* const ROUTE_TABLE_NAME_LEGACY_NETWORK = "legacy_network";
70 const char* const ROUTE_TABLE_NAME_LEGACY_SYSTEM = "legacy_system";
72 const char* const ROUTE_TABLE_NAME_LOCAL = "local";
73 const char* const ROUTE_TABLE_NAME_MAIN = "main";
75 // TODO: These values aren't defined by the Linux kernel, because our UID routing changes are not
76 // upstream (yet?), so we can't just pick them up from kernel headers. When (if?) the changes make
77 // it upstream, we'll remove this and rely on the kernel header values. For now, add a static assert
78 // that will warn us if upstream has given these values some other meaning.
79 const uint16_t FRA_UID_START = 18;
80 const uint16_t FRA_UID_END = 19;
81 static_assert(FRA_UID_START > FRA_MAX,
82 "Android-specific FRA_UID_{START,END} values also assigned in Linux uapi. "
83 "Check that these values match what the kernel does and then update this assertion.");
85 const uint16_t NETLINK_REQUEST_FLAGS = NLM_F_REQUEST | NLM_F_ACK;
86 const uint16_t NETLINK_CREATE_REQUEST_FLAGS = NETLINK_REQUEST_FLAGS | NLM_F_CREATE | NLM_F_EXCL;
88 const sockaddr_nl NETLINK_ADDRESS = {AF_NETLINK, 0, 0, 0};
90 const uint8_t AF_FAMILIES[] = {AF_INET, AF_INET6};
92 const char* const IP_VERSIONS[] = {"-4", "-6"};
94 const uid_t UID_ROOT = 0;
95 const char* const IIF_LOOPBACK = "lo";
96 const char* const IIF_NONE = NULL;
97 const char* const OIF_NONE = NULL;
98 const bool ACTION_ADD = true;
99 const bool ACTION_DEL = false;
100 const bool MODIFY_NON_UID_BASED_RULES = true;
102 const char* const RT_TABLES_PATH = "/data/misc/net/rt_tables";
103 const mode_t RT_TABLES_MODE = S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH; // mode 0644, rw-r--r--
105 const unsigned ROUTE_FLUSH_ATTEMPTS = 2;
107 // Avoids "non-constant-expression cannot be narrowed from type 'unsigned int' to 'unsigned short'"
108 // warnings when using RTA_LENGTH(x) inside static initializers (even when x is already uint16_t).
109 constexpr uint16_t U16_RTA_LENGTH(uint16_t x) {
110 return RTA_LENGTH(x);
113 // These are practically const, but can't be declared so, because they are used to initialize
114 // non-const pointers ("void* iov_base") in iovec arrays.
115 rtattr FRATTR_PRIORITY = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_PRIORITY };
116 rtattr FRATTR_TABLE = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_TABLE };
117 rtattr FRATTR_FWMARK = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_FWMARK };
118 rtattr FRATTR_FWMASK = { U16_RTA_LENGTH(sizeof(uint32_t)), FRA_FWMASK };
119 rtattr FRATTR_UID_START = { U16_RTA_LENGTH(sizeof(uid_t)), FRA_UID_START };
120 rtattr FRATTR_UID_END = { U16_RTA_LENGTH(sizeof(uid_t)), FRA_UID_END };
122 rtattr RTATTR_TABLE = { U16_RTA_LENGTH(sizeof(uint32_t)), RTA_TABLE };
123 rtattr RTATTR_OIF = { U16_RTA_LENGTH(sizeof(uint32_t)), RTA_OIF };
125 uint8_t PADDING_BUFFER[RTA_ALIGNTO] = {0, 0, 0, 0};
127 // END CONSTANTS ----------------------------------------------------------------------------------
129 // No locks needed because RouteController is accessed only from one thread (in CommandListener).
130 std::map<std::string, uint32_t> interfaceToTable;
132 uint32_t getRouteTableForInterface(const char* interface) {
133 uint32_t index = if_nametoindex(interface);
135 index += RouteController::ROUTE_TABLE_OFFSET_FROM_INDEX;
136 interfaceToTable[interface] = index;
139 // If the interface goes away if_nametoindex() will return 0 but we still need to know
140 // the index so we can remove the rules and routes.
141 auto iter = interfaceToTable.find(interface);
142 if (iter == interfaceToTable.end()) {
143 ALOGE("cannot find interface %s", interface);
144 return RT_TABLE_UNSPEC;
149 void addTableName(uint32_t table, const std::string& name, std::string* contents) {
150 char tableString[UINT32_STRLEN];
151 snprintf(tableString, sizeof(tableString), "%u", table);
152 *contents += tableString;
158 // Doesn't return success/failure as the file is optional; it's okay if we fail to update it.
