]>
Commit | Line | Data |
---|---|---|
718e3744 | 1 | /* Kernel communication using netlink interface. |
2 | * Copyright (C) 1999 Kunihiro Ishiguro | |
3 | * | |
4 | * This file is part of GNU Zebra. | |
5 | * | |
6 | * GNU Zebra is free software; you can redistribute it and/or modify it | |
7 | * under the terms of the GNU General Public License as published by the | |
8 | * Free Software Foundation; either version 2, or (at your option) any | |
9 | * later version. | |
10 | * | |
11 | * GNU Zebra is distributed in the hope that it will be useful, but | |
12 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | * General Public License for more details. | |
15 | * | |
896014f4 DL |
16 | * You should have received a copy of the GNU General Public License along |
17 | * with this program; see the file COPYING; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
718e3744 | 19 | */ |
1fdc9eae | 20 | |
21 | #include <zebra.h> | |
22 | ||
81a2f870 SW |
23 | #if defined(HANDLE_ZAPI_FUZZING) |
24 | #include <stdio.h> | |
25 | #include <string.h> | |
26 | #endif | |
27 | ||
ddfeb486 DL |
28 | #ifdef HAVE_NETLINK |
29 | ||
1fdc9eae | 30 | #include "linklist.h" |
31 | #include "if.h" | |
32 | #include "log.h" | |
33 | #include "prefix.h" | |
34 | #include "connected.h" | |
35 | #include "table.h" | |
36 | #include "memory.h" | |
37 | #include "zebra_memory.h" | |
38 | #include "rib.h" | |
39 | #include "thread.h" | |
40 | #include "privs.h" | |
41 | #include "nexthop.h" | |
42 | #include "vrf.h" | |
43 | #include "mpls.h" | |
44 | ||
45 | #include "zebra/zserv.h" | |
46 | #include "zebra/zebra_ns.h" | |
47 | #include "zebra/zebra_vrf.h" | |
05f7f5db | 48 | #include "zebra/rt.h" |
1fdc9eae | 49 | #include "zebra/debug.h" |
50 | #include "zebra/kernel_netlink.h" | |
51 | #include "zebra/rt_netlink.h" | |
52 | #include "zebra/if_netlink.h" | |
942bf97b | 53 | #include "zebra/rule_netlink.h" |
1fdc9eae | 54 | |
55 | #ifndef SO_RCVBUFFORCE | |
56 | #define SO_RCVBUFFORCE (33) | |
57 | #endif | |
58 | ||
59 | /* Hack for GNU libc version 2. */ | |
60 | #ifndef MSG_TRUNC | |
61 | #define MSG_TRUNC 0x20 | |
62 | #endif /* MSG_TRUNC */ | |
63 | ||
64 | #ifndef NLMSG_TAIL | |
d62a17ae | 65 | #define NLMSG_TAIL(nmsg) \ |
d7c0a89a QY |
66 | ((struct rtattr *)(((uint8_t *)(nmsg)) \ |
67 | + NLMSG_ALIGN((nmsg)->nlmsg_len))) | |
1fdc9eae | 68 | #endif |
69 | ||
70 | #ifndef RTA_TAIL | |
d62a17ae | 71 | #define RTA_TAIL(rta) \ |
d7c0a89a | 72 | ((struct rtattr *)(((uint8_t *)(rta)) + RTA_ALIGN((rta)->rta_len))) |
1fdc9eae | 73 | #endif |
74 | ||
f909c673 DS |
75 | #ifndef RTNL_FAMILY_IP6MR |
76 | #define RTNL_FAMILY_IP6MR 129 | |
77 | #endif | |
78 | ||
79 | #ifndef RTPROT_MROUTED | |
80 | #define RTPROT_MROUTED 17 | |
81 | #endif | |
82 | ||
d62a17ae | 83 | static const struct message nlmsg_str[] = {{RTM_NEWROUTE, "RTM_NEWROUTE"}, |
84 | {RTM_DELROUTE, "RTM_DELROUTE"}, | |
85 | {RTM_GETROUTE, "RTM_GETROUTE"}, | |
86 | {RTM_NEWLINK, "RTM_NEWLINK"}, | |
87 | {RTM_DELLINK, "RTM_DELLINK"}, | |
88 | {RTM_GETLINK, "RTM_GETLINK"}, | |
89 | {RTM_NEWADDR, "RTM_NEWADDR"}, | |
90 | {RTM_DELADDR, "RTM_DELADDR"}, | |
91 | {RTM_GETADDR, "RTM_GETADDR"}, | |
92 | {RTM_NEWNEIGH, "RTM_NEWNEIGH"}, | |
93 | {RTM_DELNEIGH, "RTM_DELNEIGH"}, | |
94 | {RTM_GETNEIGH, "RTM_GETNEIGH"}, | |
942bf97b | 95 | {RTM_NEWRULE, "RTM_NEWRULE"}, |
96 | {RTM_DELRULE, "RTM_DELRULE"}, | |
