1/*
2 * Copyright (C) 2008 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 *      http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17#include <stdlib.h>
18#include <errno.h>
19#include <fcntl.h>
20#include <string.h>
21
22#include <sys/socket.h>
23#include <sys/stat.h>
24#include <sys/types.h>
25#include <sys/wait.h>
26
27#include <netinet/in.h>
28#include <arpa/inet.h>
29
30#define LOG_TAG "SecondaryTablController"
31#include <cutils/log.h>
32#include <cutils/properties.h>
33#include <logwrap/logwrap.h>
34
35#include "ResponseCode.h"
36#include "NetdConstants.h"
37#include "SecondaryTableController.h"
38
39const char* SecondaryTableController::LOCAL_MANGLE_OUTPUT = "st_mangle_OUTPUT";
40const char* SecondaryTableController::LOCAL_MANGLE_POSTROUTING = "st_mangle_POSTROUTING";
41const char* SecondaryTableController::LOCAL_NAT_POSTROUTING = "st_nat_POSTROUTING";
42
43SecondaryTableController::SecondaryTableController(UidMarkMap *map) : mUidMarkMap(map) {
44    int i;
45    for (i=0; i < INTERFACES_TRACKED; i++) {
46        mInterfaceTable[i][0] = 0;
47        // TODO - use a hashtable or other prebuilt container class
48        mInterfaceRuleCount[i] = 0;
49    }
50}
51
52SecondaryTableController::~SecondaryTableController() {
53}
54
55int SecondaryTableController::setupIptablesHooks() {
56    int res = execIptables(V4V6,
57            "-t",
58            "mangle",
59            "-F",
60            LOCAL_MANGLE_OUTPUT,
61            NULL);
62    // Do not mark sockets that have already been marked elsewhere(for example in DNS or protect).
63    res |= execIptables(V4V6,
64            "-t",
65            "mangle",
66            "-A",
67            LOCAL_MANGLE_OUTPUT,
68            "-m",
69            "mark",
70            "!",
71            "--mark",
72            "0",
73            "-j",
74            "RETURN",
75            NULL);
76
77    // protect the legacy VPN daemons from routes.
78    // TODO: Remove this when legacy VPN's are removed.
79    res |= execIptables(V4V6,
80            "-t",
81            "mangle",
82            "-A",
83            LOCAL_MANGLE_OUTPUT,
84            "-m",
85            "owner",
86            "--uid-owner",
87            "vpn",
88            "-j",
89            "RETURN",
90            NULL);
91    return res;
92}
93
94int SecondaryTableController::findTableNumber(const char *iface) {
95    int i;
96    for (i = 0; i < INTERFACES_TRACKED; i++) {
97        // compare through the final null, hence +1
98        if (strncmp(iface, mInterfaceTable[i], IFNAMSIZ + 1) == 0) {
99            return i;
100        }
101    }
102    return -1;
103}
104
105int SecondaryTableController::addRoute(SocketClient *cli, char *iface, char *dest, int prefix,
106        char *gateway) {
107    int tableIndex = findTableNumber(iface);
108    if (tableIndex == -1) {
109        tableIndex = findTableNumber(""); // look for an empty slot
110        if (tableIndex == -1) {
111            ALOGE("Max number of NATed interfaces reached");
112            errno = ENODEV;
113            cli->sendMsg(ResponseCode::OperationFailed, "Max number NATed", true);
114            return -1;
115        }
116        strncpy(mInterfaceTable[tableIndex], iface, IFNAMSIZ);
117        // Ensure null termination even if truncation happened
118        mInterfaceTable[tableIndex][IFNAMSIZ] = 0;
119    }
120
121    return modifyRoute(cli, ADD, iface, dest, prefix, gateway, tableIndex);
122}
123
124int SecondaryTableController::modifyRoute(SocketClient *cli, const char *action, char *iface,
125        char *dest, int prefix, char *gateway, int tableIndex) {
126    char dest_str[44]; // enough to store an IPv6 address + 3 character bitmask
127    char tableIndex_str[11];
128    int ret;
129
130    //  IP tool doesn't like "::" - the equiv of 0.0.0.0 that it accepts for ipv4
131    snprintf(dest_str, sizeof(dest_str), "%s/%d", dest, prefix);
132    snprintf(tableIndex_str, sizeof(tableIndex_str), "%d", tableIndex + BASE_TABLE_NUMBER);
133
134    if (strcmp("::", gateway) == 0) {
135        const char *cmd[] = {