159 void updateTableNamesFile() {
160 std::string contents;
162 addTableName(RT_TABLE_LOCAL, ROUTE_TABLE_NAME_LOCAL, &contents);
163 addTableName(RT_TABLE_MAIN, ROUTE_TABLE_NAME_MAIN, &contents);
165 addTableName(ROUTE_TABLE_LOCAL_NETWORK, ROUTE_TABLE_NAME_LOCAL_NETWORK, &contents);
166 addTableName(ROUTE_TABLE_LEGACY_NETWORK, ROUTE_TABLE_NAME_LEGACY_NETWORK, &contents);
167 addTableName(ROUTE_TABLE_LEGACY_SYSTEM, ROUTE_TABLE_NAME_LEGACY_SYSTEM, &contents);
169 for (const auto& entry : interfaceToTable) {
170 addTableName(entry.second, entry.first, &contents);
173 if (!WriteStringToFile(contents, RT_TABLES_PATH, RT_TABLES_MODE, AID_SYSTEM, AID_WIFI)) {
174 ALOGE("failed to write to %s (%s)", RT_TABLES_PATH, strerror(errno));
179 // Sends a netlink request and expects an ack.
180 // |iov| is an array of struct iovec that contains the netlink message payload.
181 // The netlink header is generated by this function based on |action| and |flags|.
182 // Returns -errno if there was an error or if the kernel reported an error.
184 // Disable optimizations in ASan build.
185 // ASan reports an out-of-bounds 32-bit(!) access in the first loop of the
186 // function (over iov[]).
188 #if __has_feature(address_sanitizer)
189 __attribute__((optnone))
192 WARN_UNUSED_RESULT int sendNetlinkRequest(uint16_t action, uint16_t flags, iovec* iov, int iovlen) {
194 .nlmsg_type = action,
195 .nlmsg_flags = flags,
197 iov[0].iov_base = &nlmsg;
198 iov[0].iov_len = sizeof(nlmsg);
199 for (int i = 0; i < iovlen; ++i) {
200 nlmsg.nlmsg_len += iov[i].iov_len;
209 int sock = socket(AF_NETLINK, SOCK_DGRAM | SOCK_CLOEXEC, NETLINK_ROUTE);
211 connect(sock, reinterpret_cast<const sockaddr*>(&NETLINK_ADDRESS),
212 sizeof(NETLINK_ADDRESS)) != -1 &&
213 writev(sock, iov, iovlen) != -1 &&
214 (ret = recv(sock, &response, sizeof(response), 0)) != -1) {
215 if (ret == sizeof(response)) {
216 ret = response.err.error; // Netlink errors are negative errno.
218 ALOGE("netlink response contains error (%s)", strerror(-ret));
221 ALOGE("bad netlink response message size (%d != %zu)", ret, sizeof(response));
225 ALOGE("netlink socket/connect/writev/recv failed (%s)", strerror(errno));
236 // Returns 0 on success or negative errno on failure.
237 int padInterfaceName(const char* input, char* name, size_t* length, uint16_t* padding) {
243 *length = strlcpy(name, input, IFNAMSIZ) + 1;
244 if (*length > IFNAMSIZ) {
245 ALOGE("interface name too long (%zu > %u)", *length, IFNAMSIZ);
246 return -ENAMETOOLONG;
248 *padding = RTA_SPACE(*length) - RTA_LENGTH(*length);
252 // Adds or removes a routing rule for IPv4 and IPv6.
254 // + If |table| is non-zero, the rule points at the specified routing table. Otherwise, the table is
255 // unspecified. An unspecified table is not allowed when creating an FR_ACT_TO_TBL rule.
256 // + If |mask| is non-zero, the rule matches the specified fwmark and mask. Otherwise, |fwmark| is
258 // + If |iif| is non-NULL, the rule matches the specified incoming interface.
259 // + If |oif| is non-NULL, the rule matches the specified outgoing interface.
260 // + If |uidStart| and |uidEnd| are not INVALID_UID, the rule matches packets from UIDs in that
261 // range (inclusive). Otherwise, the rule matches packets from all UIDs.
263 // Returns 0 on success or negative errno on failure.
264 WARN_UNUSED_RESULT int modifyIpRule(uint16_t action, uint32_t priority, uint8_t ruleType,
265 uint32_t table, uint32_t fwmark, uint32_t mask, const char* iif,
266 const char* oif, uid_t uidStart, uid_t uidEnd) {
267 // Ensure that if you set a bit in the fwmark, it's not being ignored by the mask.
268 if (fwmark & ~mask) {
269 ALOGE("mask 0x%x does not select all the bits set in fwmark 0x%x", mask, fwmark);
273 // Interface names must include exactly one terminating NULL and be properly padded, or older
274 // kernels will refuse to delete rules.
275 char iifName[IFNAMSIZ], oifName[IFNAMSIZ];
276 size_t iifLength, oifLength;
277 uint16_t iifPadding, oifPadding;
278 if (int ret = padInterfaceName(iif, iifName, &iifLength, &iifPadding)) {
281 if (int ret = padInterfaceName(oif, oifName, &oifLength, &oifPadding)) {
285 // Either both start and end UID must be specified, or neither.