97 | {RTM_GETRULE, "RTM_GETRULE"}, | |
d62a17ae | 98 | {0}}; |
1fdc9eae | 99 | |
100 | static const struct message rtproto_str[] = { | |
d62a17ae | 101 | {RTPROT_REDIRECT, "redirect"}, |
102 | {RTPROT_KERNEL, "kernel"}, | |
103 | {RTPROT_BOOT, "boot"}, | |
104 | {RTPROT_STATIC, "static"}, | |
105 | {RTPROT_GATED, "GateD"}, | |
106 | {RTPROT_RA, "router advertisement"}, | |
107 | {RTPROT_MRT, "MRT"}, | |
108 | {RTPROT_ZEBRA, "Zebra"}, | |
1fdc9eae | 109 | #ifdef RTPROT_BIRD |
d62a17ae | 110 | {RTPROT_BIRD, "BIRD"}, |
1fdc9eae | 111 | #endif /* RTPROT_BIRD */ |
d62a17ae | 112 | {RTPROT_MROUTED, "mroute"}, |
113 | {RTPROT_BGP, "BGP"}, | |
114 | {RTPROT_OSPF, "OSPF"}, | |
115 | {RTPROT_ISIS, "IS-IS"}, | |
116 | {RTPROT_RIP, "RIP"}, | |
117 | {RTPROT_RIPNG, "RIPNG"}, | |
d4d71f11 | 118 | {RTPROT_ZSTATIC, "static"}, |
d62a17ae | 119 | {0}}; |
120 | ||
121 | static const struct message family_str[] = {{AF_INET, "ipv4"}, | |
122 | {AF_INET6, "ipv6"}, | |
123 | {AF_BRIDGE, "bridge"}, | |
124 | {RTNL_FAMILY_IPMR, "ipv4MR"}, | |
125 | {RTNL_FAMILY_IP6MR, "ipv6MR"}, | |
126 | {0}}; | |
127 | ||
128 | static const struct message rttype_str[] = {{RTN_UNICAST, "unicast"}, | |
129 | {RTN_MULTICAST, "multicast"}, | |
130 | {0}}; | |
b339bde7 | 131 | |
1fdc9eae | 132 | extern struct thread_master *master; |
d7c0a89a | 133 | extern uint32_t nl_rcvbufsize; |
1fdc9eae | 134 | |
135 | extern struct zebra_privs_t zserv_privs; | |
136 | ||
2414abd3 | 137 | int netlink_talk_filter(struct nlmsghdr *h, ns_id_t ns_id, int startup) |
1fdc9eae | 138 | { |
3575d9e8 DS |
139 | /* |
140 | * This is an error condition that must be handled during | |
141 | * development. | |
142 | * | |
143 | * The netlink_talk_filter function is used for communication | |
144 | * down the netlink_cmd pipe and we are expecting | |
145 | * an ack being received. So if we get here | |
146 | * then we did not receive the ack and instead | |
147 | * received some other message in an unexpected | |
148 | * way. | |
149 | */ | |
150 | zlog_err("%s: ignoring message type 0x%04x(%s) NS %u", | |
151 | __PRETTY_FUNCTION__, h->nlmsg_type, | |
152 | nl_msg_type_to_str(h->nlmsg_type), ns_id); | |
d62a17ae | 153 | return 0; |
1fdc9eae | 154 | } |
155 | ||
d62a17ae | 156 | static int netlink_recvbuf(struct nlsock *nl, uint32_t newsize) |
1fdc9eae | 157 | { |
d7c0a89a | 158 | uint32_t oldsize; |
d62a17ae | 159 | socklen_t newlen = sizeof(newsize); |
160 | socklen_t oldlen = sizeof(oldsize); | |
161 | int ret; | |
162 | ||
163 | ret = getsockopt(nl->sock, SOL_SOCKET, SO_RCVBUF, &oldsize, &oldlen); | |
164 | if (ret < 0) { | |
165 | zlog_err("Can't get %s receive buffer size: %s", nl->name, | |
166 | safe_strerror(errno)); | |
167 | return -1; | |
168 | } | |
169 | ||
170 | /* Try force option (linux >= 2.6.14) and fall back to normal set */ | |
171 | if (zserv_privs.change(ZPRIVS_RAISE)) | |
172 | zlog_err("routing_socket: Can't raise privileges"); | |
173 | ret = setsockopt(nl->sock, SOL_SOCKET, SO_RCVBUFFORCE, &nl_rcvbufsize, | |
174 | sizeof(nl_rcvbufsize)); | |
175 | if (zserv_privs.change(ZPRIVS_LOWER)) | |
176 | zlog_err("routing_socket: Can't lower privileges"); | |
177 | if (ret < 0) | |
178 | ret = setsockopt(nl->sock, SOL_SOCKET, SO_RCVBUF, | |
179 | &nl_rcvbufsize, sizeof(nl_rcvbufsize)); | |
180 | if (ret < 0) { | |
181 | zlog_err("Can't set %s receive buffer size: %s", nl->name, | |