136                IP_PATH,
137                "route",
138                action,
139                dest_str,
140                "dev",
141                iface,
142                "table",
143                tableIndex_str
144        };
145        ret = runCmd(ARRAY_SIZE(cmd), cmd);
146    } else {
147        const char *cmd[] = {
148                IP_PATH,
149                "route",
150                action,
151                dest_str,
152                "via",
153                gateway,
154                "dev",
155                iface,
156                "table",
157                tableIndex_str
158        };
159        ret = runCmd(ARRAY_SIZE(cmd), cmd);
160    }
161
162    if (ret) {
163        ALOGE("ip route %s failed: %s route %s %s/%d via %s dev %s table %d", action,
164                IP_PATH, action, dest, prefix, gateway, iface, tableIndex+BASE_TABLE_NUMBER);
165        errno = ENODEV;
166        cli->sendMsg(ResponseCode::OperationFailed, "ip route modification failed", true);
167        return -1;
168    }
169
170    if (strcmp(action, ADD) == 0) {
171        mInterfaceRuleCount[tableIndex]++;
172    } else {
173        if (--mInterfaceRuleCount[tableIndex] < 1) {
174            mInterfaceRuleCount[tableIndex] = 0;
175            mInterfaceTable[tableIndex][0] = 0;
176        }
177    }
178    modifyRuleCount(tableIndex, action);
179    cli->sendMsg(ResponseCode::CommandOkay, "Route modified", false);
180    return 0;
181}
182
183void SecondaryTableController::modifyRuleCount(int tableIndex, const char *action) {
184    if (strcmp(action, ADD) == 0) {
185        mInterfaceRuleCount[tableIndex]++;
186    } else {
187        if (--mInterfaceRuleCount[tableIndex] < 1) {
188            mInterfaceRuleCount[tableIndex] = 0;
189            mInterfaceTable[tableIndex][0] = 0;
190        }
191    }
192}
193
194int SecondaryTableController::verifyTableIndex(int tableIndex) {
195    if ((tableIndex < 0) ||
196            (tableIndex >= INTERFACES_TRACKED) ||
197            (mInterfaceTable[tableIndex][0] == 0)) {
198        return -1;
199    } else {
200        return 0;
201    }
202}
203
204const char *SecondaryTableController::getVersion(const char *addr) {
205    if (strchr(addr, ':') != NULL) {
206        return "-6";
207    } else {
208        return "-4";
209    }
210}
211
212IptablesTarget SecondaryTableController::getIptablesTarget(const char *addr) {
213    if (strchr(addr, ':') != NULL) {
214        return V6;
215    } else {
216        return V4;
217    }
218}
219
220int SecondaryTableController::removeRoute(SocketClient *cli, char *iface, char *dest, int prefix,
221        char *gateway) {
222    int tableIndex = findTableNumber(iface);
223    if (tableIndex == -1) {
224        ALOGE("Interface not found");
225        errno = ENODEV;
226        cli->sendMsg(ResponseCode::OperationFailed, "Interface not found", true);
227        return -1;
228    }
229
230    return modifyRoute(cli, DEL, iface, dest, prefix, gateway, tableIndex);
231}
232
233int SecondaryTableController::modifyFromRule(int tableIndex, const char *action,
234        const char *addr) {
235    char tableIndex_str[11];
236
237    if (verifyTableIndex(tableIndex)) {
238        return -1;
239    }
240
241    snprintf(tableIndex_str, sizeof(tableIndex_str), "%d", tableIndex +
242            BASE_TABLE_NUMBER);
243    const char *cmd[] = {
244            IP_PATH,
245            getVersion(addr),
246            "rule",
247            action,
248            "from",
249            addr,
250            "table",
251            tableIndex_str
252    };
253    if (runCmd(ARRAY_SIZE(cmd), cmd)) {
254        return -1;
255    }
256
257    modifyRuleCount(tableIndex, action);
258    return 0;
259}
260
261int SecondaryTableController::modifyLocalRoute(int tableIndex, const char *action,
262        const char *iface, const char *addr) {
263    char tableIndex_str[11];
264
265    if (verifyTableIndex(tableIndex)) {
266        return -1;
267    }
268
269    modifyRuleCount(tableIndex, action); // some del's will fail as the iface is already gone.