286 if ((uidStart == INVALID_UID) != (uidEnd == INVALID_UID)) {
287 ALOGE("incompatible start and end UIDs (%u vs %u)", uidStart, uidEnd);
291 bool isUidRule = (uidStart != INVALID_UID);
293 // Assemble a rule request and put it in an array of iovec structures.
294 fib_rule_hdr rule = {
296 // Note that here we're implicitly setting rule.table to 0. When we want to specify a
297 // non-zero table, we do this via the FRATTR_TABLE attribute.
300 // Don't ever create a rule that looks up table 0, because table 0 is the local table.
301 // It's OK to specify a table ID of 0 when deleting a rule, because that doesn't actually select
302 // table 0, it's a wildcard that matches anything.
303 if (table == RT_TABLE_UNSPEC && rule.action == FR_ACT_TO_TBL && action != RTM_DELRULE) {
304 ALOGE("RT_TABLE_UNSPEC only allowed when deleting rules");
308 rtattr fraIifName = { U16_RTA_LENGTH(iifLength), FRA_IIFNAME };
309 rtattr fraOifName = { U16_RTA_LENGTH(oifLength), FRA_OIFNAME };
313 { &rule, sizeof(rule) },
314 { &FRATTR_PRIORITY, sizeof(FRATTR_PRIORITY) },
315 { &priority, sizeof(priority) },
316 { &FRATTR_TABLE, table != RT_TABLE_UNSPEC ? sizeof(FRATTR_TABLE) : 0 },
317 { &table, table != RT_TABLE_UNSPEC ? sizeof(table) : 0 },
318 { &FRATTR_FWMARK, mask ? sizeof(FRATTR_FWMARK) : 0 },
319 { &fwmark, mask ? sizeof(fwmark) : 0 },
320 { &FRATTR_FWMASK, mask ? sizeof(FRATTR_FWMASK) : 0 },
321 { &mask, mask ? sizeof(mask) : 0 },
322 { &FRATTR_UID_START, isUidRule ? sizeof(FRATTR_UID_START) : 0 },
323 { &uidStart, isUidRule ? sizeof(uidStart) : 0 },
324 { &FRATTR_UID_END, isUidRule ? sizeof(FRATTR_UID_END) : 0 },
325 { &uidEnd, isUidRule ? sizeof(uidEnd) : 0 },
326 { &fraIifName, iif != IIF_NONE ? sizeof(fraIifName) : 0 },
327 { iifName, iifLength },
328 { PADDING_BUFFER, iifPadding },
329 { &fraOifName, oif != OIF_NONE ? sizeof(fraOifName) : 0 },
330 { oifName, oifLength },
331 { PADDING_BUFFER, oifPadding },
334 uint16_t flags = (action == RTM_NEWRULE) ? NETLINK_CREATE_REQUEST_FLAGS : NETLINK_REQUEST_FLAGS;
335 for (size_t i = 0; i < ARRAY_SIZE(AF_FAMILIES); ++i) {
336 rule.family = AF_FAMILIES[i];
337 if (int ret = sendNetlinkRequest(action, flags, iov, ARRAY_SIZE(iov))) {
345 WARN_UNUSED_RESULT int modifyIpRule(uint16_t action, uint32_t priority, uint32_t table,
346 uint32_t fwmark, uint32_t mask, const char* iif,
347 const char* oif, uid_t uidStart, uid_t uidEnd) {
348 return modifyIpRule(action, priority, FR_ACT_TO_TBL, table, fwmark, mask, iif, oif, uidStart,
352 WARN_UNUSED_RESULT int modifyIpRule(uint16_t action, uint32_t priority, uint32_t table,
353 uint32_t fwmark, uint32_t mask) {
354 return modifyIpRule(action, priority, table, fwmark, mask, IIF_NONE, OIF_NONE, INVALID_UID,
358 // Adds or deletes an IPv4 or IPv6 route.
359 // Returns 0 on success or negative errno on failure.
360 WARN_UNUSED_RESULT int modifyIpRoute(uint16_t action, uint32_t table, const char* interface,
361 const char* destination, const char* nexthop) {
362 // At least the destination must be non-null.
364 ALOGE("null destination");
369 uint8_t rawAddress[sizeof(in6_addr)];
371 uint8_t prefixLength;
372 int rawLength = parsePrefix(destination, &family, rawAddress, sizeof(rawAddress),
375 ALOGE("parsePrefix failed for destination %s (%s)", destination, strerror(-rawLength));
379 if (static_cast<size_t>(rawLength) > sizeof(rawAddress)) {
380 ALOGE("impossible! address too long (%d vs %zu)", rawLength, sizeof(rawAddress));
381 return -ENOBUFS; // Cannot happen; parsePrefix only supports IPv4 and IPv6.
384 uint8_t type = RTN_UNICAST;
386 uint8_t rawNexthop[sizeof(in6_addr)];
388 if (nexthop && !strcmp(nexthop, "unreachable")) {
389 type = RTN_UNREACHABLE;
390 // 'interface' is likely non-NULL, as the caller (modifyRoute()) likely used it to lookup
391 // the table number. But it's an error to specify an interface ("dev ...") or a nexthop for
392 // unreachable routes, so nuke them. (IPv6 allows them to be specified; IPv4 doesn't.)