182 | safe_strerror(errno)); | |
183 | return -1; | |
184 | } | |
185 | ||
186 | ret = getsockopt(nl->sock, SOL_SOCKET, SO_RCVBUF, &newsize, &newlen); | |
187 | if (ret < 0) { | |
188 | zlog_err("Can't get %s receive buffer size: %s", nl->name, | |
189 | safe_strerror(errno)); | |
190 | return -1; | |
191 | } | |
192 | ||
193 | zlog_info("Setting netlink socket receive buffer size: %u -> %u", | |
194 | oldsize, newsize); | |
195 | return 0; | |
1fdc9eae | 196 | } |
197 | ||
198 | /* Make socket for Linux netlink interface. */ | |
d62a17ae | 199 | static int netlink_socket(struct nlsock *nl, unsigned long groups, |
200 | ns_id_t ns_id) | |
1fdc9eae | 201 | { |
d62a17ae | 202 | int ret; |
203 | struct sockaddr_nl snl; | |
204 | int sock; | |
205 | int namelen; | |
206 | int save_errno; | |
207 | ||
208 | if (zserv_privs.change(ZPRIVS_RAISE)) { | |
209 | zlog_err("Can't raise privileges"); | |
210 | return -1; | |
211 | } | |
212 | ||
fe533c56 | 213 | sock = ns_socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE, ns_id); |
d62a17ae | 214 | if (sock < 0) { |
215 | zlog_err("Can't open %s socket: %s", nl->name, | |
216 | safe_strerror(errno)); | |
217 | return -1; | |
218 | } | |
219 | ||
220 | memset(&snl, 0, sizeof snl); | |
221 | snl.nl_family = AF_NETLINK; | |
222 | snl.nl_groups = groups; | |
223 | ||
224 | /* Bind the socket to the netlink structure for anything. */ | |
225 | ret = bind(sock, (struct sockaddr *)&snl, sizeof snl); | |
226 | save_errno = errno; | |
227 | if (zserv_privs.change(ZPRIVS_LOWER)) | |
228 | zlog_err("Can't lower privileges"); | |
229 | ||
230 | if (ret < 0) { | |
231 | zlog_err("Can't bind %s socket to group 0x%x: %s", nl->name, | |
232 | snl.nl_groups, safe_strerror(save_errno)); | |
233 | close(sock); | |
234 | return -1; | |
235 | } | |
236 | ||
237 | /* multiple netlink sockets will have different nl_pid */ | |
238 | namelen = sizeof snl; | |
239 | ret = getsockname(sock, (struct sockaddr *)&snl, (socklen_t *)&namelen); | |
240 | if (ret < 0 || namelen != sizeof snl) { | |
241 | zlog_err("Can't get %s socket name: %s", nl->name, | |
242 | safe_strerror(errno)); | |
243 | close(sock); | |
244 | return -1; | |
245 | } | |
246 | ||
247 | nl->snl = snl; | |
248 | nl->sock = sock; | |
249 | return ret; | |
1fdc9eae | 250 | } |
251 | ||
2414abd3 | 252 | static int netlink_information_fetch(struct nlmsghdr *h, ns_id_t ns_id, |
d62a17ae | 253 | int startup) |
1fdc9eae | 254 | { |
3575d9e8 DS |
255 | /* |
256 | * When we handle new message types here | |
257 | * because we are starting to install them | |
258 | * then lets check the netlink_install_filter | |
259 | * and see if we should add the corresponding | |
260 | * allow through entry there. | |
261 | * Probably not needed to do but please | |
262 | * think about it. | |
263 | */ | |
d62a17ae | 264 | switch (h->nlmsg_type) { |
265 | case RTM_NEWROUTE: | |
2414abd3 | 266 | return netlink_route_change(h, ns_id, startup); |
d62a17ae | 267 | case RTM_DELROUTE: |
2414abd3 | 268 | return netlink_route_change(h, ns_id, startup); |
d62a17ae | 269 | case RTM_NEWLINK: |
2414abd3 | 270 | return netlink_link_change(h, ns_id, startup); |
d62a17ae | 271 | case RTM_DELLINK: |
2414abd3 | 272 | return netlink_link_change(h, ns_id, startup); |
d62a17ae | 273 | case RTM_NEWADDR: |
2414abd3 | 274 | return netlink_interface_addr(h, ns_id, startup); |