270
271    snprintf(tableIndex_str, sizeof(tableIndex_str), "%d", tableIndex +
272            BASE_TABLE_NUMBER);
273    const char *cmd[] = {
274            IP_PATH,
275            "route",
276            action,
277            addr,
278            "dev",
279            iface,
280            "table",
281            tableIndex_str
282    };
283
284    return runCmd(ARRAY_SIZE(cmd), cmd);
285}
286int SecondaryTableController::addFwmarkRule(const char *iface) {
287    return setFwmarkRule(iface, true);
288}
289
290int SecondaryTableController::removeFwmarkRule(const char *iface) {
291    return setFwmarkRule(iface, false);
292}
293
294int SecondaryTableController::setFwmarkRule(const char *iface, bool add) {
295    int tableIndex = findTableNumber(iface);
296    if (tableIndex == -1) {
297        tableIndex = findTableNumber(""); // look for an empty slot
298        if (tableIndex == -1) {
299            ALOGE("Max number of NATed interfaces reached");
300            errno = ENODEV;
301            return -1;
302        }
303        strncpy(mInterfaceTable[tableIndex], iface, IFNAMSIZ);
304        // Ensure null termination even if truncation happened
305        mInterfaceTable[tableIndex][IFNAMSIZ] = 0;
306    }
307    int mark = tableIndex + BASE_TABLE_NUMBER;
308    char mark_str[11];
309    int ret;
310
311    //fail fast if any rules already exist for this interface
312    if (mUidMarkMap->anyRulesForMark(mark)) {
313        errno = EBUSY;
314        return -1;
315    }
316
317    snprintf(mark_str, sizeof(mark_str), "%d", mark);
318    // Flush any marked routes we added
319    if (!add) {
320        // iproute2 rule del will delete anything that matches, but only one rule at a time.
321        // So clearing the rules requires a bunch of calls.
322        // ip rule del will fail once there are no remaining rules that match.
323        const char *v4_cmd[] = {
324            IP_PATH,
325            "-4",
326            "rule",
327            "del",
328            "fwmark",
329            mark_str,
330            "table",
331            mark_str
332        };
333        while(!runCmd(ARRAY_SIZE(v4_cmd), v4_cmd)) {}
334
335        const char *v6_cmd[] = {
336            IP_PATH,
337            "-6",
338            "rule",
339            "del",
340            "fwmark",
341            mark_str,
342            "table",
343            mark_str
344        };
345        while(!runCmd(ARRAY_SIZE(v6_cmd), v6_cmd)) {}
346    }
347    // Add a route to the table to send all traffic to iface.
348    // We only need a default route because this table is only selected if a packet matches an
349    // IP rule that checks both the route and the mark.
350    const char *route_cmd[] = {
351        IP_PATH,
352        "route",
353        add ? "add" : "del",
354        "default",
355        "dev",
356        iface,
357        "table",
358        mark_str
359    };
360    ret = runCmd(ARRAY_SIZE(route_cmd), route_cmd);
361    // The command might fail during delete if the iface is gone
362    if (add && ret) return ret;
363
364    // As above for IPv6
365    const char *route6_cmd[] = {
366        IP_PATH,
367        "-6",
368        "route",
369        add ? "add" : "del",
370        "default",
371        "dev",
372        iface,
373        "table",
374        mark_str
375    };
376    ret = runCmd(ARRAY_SIZE(route6_cmd), route6_cmd);
377    // The command might fail during delete if the iface is gone
378    if (add && ret) return ret;
379
380    /* Best effort, because some kernels might not have the needed TCPMSS */
381    execIptables(V4V6,
382            "-t",
383            "mangle",
384            add ? "-A" : "-D",
385            LOCAL_MANGLE_POSTROUTING,
386            "-p", "tcp", "-o", iface, "--tcp-flags", "SYN,RST", "SYN",
387            "-j",
388            "TCPMSS",
389            "--clamp-mss-to-pmtu",
390            NULL);
391
392    // Because the mark gets set after the intial routing decision the source IP address is that
393    // of the original out interface. The only way to change the source IP address to that of the
394    // VPN iface is using source NAT.
395    // TODO: Remove this when we get the mark set correctly before the first routing pass.
396    ret = execIptables(V4,
397            "-t",
398            "nat",
399            add ? "-A" : "-D",
400            LOCAL_NAT_POSTROUTING,
401            "-o",
402            iface,
403            "-m",
404            "mark",
405            "--mark",
406            mark_str,
407            "-j",
408            "MASQUERADE",
409            NULL);
410
411    if (ret) return ret;
412
413    // Try and set up NAT for IPv6 as well. This was only added in Linux 3.7 so this may fail.
414    ret = execIptables(V6,
415            "-t",
416            "nat",
417            add ? "-A" : "-D",
418            LOCAL_NAT_POSTROUTING,
419            "-o",
420            iface,
421            "-m",
422            "mark",
423            "--mark",
424            mark_str,
425            "-j",
426            "MASQUERADE",
427            NULL);
428    if (ret) {
429        // Without V6 NAT we can't do V6 over VPNs. If an IPv6 packet matches a VPN rule, then it
430        // will go out on the VPN interface, but without NAT, it will have the wrong source
431        // address. So reject all these packets.
432        // Due to rule application by the time the connection hits the output filter chain the
433        // routing pass based on the new mark has not yet happened. Reject in ip instead.
434        // TODO: Make the VPN code refuse to install IPv6 routes until we don't need IPv6 NAT.