393 interface = OIF_NONE;
395 } else if (nexthop && !strcmp(nexthop, "throw")) {
397 interface = OIF_NONE;
400 // If an interface was specified, find the ifindex.
401 if (interface != OIF_NONE) {
402 ifindex = if_nametoindex(interface);
404 ALOGE("cannot find interface %s", interface);
409 // If a nexthop was specified, parse it as the same family as the prefix.
410 if (nexthop && inet_pton(family, nexthop, rawNexthop) <= 0) {
411 ALOGE("inet_pton failed for nexthop %s", nexthop);
416 // Assemble a rtmsg and put it in an array of iovec structures.
418 .rtm_protocol = RTPROT_STATIC,
420 .rtm_family = family,
421 .rtm_dst_len = prefixLength,
422 .rtm_scope = static_cast<uint8_t>(nexthop ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK),
425 rtattr rtaDst = { U16_RTA_LENGTH(rawLength), RTA_DST };
426 rtattr rtaGateway = { U16_RTA_LENGTH(rawLength), RTA_GATEWAY };
430 { &route, sizeof(route) },
431 { &RTATTR_TABLE, sizeof(RTATTR_TABLE) },
432 { &table, sizeof(table) },
433 { &rtaDst, sizeof(rtaDst) },
434 { rawAddress, static_cast<size_t>(rawLength) },
435 { &RTATTR_OIF, interface != OIF_NONE ? sizeof(RTATTR_OIF) : 0 },
436 { &ifindex, interface != OIF_NONE ? sizeof(ifindex) : 0 },
437 { &rtaGateway, nexthop ? sizeof(rtaGateway) : 0 },
438 { rawNexthop, nexthop ? static_cast<size_t>(rawLength) : 0 },
441 uint16_t flags = (action == RTM_NEWROUTE) ? NETLINK_CREATE_REQUEST_FLAGS :
442 NETLINK_REQUEST_FLAGS;
443 return sendNetlinkRequest(action, flags, iov, ARRAY_SIZE(iov));
446 // An iptables rule to mark incoming packets on a network with the netId of the network.
448 // This is so that the kernel can:
449 // + Use the right fwmark for (and thus correctly route) replies (e.g.: TCP RST, ICMP errors, ping
450 // replies, SYN-ACKs, etc).
451 // + Mark sockets that accept connections from this interface so that the connection stays on the
453 WARN_UNUSED_RESULT int modifyIncomingPacketMark(unsigned netId, const char* interface,
454 Permission permission, bool add) {
457 fwmark.netId = netId;
458 fwmark.explicitlySelected = true;
459 fwmark.protectedFromVpn = true;
460 fwmark.permission = permission;
462 char markString[UINT32_HEX_STRLEN];
463 snprintf(markString, sizeof(markString), "0x%x", fwmark.intValue);
465 if (execIptables(V4V6, "-t", "mangle", add ? "-A" : "-D", "INPUT", "-i", interface, "-j",
466 "MARK", "--set-mark", markString, NULL)) {
467 ALOGE("failed to change iptables rule that sets incoming packet mark");
474 // A rule to route responses to the local network forwarded via the VPN.
476 // When a VPN is in effect, packets from the local network to upstream networks are forwarded into
477 // the VPN's tunnel interface. When the VPN forwards the responses, they emerge out of the tunnel.
478 WARN_UNUSED_RESULT int modifyVpnOutputToLocalRule(const char* vpnInterface, bool add) {
479 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_VPN_OUTPUT_TO_LOCAL,
480 ROUTE_TABLE_LOCAL_NETWORK, MARK_UNSET, MARK_UNSET, vpnInterface, OIF_NONE,
481 INVALID_UID, INVALID_UID);
484 // A rule to route all traffic from a given set of UIDs to go over the VPN.
486 // Notice that this rule doesn't use the netId. I.e., no matter what netId the user's socket may
487 // have, if they are subject to this VPN, their traffic has to go through it. Allows the traffic to
488 // bypass the VPN if the protectedFromVpn bit is set.
489 WARN_UNUSED_RESULT int modifyVpnUidRangeRule(uint32_t table, uid_t uidStart, uid_t uidEnd,
490 bool secure, bool add) {
494 fwmark.protectedFromVpn = false;
495 mask.protectedFromVpn = true;
500 priority = RULE_PRIORITY_SECURE_VPN;
502 priority = RULE_PRIORITY_BYPASSABLE_VPN;
504 fwmark.explicitlySelected = false;
505 mask.explicitlySelected = true;
508 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, priority, table, fwmark.intValue,
509 mask.intValue, IIF_LOOPBACK, OIF_NONE, uidStart, uidEnd);
512 // A rule to allow system apps to send traffic over this VPN even if they are not part of the target
515 // This is needed for DnsProxyListener to correctly resolve a request for a user who is in the
516 // target set, but where the DnsProxyListener itself is not.