d62a17ae | 275 | case RTM_DELADDR: |
2414abd3 | 276 | return netlink_interface_addr(h, ns_id, startup); |
d62a17ae | 277 | case RTM_NEWNEIGH: |
2414abd3 | 278 | return netlink_neigh_change(h, ns_id); |
d62a17ae | 279 | case RTM_DELNEIGH: |
2414abd3 | 280 | return netlink_neigh_change(h, ns_id); |
942bf97b | 281 | case RTM_NEWRULE: |
2414abd3 | 282 | return netlink_rule_change(h, ns_id, startup); |
942bf97b | 283 | case RTM_DELRULE: |
2414abd3 | 284 | return netlink_rule_change(h, ns_id, startup); |
d62a17ae | 285 | default: |
3575d9e8 DS |
286 | /* |
287 | * If we have received this message then | |
288 | * we have made a mistake during development | |
289 | * and we need to write some code to handle | |
290 | * this message type or not ask for | |
291 | * it to be sent up to us | |
292 | */ | |
293 | zlog_err("Unknown netlink nlmsg_type %s(%d) vrf %u\n", | |
294 | nl_msg_type_to_str(h->nlmsg_type), h->nlmsg_type, | |
295 | ns_id); | |
d62a17ae | 296 | break; |
297 | } | |
298 | return 0; | |
1fdc9eae | 299 | } |
300 | ||
81a2f870 SW |
301 | #if defined(HANDLE_ZAPI_FUZZING) |
302 | /* Using globals here to avoid adding function parameters */ | |
303 | ||
304 | /* Keep distinct filenames for netlink fuzzy collection */ | |
305 | static unsigned int netlink_file_counter = 1; | |
306 | ||
307 | /* File name to read fuzzed netlink from */ | |
308 | static char netlink_fuzz_file[MAXPATHLEN] = ""; | |
309 | ||
310 | /* Flag for whether to read from file or not */ | |
311 | static int netlink_read = 0; | |
312 | ||
313 | /** | |
314 | * netlink_set_read() - Sets the read flag | |
315 | * @flag: Flag value. | |
316 | * | |
317 | */ | |
318 | void set_netlink_read(int flag) | |
319 | { | |
320 | netlink_read = flag; | |
321 | } | |
322 | ||
323 | /** | |
324 | * netlink_read_init() - Starts the message parser | |
325 | * @fname: Filename to read. | |
326 | */ | |
327 | void netlink_read_init(const char *fname) | |
328 | { | |
329 | snprintf(netlink_fuzz_file, MAXPATHLEN, "%s", fname); | |
330 | /* Creating this fake socket for testing purposes */ | |
331 | struct zebra_ns zns = { | |
332 | .ns_id = 0, | |
333 | .netlink_cmd = {.sock = -1, | |
334 | .seq = -1, | |
335 | .name = "fuzzer_command_channel"}, | |
336 | .netlink = {.sock = -1, | |
337 | .seq = -1, | |
338 | .name = "fuzzer_kernel_message"} | |
339 | }; | |
340 | netlink_parse_info(netlink_information_fetch, &zns.netlink, &zns, 1, 0); | |
341 | } | |
342 | ||
343 | /** | |
344 | * netlink_write_incoming() - Writes all data received from netlink to a file | |
345 | * @buf: Data from netlink. | |
346 | * @size: Size of data. | |
347 | * @counter: Counter for keeping filenames distinct. | |
348 | */ | |
349 | static void netlink_write_incoming(const char *buf, const unsigned int size, | |
350 | unsigned int counter) | |
351 | { | |
352 | char fname[MAXPATHLEN]; | |
353 | FILE *f; | |
354 | ||
355 | zserv_privs.change(ZPRIVS_RAISE); | |
356 | snprintf(fname, MAXPATHLEN, "%s/%s_%u", DAEMON_VTY_DIR, "netlink", | |
357 | counter); | |
358 | f = fopen(fname, "w"); | |
359 | if (f) { | |
360 | fwrite(buf, 1, size, f); | |
361 | fclose(f); | |
362 | } | |
363 | zserv_privs.change(ZPRIVS_LOWER); | |
364 | } | |
365 | ||
366 | /** | |
367 | * netlink_read_file() - Reads netlink data from file | |
368 | * @buf: Netlink buffer being overwritten. | |