435        const char *reject_cmd[] = {
436            IP_PATH,
437            "-6",
438            "route",
439            add ? "replace" : "del",
440            "unreachable",
441            "default",
442            "table",
443            mark_str
444        };
445        ret = runCmd(ARRAY_SIZE(reject_cmd), reject_cmd);
446        // The command might fail during delete if the iface is gone
447        if (add && ret) return ret;
448
449    }
450    return 0;
451
452}
453
454int SecondaryTableController::addFwmarkRoute(const char* iface, const char *dest, int prefix) {
455    return setFwmarkRoute(iface, dest, prefix, true);
456}
457
458int SecondaryTableController::removeFwmarkRoute(const char* iface, const char *dest, int prefix) {
459    return setFwmarkRoute(iface, dest, prefix, true);
460}
461
462int SecondaryTableController::setFwmarkRoute(const char* iface, const char *dest, int prefix,
463                                             bool add) {
464    int tableIndex = findTableNumber(iface);
465    if (tableIndex == -1) {
466        errno = EINVAL;
467        return -1;
468    }
469    int mark = tableIndex + BASE_TABLE_NUMBER;
470    char mark_str[11] = {0};
471    char dest_str[44]; // enough to store an IPv6 address + 3 character bitmask
472
473    snprintf(mark_str, sizeof(mark_str), "%d", mark);
474    snprintf(dest_str, sizeof(dest_str), "%s/%d", dest, prefix);
475    const char *rule_cmd[] = {
476        IP_PATH,
477        getVersion(dest_str),
478        "rule",
479        add ? "add" : "del",
480        "prio",
481        RULE_PRIO,
482        "to",
483        dest_str,
484        "fwmark",
485        mark_str,
486        "table",
487        mark_str
488    };
489    return runCmd(ARRAY_SIZE(rule_cmd), rule_cmd);
490}
491
492int SecondaryTableController::addUidRule(const char *iface, int uid_start, int uid_end) {
493    return setUidRule(iface, uid_start, uid_end, true);
494}
495
496int SecondaryTableController::removeUidRule(const char *iface, int uid_start, int uid_end) {
497    return setUidRule(iface, uid_start, uid_end, false);
498}
499
500int SecondaryTableController::setUidRule(const char *iface, int uid_start, int uid_end, bool add) {
501    int tableIndex = findTableNumber(iface);
502    if (tableIndex == -1) {
503        errno = EINVAL;
504        return -1;
505    }
506    int mark = tableIndex + BASE_TABLE_NUMBER;
507    if (add) {
508        if (!mUidMarkMap->add(uid_start, uid_end, mark)) {
509            errno = EINVAL;
510            return -1;
511        }
512    } else {
513        if (!mUidMarkMap->remove(uid_start, uid_end, mark)) {
514            errno = EINVAL;
515            return -1;
516        }
517    }
518    char uid_str[24] = {0};
519    snprintf(uid_str, sizeof(uid_str), "%d-%d", uid_start, uid_end);
520    char mark_str[11] = {0};
521    snprintf(mark_str, sizeof(mark_str), "%d", mark);
522    return execIptables(V4V6,
523            "-t",
524            "mangle",
525            add ? "-A" : "-D",
526            LOCAL_MANGLE_OUTPUT,
527            "-m",
528            "owner",
529            "--uid-owner",
530            uid_str,
531            "-j",
532            "MARK",
533            "--set-mark",
534            mark_str,
535            NULL);
536}
537
538int SecondaryTableController::addHostExemption(const char *host) {
539    return setHostExemption(host, true);
540}
541
542int SecondaryTableController::removeHostExemption(const char *host) {
543    return setHostExemption(host, false);
544}
545
546int SecondaryTableController::setHostExemption(const char *host, bool add) {
547    const char *cmd[] = {
548        IP_PATH,
549        getVersion(host),
550        "rule",
551        add ? "add" : "del",
552        "prio",
553        EXEMPT_PRIO,
554        "to",
555        host,
556        "table",
557        "main"
558    };
559    return runCmd(ARRAY_SIZE(cmd), cmd);
560}
561
562void SecondaryTableController::getUidMark(SocketClient *cli, int uid) {
563    int mark = mUidMarkMap->getMark(uid);
564    char mark_str[11];
565    snprintf(mark_str, sizeof(mark_str), "%d", mark);
566    cli->sendMsg(ResponseCode::GetMarkResult, mark_str, false);
567}
568
569void SecondaryTableController::getProtectMark(SocketClient *cli) {
570    char protect_mark_str[11];
571    snprintf(protect_mark_str, sizeof(protect_mark_str), "%d", PROTECT_MARK);
572    cli->sendMsg(ResponseCode::GetMarkResult, protect_mark_str, false);
573}
574
575int SecondaryTableController::runCmd(int argc, const char **argv) {
576    int ret = 0;
577
578    ret = android_fork_execvp(argc, (char **)argv, NULL, false, false);
579    return ret;
580}
581