517 WARN_UNUSED_RESULT int modifyVpnSystemPermissionRule(unsigned netId, uint32_t table, bool secure,
522 fwmark.netId = netId;
523 mask.netId = FWMARK_NET_ID_MASK;
525 fwmark.permission = PERMISSION_SYSTEM;
526 mask.permission = PERMISSION_SYSTEM;
528 uint32_t priority = secure ? RULE_PRIORITY_SECURE_VPN : RULE_PRIORITY_BYPASSABLE_VPN;
530 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, priority, table, fwmark.intValue,
534 // A rule to route traffic based on an explicitly chosen network.
536 // Supports apps that use the multinetwork APIs to restrict their traffic to a network.
538 // Even though we check permissions at the time we set a netId into the fwmark of a socket, we need
539 // to check it again in the rules here, because a network's permissions may have been updated via
540 // modifyNetworkPermission().
541 WARN_UNUSED_RESULT int modifyExplicitNetworkRule(unsigned netId, uint32_t table,
542 Permission permission, uid_t uidStart,
543 uid_t uidEnd, bool add) {
547 fwmark.netId = netId;
548 mask.netId = FWMARK_NET_ID_MASK;
550 fwmark.explicitlySelected = true;
551 mask.explicitlySelected = true;
553 fwmark.permission = permission;
554 mask.permission = permission;
556 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_EXPLICIT_NETWORK, table,
557 fwmark.intValue, mask.intValue, IIF_NONE, OIF_NONE, uidStart, uidEnd);
560 // A rule to route traffic based on a chosen outgoing interface.
562 // Supports apps that use SO_BINDTODEVICE or IP_PKTINFO options and the kernel that already knows
563 // the outgoing interface (typically for link-local communications).
564 WARN_UNUSED_RESULT int modifyOutputInterfaceRules(const char* interface, uint32_t table,
565 Permission permission, uid_t uidStart,
566 uid_t uidEnd, bool add) {
570 fwmark.permission = permission;
571 mask.permission = permission;
573 // If this rule does not specify a UID range, then also add a corresponding high-priority rule
574 // for UID. This covers forwarded packets and system daemons such as the tethering DHCP server.
575 if (uidStart == INVALID_UID && uidEnd == INVALID_UID) {
576 if (int ret = modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_VPN_OVERRIDE_OIF,
577 table, fwmark.intValue, mask.intValue, IIF_NONE, interface,
578 UID_ROOT, UID_ROOT)) {
583 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_OUTPUT_INTERFACE, table,
584 fwmark.intValue, mask.intValue, IIF_NONE, interface, uidStart, uidEnd);
587 // A rule to route traffic based on the chosen network.
589 // This is for sockets that have not explicitly requested a particular network, but have been
590 // bound to one when they called connect(). This ensures that sockets connected on a particular
591 // network stay on that network even if the default network changes.
592 WARN_UNUSED_RESULT int modifyImplicitNetworkRule(unsigned netId, uint32_t table,
593 Permission permission, bool add) {
597 fwmark.netId = netId;
598 mask.netId = FWMARK_NET_ID_MASK;
600 fwmark.explicitlySelected = false;
601 mask.explicitlySelected = true;
603 fwmark.permission = permission;
604 mask.permission = permission;
606 return modifyIpRule(add ? RTM_NEWRULE : RTM_DELRULE, RULE_PRIORITY_IMPLICIT_NETWORK, table,
607 fwmark.intValue, mask.intValue);
610 // A rule to enable split tunnel VPNs.
612 // If a packet with a VPN's netId doesn't find a route in the VPN's routing table, it's allowed to
613 // go over the default network, provided it wasn't explicitly restricted to the VPN and has the
614 // permissions required by the default network.
615 WARN_UNUSED_RESULT int modifyVpnFallthroughRule(uint16_t action, unsigned vpnNetId,
616 const char* physicalInterface,
617 Permission permission) {
618 uint32_t table = getRouteTableForInterface(physicalInterface);
619 if (table == RT_TABLE_UNSPEC) {
626 fwmark.netId = vpnNetId;
627 mask.netId = FWMARK_NET_ID_MASK;
629 fwmark.explicitlySelected = false;
630 mask.explicitlySelected = true;
632 fwmark.permission = permission;
633 mask.permission = permission;
635 return modifyIpRule(action, RULE_PRIORITY_VPN_FALLTHROUGH, table, fwmark.intValue,
639 // Add rules to allow legacy routes added through the requestRouteToHost() API.
640 WARN_UNUSED_RESULT int addLegacyRouteRules() {
644 fwmark.explicitlySelected = false;
645 mask.explicitlySelected = true;
647 // Rules to allow legacy routes to override the default network.
648 if (int ret = modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_LEGACY_SYSTEM, ROUTE_TABLE_LEGACY_SYSTEM,
649 fwmark.intValue, mask.intValue)) {
652 if (int ret = modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_LEGACY_NETWORK,
653 ROUTE_TABLE_LEGACY_NETWORK, fwmark.intValue, mask.intValue)) {
657 fwmark.permission = PERMISSION_SYSTEM;
658 mask.permission = PERMISSION_SYSTEM;
660 // A rule to allow legacy routes from system apps to override VPNs.