369 | * @fname: File name to read from. | |
370 | * | |
371 | * Return: Size of file. | |
372 | */ | |
373 | static long netlink_read_file(char *buf, const char *fname) | |
374 | { | |
375 | FILE *f; | |
376 | long file_bytes = -1; | |
377 | zserv_privs.change(ZPRIVS_RAISE); | |
378 | f = fopen(fname, "r"); | |
379 | if (f) { | |
380 | fseek(f, 0, SEEK_END); | |
381 | file_bytes = ftell(f); | |
382 | rewind(f); | |
383 | fread(buf, file_bytes, 1, f); | |
384 | fclose(f); | |
385 | } | |
386 | zserv_privs.change(ZPRIVS_LOWER); | |
387 | return file_bytes; | |
388 | } | |
389 | ||
390 | #endif | |
391 | ||
d62a17ae | 392 | static int kernel_read(struct thread *thread) |
1fdc9eae | 393 | { |
d62a17ae | 394 | struct zebra_ns *zns = (struct zebra_ns *)THREAD_ARG(thread); |
395 | netlink_parse_info(netlink_information_fetch, &zns->netlink, zns, 5, 0); | |
396 | zns->t_netlink = NULL; | |
397 | thread_add_read(zebrad.master, kernel_read, zns, zns->netlink.sock, | |
398 | &zns->t_netlink); | |
1fdc9eae | 399 | |
d62a17ae | 400 | return 0; |
1fdc9eae | 401 | } |
402 | ||
3575d9e8 DS |
403 | /* |
404 | * Filter out messages from self that occur on listener socket, | |
1fdc9eae | 405 | * caused by our actions on the command socket |
3575d9e8 DS |
406 | * |
407 | * When we add new Netlink message types we probably | |
408 | * do not need to add them here as that we are filtering | |
409 | * on the routes we actually care to receive( which is rarer | |
410 | * then the normal course of operations). We are intentionally | |
411 | * allowing some messages from ourselves through | |
412 | * ( I'm looking at you Interface based netlink messages ) | |
413 | * so that we only had to write one way to handle incoming | |
414 | * address add/delete changes. | |
1fdc9eae | 415 | */ |
d62a17ae | 416 | static void netlink_install_filter(int sock, __u32 pid) |
1fdc9eae | 417 | { |
3575d9e8 DS |
418 | /* |
419 | * BPF_JUMP instructions and where you jump to are based upon | |
420 | * 0 as being the next statement. So count from 0. Writing | |
421 | * this down because every time I look at this I have to | |
422 | * re-remember it. | |
423 | */ | |
d62a17ae | 424 | struct sock_filter filter[] = { |
3575d9e8 DS |
425 | /* |
426 | * Logic: | |
427 | * if (nlmsg_pid == pid) { | |
428 | * if (the incoming nlmsg_type == | |
429 | * RTM_NEWADDR | RTM_DELADDR) | |
430 | * keep this message | |
431 | * else | |
432 | * skip this message | |
433 | * } else | |
434 | * keep this netlink message | |
435 | */ | |
436 | /* | |
437 | * 0: Load the nlmsg_pid into the BPF register | |
438 | */ | |
d62a17ae | 439 | BPF_STMT(BPF_LD | BPF_ABS | BPF_W, |
440 | offsetof(struct nlmsghdr, nlmsg_pid)), | |
3575d9e8 DS |
441 | /* |
442 | * 1: Compare to pid | |
443 | */ | |
444 | BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htonl(pid), 0, 4), | |
445 | /* | |
446 | * 2: Load the nlmsg_type into BPF register | |
447 | */ | |
448 | BPF_STMT(BPF_LD | BPF_ABS | BPF_H, | |
449 | offsetof(struct nlmsghdr, nlmsg_type)), | |
450 | /* | |
451 | * 3: Compare to RTM_NEWADDR | |
452 | */ | |
453 | BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_NEWADDR), 2, 0), | |
454 | /* | |
455 | * 4: Compare to RTM_DELADDR | |
456 | */ | |
457 | BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_DELADDR), 1, 0), | |
458 | /* | |
459 | * 5: This is the end state of we want to skip the | |