661 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_VPN_OVERRIDE_SYSTEM, ROUTE_TABLE_LEGACY_SYSTEM,
662 fwmark.intValue, mask.intValue);
665 // Add rules to lookup the local network when specified explicitly or otherwise.
666 WARN_UNUSED_RESULT int addLocalNetworkRules(unsigned localNetId) {
667 if (int ret = modifyExplicitNetworkRule(localNetId, ROUTE_TABLE_LOCAL_NETWORK, PERMISSION_NONE,
668 INVALID_UID, INVALID_UID, ACTION_ADD)) {
675 fwmark.explicitlySelected = false;
676 mask.explicitlySelected = true;
678 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_LOCAL_NETWORK, ROUTE_TABLE_LOCAL_NETWORK,
679 fwmark.intValue, mask.intValue);
682 int configureDummyNetwork() {
683 const char *interface = DummyNetwork::INTERFACE_NAME;
684 uint32_t table = getRouteTableForInterface(interface);
685 if (table == RT_TABLE_UNSPEC) {
686 // getRouteTableForInterface has already looged an error.
691 int ret = ifc_up(interface);
694 ALOGE("Can't bring up %s: %s", interface, strerror(errno));
698 if ((ret = modifyOutputInterfaceRules(interface, table, PERMISSION_NONE,
699 INVALID_UID, INVALID_UID, ACTION_ADD))) {
700 ALOGE("Can't create oif rules for %s: %s", interface, strerror(-ret));
704 if ((ret = modifyIpRoute(RTM_NEWROUTE, table, interface, "0.0.0.0/0", NULL))) {
705 ALOGE("Can't add IPv4 default route to %s: %s", interface, strerror(-ret));
709 if ((ret = modifyIpRoute(RTM_NEWROUTE, table, interface, "::/0", NULL))) {
710 ALOGE("Can't add IPv6 default route to %s: %s", interface, strerror(-ret));
717 // Add a new rule to look up the 'main' table, with the same selectors as the "default network"
718 // rule, but with a lower priority. We will never create routes in the main table; it should only be
719 // used for directly-connected routes implicitly created by the kernel when adding IP addresses.
720 // This is necessary, for example, when adding a route through a directly-connected gateway: in
721 // order to add the route, there must already be a directly-connected route that covers the gateway.
722 WARN_UNUSED_RESULT int addDirectlyConnectedRule() {
726 fwmark.netId = NETID_UNSET;
727 mask.netId = FWMARK_NET_ID_MASK;
729 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_DIRECTLY_CONNECTED, RT_TABLE_MAIN,
730 fwmark.intValue, mask.intValue, IIF_NONE, OIF_NONE, UID_ROOT, UID_ROOT);
733 // Add an explicit unreachable rule close to the end of the prioriy list to make it clear that
734 // relying on the kernel-default "from all lookup main" rule at priority 32766 is not intended
735 // behaviour. We do flush the kernel-default rules at startup, but having an explicit unreachable
736 // rule will hopefully make things even clearer.
737 WARN_UNUSED_RESULT int addUnreachableRule() {
738 return modifyIpRule(RTM_NEWRULE, RULE_PRIORITY_UNREACHABLE, RT_TABLE_UNSPEC, MARK_UNSET,
742 WARN_UNUSED_RESULT int modifyLocalNetwork(unsigned netId, const char* interface, bool add) {
743 if (int ret = modifyIncomingPacketMark(netId, interface, PERMISSION_NONE, add)) {
746 return modifyOutputInterfaceRules(interface, ROUTE_TABLE_LOCAL_NETWORK, PERMISSION_NONE,
747 INVALID_UID, INVALID_UID, add);
750 WARN_UNUSED_RESULT int modifyPhysicalNetwork(unsigned netId, const char* interface,
751 Permission permission, bool add) {
752 uint32_t table = getRouteTableForInterface(interface);
753 if (table == RT_TABLE_UNSPEC) {
757 if (int ret = modifyIncomingPacketMark(netId, interface, permission, add)) {
760 if (int ret = modifyExplicitNetworkRule(netId, table, permission, INVALID_UID, INVALID_UID,
764 if (int ret = modifyOutputInterfaceRules(interface, table, permission, INVALID_UID, INVALID_UID,
768 return modifyImplicitNetworkRule(netId, table, permission, add);
771 WARN_UNUSED_RESULT int modifyVirtualNetwork(unsigned netId, const char* interface,
772 const UidRanges& uidRanges, bool secure, bool add,
773 bool modifyNonUidBasedRules) {
774 uint32_t table = getRouteTableForInterface(interface);
775 if (table == RT_TABLE_UNSPEC) {
779 for (const UidRanges::Range& range : uidRanges.getRanges()) {
780 if (int ret = modifyVpnUidRangeRule(table, range.first, range.second, secure, add)) {
783 if (int ret = modifyExplicitNetworkRule(netId, table, PERMISSION_NONE, range.first,
784 range.second, add)) {
787 if (int ret = modifyOutputInterfaceRules(interface, table, PERMISSION_NONE, range.first,
788 range.second, add)) {
793 if (modifyNonUidBasedRules) {
794 if (int ret = modifyIncomingPacketMark(netId, interface, PERMISSION_NONE, add)) {
797 if (int ret = modifyVpnOutputToLocalRule(interface, add)) {
800 if (int ret = modifyVpnSystemPermissionRule(netId, table, secure, add)) {
803 return modifyExplicitNetworkRule(netId, table, PERMISSION_NONE, UID_ROOT, UID_ROOT, add);
809 WARN_UNUSED_RESULT int modifyDefaultNetwork(uint16_t action, const char* interface,
810 Permission permission) {
811 uint32_t table = getRouteTableForInterface(interface);
812 if (table == RT_TABLE_UNSPEC) {
819 fwmark.netId = NETID_UNSET;
820 mask.netId = FWMARK_NET_ID_MASK;
822 fwmark.permission = permission;
823 mask.permission = permission;
825 return modifyIpRule(action, RULE_PRIORITY_DEFAULT_NETWORK, table, fwmark.intValue,
829 WARN_UNUSED_RESULT int modifyTetheredNetwork(uint16_t action, const char* inputInterface,
830 const char* outputInterface) {
831 uint32_t table = getRouteTableForInterface(outputInterface);
832 if (table == RT_TABLE_UNSPEC) {
836 return modifyIpRule(action, RULE_PRIORITY_TETHERING, table, MARK_UNSET, MARK_UNSET,
837 inputInterface, OIF_NONE, INVALID_UID, INVALID_UID);
840 // Returns 0 on success or negative errno on failure.
841 WARN_UNUSED_RESULT int flushRules() {
842 for (size_t i = 0; i < ARRAY_SIZE(IP_VERSIONS); ++i) {
843 const char* argv[] = {
849 if (android_fork_execvp(ARRAY_SIZE(argv), const_cast<char**>(argv), NULL, false, false)) {
850 ALOGE("failed to flush rules");
857 // Adds or removes an IPv4 or IPv6 route to the specified table and, if it's a directly-connected
858 // route, to the main table as well.
859 // Returns 0 on success or negative errno on failure.
860 WARN_UNUSED_RESULT int modifyRoute(uint16_t action, const char* interface, const char* destination,
861 const char* nexthop, RouteController::TableType tableType) {
864 case RouteController::INTERFACE: {
865 table = getRouteTableForInterface(interface);
866 if (table == RT_TABLE_UNSPEC) {
871 case RouteController::LOCAL_NETWORK: {
872 table = ROUTE_TABLE_LOCAL_NETWORK;
875 case RouteController::LEGACY_NETWORK: {
876 table = ROUTE_TABLE_LEGACY_NETWORK;
879 case RouteController::LEGACY_SYSTEM: {
880 table = ROUTE_TABLE_LEGACY_SYSTEM;
885 int ret = modifyIpRoute(action, table, interface, destination, nexthop);
886 // Trying to add a route that already exists shouldn't cause an error.
887 if (ret && !(action == RTM_NEWROUTE && ret == -EEXIST)) {
894 // Returns 0 on success or negative errno on failure.
895 WARN_UNUSED_RESULT int flushRoutes(const char* interface) {
896 uint32_t table = getRouteTableForInterface(interface);
897 if (table == RT_TABLE_UNSPEC) {
901 char tableString[UINT32_STRLEN];
902 snprintf(tableString, sizeof(tableString), "%u", table);
905 for (size_t i = 0; i < ARRAY_SIZE(IP_VERSIONS); ++i) {
906 const char* argv[] = {
915 // A flush works by dumping routes and deleting each route as it's returned, and it can
916 // fail if something else deletes the route between the dump and the delete. This can
917 // happen, for example, if an interface goes down while we're trying to flush its routes.
918 // So try multiple times and only return an error if the last attempt fails.
920 // TODO: replace this with our own netlink code.
921 unsigned attempts = 0;
924 err = android_fork_execvp(ARRAY_SIZE(argv), const_cast<char**>(argv),
927 } while (err != 0 && attempts < ROUTE_FLUSH_ATTEMPTS);
929 ALOGE("failed to flush %s routes in table %s after %d attempts",
930 IP_VERSIONS[i], tableString, attempts);
935 // If we failed to flush routes, the caller may elect to keep this interface around, so keep
936 // track of its name.
938 interfaceToTable.erase(interface);
944 WARN_UNUSED_RESULT int clearTetheringRules(const char* inputInterface) {
947 ret = modifyIpRule(RTM_DELRULE, RULE_PRIORITY_TETHERING, 0, MARK_UNSET, MARK_UNSET,
948 inputInterface, OIF_NONE, INVALID_UID, INVALID_UID);
951 if (ret == -ENOENT) {
960 int RouteController::Init(unsigned localNetId) {
961 if (int ret = flushRules()) {
964 if (int ret = addLegacyRouteRules()) {
967 if (int ret = addLocalNetworkRules(localNetId)) {
970 if (int ret = addDirectlyConnectedRule()) {
973 if (int ret = addUnreachableRule()) {
976 // Don't complain if we can't add the dummy network, since not all devices support it.