460 | * message | |
461 | */ | |
d62a17ae | 462 | BPF_STMT(BPF_RET | BPF_K, 0), |
3575d9e8 DS |
463 | /* 6: This is the end state of we want to keep |
464 | * the message | |
465 | */ | |
d62a17ae | 466 | BPF_STMT(BPF_RET | BPF_K, 0xffff), |
467 | }; | |
468 | ||
469 | struct sock_fprog prog = { | |
9d303b37 | 470 | .len = array_size(filter), .filter = filter, |
d62a17ae | 471 | }; |
472 | ||
473 | if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER, &prog, sizeof(prog)) | |
474 | < 0) | |
475 | zlog_warn("Can't install socket filter: %s\n", | |
476 | safe_strerror(errno)); | |
1fdc9eae | 477 | } |
478 | ||
d62a17ae | 479 | void netlink_parse_rtattr(struct rtattr **tb, int max, struct rtattr *rta, |
480 | int len) | |
1fdc9eae | 481 | { |
d62a17ae | 482 | while (RTA_OK(rta, len)) { |
483 | if (rta->rta_type <= max) | |
484 | tb[rta->rta_type] = rta; | |
485 | rta = RTA_NEXT(rta, len); | |
486 | } | |
1fdc9eae | 487 | } |
488 | ||
86391e56 MS |
489 | int addattr_l(struct nlmsghdr *n, unsigned int maxlen, int type, |
490 | const void *data, unsigned int alen) | |
1fdc9eae | 491 | { |
d62a17ae | 492 | int len; |
493 | struct rtattr *rta; | |
1fdc9eae | 494 | |
d62a17ae | 495 | len = RTA_LENGTH(alen); |
1fdc9eae | 496 | |
d62a17ae | 497 | if (NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len) > maxlen) |
498 | return -1; | |
1fdc9eae | 499 | |
d62a17ae | 500 | rta = (struct rtattr *)(((char *)n) + NLMSG_ALIGN(n->nlmsg_len)); |
501 | rta->rta_type = type; | |
502 | rta->rta_len = len; | |
4b2792b5 | 503 | |
d62a17ae | 504 | if (data) |
505 | memcpy(RTA_DATA(rta), data, alen); | |
506 | else | |
507 | assert(alen == 0); | |
4b2792b5 | 508 | |
d62a17ae | 509 | n->nlmsg_len = NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len); |
1fdc9eae | 510 | |
d62a17ae | 511 | return 0; |
1fdc9eae | 512 | } |
513 | ||
86391e56 MS |
514 | int rta_addattr_l(struct rtattr *rta, unsigned int maxlen, int type, |
515 | const void *data, unsigned int alen) | |
1fdc9eae | 516 | { |
d62a17ae | 517 | unsigned int len; |
518 | struct rtattr *subrta; | |
1fdc9eae | 519 | |
d62a17ae | 520 | len = RTA_LENGTH(alen); |
1fdc9eae | 521 | |
d62a17ae | 522 | if (RTA_ALIGN(rta->rta_len) + RTA_ALIGN(len) > maxlen) |
523 | return -1; | |
1fdc9eae | 524 | |
d62a17ae | 525 | subrta = (struct rtattr *)(((char *)rta) + RTA_ALIGN(rta->rta_len)); |
526 | subrta->rta_type = type; | |
527 | subrta->rta_len = len; | |
4b2792b5 | 528 | |
d62a17ae | 529 | if (data) |
530 | memcpy(RTA_DATA(subrta), data, alen); | |
531 | else | |
532 | assert(alen == 0); | |
4b2792b5 | 533 | |
d62a17ae | 534 | rta->rta_len = NLMSG_ALIGN(rta->rta_len) + RTA_ALIGN(len); |
1fdc9eae | 535 | |
d62a17ae | 536 | return 0; |
1fdc9eae | 537 | } |
538 | ||
d7c0a89a | 539 | int addattr16(struct nlmsghdr *n, unsigned int maxlen, int type, uint16_t data) |
bbc16902 | 540 | { |
d7c0a89a | 541 | return addattr_l(n, maxlen, type, &data, sizeof(uint16_t)); |
bbc16902 | 542 | } |
543 | ||
d62a17ae | 544 | int addattr32(struct nlmsghdr *n, unsigned int maxlen, int type, int data) |
1fdc9eae | 545 | { |
d7c0a89a | 546 | return addattr_l(n, maxlen, type, &data, sizeof(uint32_t)); |
1fdc9eae | 547 | } |
548 | ||
d62a17ae | 549 | struct rtattr *addattr_nest(struct nlmsghdr *n, int maxlen, int type) |
1fdc9eae | 550 | { |
d62a17ae | 551 | struct rtattr *nest = NLMSG_TAIL(n); |
1fdc9eae | 552 | |