977 configureDummyNetwork();
979 updateTableNamesFile();
983 int RouteController::addInterfaceToLocalNetwork(unsigned netId, const char* interface) {
984 return modifyLocalNetwork(netId, interface, ACTION_ADD);
987 int RouteController::removeInterfaceFromLocalNetwork(unsigned netId, const char* interface) {
988 return modifyLocalNetwork(netId, interface, ACTION_DEL);
991 int RouteController::addInterfaceToPhysicalNetwork(unsigned netId, const char* interface,
992 Permission permission) {
993 if (int ret = modifyPhysicalNetwork(netId, interface, permission, ACTION_ADD)) {
996 updateTableNamesFile();
1000 int RouteController::removeInterfaceFromPhysicalNetwork(unsigned netId, const char* interface,
1001 Permission permission) {
1002 if (int ret = modifyPhysicalNetwork(netId, interface, permission, ACTION_DEL)) {
1005 if (int ret = flushRoutes(interface)) {
1008 if (int ret = clearTetheringRules(interface)) {
1011 updateTableNamesFile();
1015 int RouteController::addInterfaceToVirtualNetwork(unsigned netId, const char* interface,
1016 bool secure, const UidRanges& uidRanges) {
1017 if (int ret = modifyVirtualNetwork(netId, interface, uidRanges, secure, ACTION_ADD,
1018 MODIFY_NON_UID_BASED_RULES)) {
1021 updateTableNamesFile();
1025 int RouteController::removeInterfaceFromVirtualNetwork(unsigned netId, const char* interface,
1026 bool secure, const UidRanges& uidRanges) {
1027 if (int ret = modifyVirtualNetwork(netId, interface, uidRanges, secure, ACTION_DEL,
1028 MODIFY_NON_UID_BASED_RULES)) {
1031 if (int ret = flushRoutes(interface)) {
1034 updateTableNamesFile();
1038 int RouteController::modifyPhysicalNetworkPermission(unsigned netId, const char* interface,
1039 Permission oldPermission,
1040 Permission newPermission) {
1041 // Add the new rules before deleting the old ones, to avoid race conditions.
1042 if (int ret = modifyPhysicalNetwork(netId, interface, newPermission, ACTION_ADD)) {
1045 return modifyPhysicalNetwork(netId, interface, oldPermission, ACTION_DEL);
1048 int RouteController::addUsersToVirtualNetwork(unsigned netId, const char* interface, bool secure,
1049 const UidRanges& uidRanges) {
1050 return modifyVirtualNetwork(netId, interface, uidRanges, secure, ACTION_ADD,
1051 !MODIFY_NON_UID_BASED_RULES);
1054 int RouteController::removeUsersFromVirtualNetwork(unsigned netId, const char* interface,
1055 bool secure, const UidRanges& uidRanges) {
1056 return modifyVirtualNetwork(netId, interface, uidRanges, secure, ACTION_DEL,
1057 !MODIFY_NON_UID_BASED_RULES);
1060 int RouteController::addInterfaceToDefaultNetwork(const char* interface, Permission permission) {
1061 return modifyDefaultNetwork(RTM_NEWRULE, interface, permission);
1064 int RouteController::removeInterfaceFromDefaultNetwork(const char* interface,
1065 Permission permission) {
1066 return modifyDefaultNetwork(RTM_DELRULE, interface, permission);
1069 int RouteController::addRoute(const char* interface, const char* destination, const char* nexthop,
1070 TableType tableType) {
1071 return modifyRoute(RTM_NEWROUTE, interface, destination, nexthop, tableType);
1074 int RouteController::removeRoute(const char* interface, const char* destination,
1075 const char* nexthop, TableType tableType) {
1076 return modifyRoute(RTM_DELROUTE, interface, destination, nexthop, tableType);
1079 int RouteController::enableTethering(const char* inputInterface, const char* outputInterface) {
1080 return modifyTetheredNetwork(RTM_NEWRULE, inputInterface, outputInterface);
1083 int RouteController::disableTethering(const char* inputInterface, const char* outputInterface) {
1084 return modifyTetheredNetwork(RTM_DELRULE, inputInterface, outputInterface);
1087 int RouteController::addVirtualNetworkFallthrough(unsigned vpnNetId, const char* physicalInterface,
1088 Permission permission) {
1089 return modifyVpnFallthroughRule(RTM_NEWRULE, vpnNetId, physicalInterface, permission);
1092 int RouteController::removeVirtualNetworkFallthrough(unsigned vpnNetId,
1093 const char* physicalInterface,
1094 Permission permission) {
1095 return modifyVpnFallthroughRule(RTM_DELRULE, vpnNetId, physicalInterface, permission);