d62a17ae | 553 | addattr_l(n, maxlen, type, NULL, 0); |
554 | return nest; | |
1fdc9eae | 555 | } |
556 | ||
d62a17ae | 557 | int addattr_nest_end(struct nlmsghdr *n, struct rtattr *nest) |
1fdc9eae | 558 | { |
d7c0a89a | 559 | nest->rta_len = (uint8_t *)NLMSG_TAIL(n) - (uint8_t *)nest; |
d62a17ae | 560 | return n->nlmsg_len; |
1fdc9eae | 561 | } |
562 | ||
d62a17ae | 563 | struct rtattr *rta_nest(struct rtattr *rta, int maxlen, int type) |
1fdc9eae | 564 | { |
d62a17ae | 565 | struct rtattr *nest = RTA_TAIL(rta); |
1fdc9eae | 566 | |
d62a17ae | 567 | rta_addattr_l(rta, maxlen, type, NULL, 0); |
568 | return nest; | |
1fdc9eae | 569 | } |
570 | ||
d62a17ae | 571 | int rta_nest_end(struct rtattr *rta, struct rtattr *nest) |
1fdc9eae | 572 | { |
d7c0a89a | 573 | nest->rta_len = (uint8_t *)RTA_TAIL(rta) - (uint8_t *)nest; |
d62a17ae | 574 | return rta->rta_len; |
1fdc9eae | 575 | } |
576 | ||
d62a17ae | 577 | const char *nl_msg_type_to_str(uint16_t msg_type) |
1fdc9eae | 578 | { |
d62a17ae | 579 | return lookup_msg(nlmsg_str, msg_type, ""); |
1fdc9eae | 580 | } |
581 | ||
d7c0a89a | 582 | const char *nl_rtproto_to_str(uint8_t rtproto) |
1fdc9eae | 583 | { |
d62a17ae | 584 | return lookup_msg(rtproto_str, rtproto, ""); |
1fdc9eae | 585 | } |
b339bde7 | 586 | |
d7c0a89a | 587 | const char *nl_family_to_str(uint8_t family) |
b339bde7 | 588 | { |
d62a17ae | 589 | return lookup_msg(family_str, family, ""); |
b339bde7 DS |
590 | } |
591 | ||
d7c0a89a | 592 | const char *nl_rttype_to_str(uint8_t rttype) |
b339bde7 | 593 | { |
d62a17ae | 594 | return lookup_msg(rttype_str, rttype, ""); |
b339bde7 DS |
595 | } |
596 | ||
5d307d5d DS |
597 | #define NL_OK(nla, len) \ |
598 | ((len) >= (int)sizeof(struct nlattr) \ | |
599 | && (nla)->nla_len >= sizeof(struct nlattr) \ | |
600 | && (nla)->nla_len <= (len)) | |
601 | #define NL_NEXT(nla, attrlen) \ | |
602 | ((attrlen) -= RTA_ALIGN((nla)->nla_len), \ | |
603 | (struct nlattr *)(((char *)(nla)) + RTA_ALIGN((nla)->nla_len))) | |
604 | #define NL_RTA(r) \ | |
605 | ((struct nlattr *)(((char *)(r)) \ | |
606 | + NLMSG_ALIGN(sizeof(struct nlmsgerr)))) | |
607 | ||
608 | static void netlink_parse_nlattr(struct nlattr **tb, int max, | |
609 | struct nlattr *nla, int len) | |
610 | { | |
611 | while (NL_OK(nla, len)) { | |
612 | if (nla->nla_type <= max) | |
613 | tb[nla->nla_type] = nla; | |
614 | nla = NL_NEXT(nla, len); | |
615 | } | |
616 | } | |
617 | ||
618 | static void netlink_parse_extended_ack(struct nlmsghdr *h) | |
619 | { | |
620 | struct nlattr *tb[NLMSGERR_ATTR_MAX + 1]; | |
621 | const struct nlmsgerr *err = | |
622 | (const struct nlmsgerr *)((uint8_t *)h | |
623 | + NLMSG_ALIGN( | |
624 | sizeof(struct nlmsghdr))); | |
625 | const struct nlmsghdr *err_nlh = NULL; | |
626 | uint32_t hlen = sizeof(*err); | |
627 | const char *msg = NULL; | |
628 | uint32_t off = 0; | |
629 | ||
630 | if (!(h->nlmsg_flags & NLM_F_CAPPED)) | |
631 | hlen += h->nlmsg_len - NLMSG_ALIGN(sizeof(struct nlmsghdr)); | |
632 | ||
633 | memset(tb, 0, sizeof(tb)); | |
634 | netlink_parse_nlattr(tb, NLMSGERR_ATTR_MAX, NL_RTA(h), hlen); | |
635 | ||
636 | if (tb[NLMSGERR_ATTR_MSG]) | |
637 | msg = (const char *)RTA_DATA(tb[NLMSGERR_ATTR_MSG]); | |
638 | ||
639 | if (tb[NLMSGERR_ATTR_OFFS]) { | |
640 | off = *(uint32_t *)RTA_DATA(tb[NLMSGERR_ATTR_OFFS]); | |
641 | ||
642 | if (off > h->nlmsg_len) { | |
643 | zlog_err("Invalid offset for NLMSGERR_ATTR_OFFS\n"); | |
5d307d5d DS |
644 | } else if (!(h->nlmsg_flags & NLM_F_CAPPED)) { |
645 | /* | |
646 | * Header of failed message | |
647 | * we are not doing anything currently with it | |
648 | * but noticing it for later. | |
649 | */ | |
650 | err_nlh = &err->msg; | |
651 | zlog_warn("%s: Received %d extended Ack", | |
652 | __PRETTY_FUNCTION__, err_nlh->nlmsg_type); | |
653 | } | |
654 | } | |
655 | ||
656 | if (msg && *msg != '\0') { | |
657 | bool is_err = !!err->error; | |
658 | ||
659 | if (is_err) | |
660 | zlog_err("Extended Error: %s", msg); | |
661 | else | |
662 | zlog_warn("Extended Warning: %s", msg); | |
663 | } | |
664 | } | |
665 | ||
936ebf0a DS |
666 | /* |
667 | * netlink_parse_info | |
668 | * | |
669 | * Receive message from netlink interface and pass those information | |
670 | * to the given function. | |
671 | * | |
672 | * filter -> Function to call to read the results | |
673 | * nl -> netlink socket information | |
674 | * zns -> The zebra namespace data | |
675 | * count -> How many we should read in, 0 means as much as possible | |
676 | * startup -> Are we reading in under startup conditions? passed to | |
677 | * the filter. | |
678 | */ | |
2414abd3 | 679 | int netlink_parse_info(int (*filter)(struct nlmsghdr *, ns_id_t, int), |
d62a17ae | 680 | struct nlsock *nl, struct zebra_ns *zns, int count, |
681 | int startup) | |
1fdc9eae | 682 | { |
d62a17ae | 683 | int status; |
684 | int ret = 0; | |
685 | int error; | |
686 | int read_in = 0; | |
687 | ||
688 | while (1) { | |
9ed7517b | 689 | char buf[NL_RCV_PKT_BUF_SIZE]; |
d62a17ae | 690 | struct iovec iov = {.iov_base = buf, .iov_len = sizeof buf}; |
691 | struct sockaddr_nl snl; | |
692 | struct msghdr msg = {.msg_name = (void *)&snl, | |
693 | .msg_namelen = sizeof snl, | |
694 | .msg_iov = &iov, | |
695 | .msg_iovlen = 1}; | |
696 | struct nlmsghdr *h; | |
697 | ||
698 | if (count && read_in >= count) | |
699 | return 0; | |
700 | ||
81a2f870 SW |
701 | #if defined(HANDLE_ZAPI_FUZZING) |
702 | /* Check if reading and filename is set */ | |
703 | if (netlink_read && '\0' != netlink_fuzz_file[0]) { | |
704 | zlog_debug("Reading netlink fuzz file"); | |
705 | status = netlink_read_file(buf, netlink_fuzz_file); | |
706 | snl.nl_pid = 0; | |
707 | } else { | |
708 | status = recvmsg(nl->sock, &msg, 0); | |
709 | } | |
710 | #else | |
d62a17ae | 711 | status = recvmsg(nl->sock, &msg, 0); |
81a2f870 | 712 | #endif |
d62a17ae | 713 | if (status < 0) { |
714 | if (errno == EINTR) | |
715 | continue; | |
716 | if (errno == EWOULDBLOCK || errno == EAGAIN) | |
717 | break; | |
718 | zlog_err("%s recvmsg overrun: %s", nl->name, | |
719 | safe_strerror(errno)); | |
720 | /* | |
721 | * In this case we are screwed. | |
722 | * There is no good way to | |
723 | * recover zebra at this point. | |
724 | */ | |
725 | exit(-1); | |
726 | continue; | |
1fdc9eae | 727 | } |
728 | ||
d62a17ae | 729 | if (status == 0) { |
730 | zlog_err("%s EOF", nl->name); | |
731 | return -1; | |
732 | } | |
733 | ||
734 | if (msg.msg_namelen != sizeof snl) { | |
735 | zlog_err("%s sender address length error: length %d", | |
736 | nl->name, msg.msg_namelen); | |
737 | return -1; | |
738 | } | |
739 | ||
740 | if (IS_ZEBRA_DEBUG_KERNEL_MSGDUMP_RECV) { | |
741 | zlog_debug("%s: << netlink message dump [recv]", | |
742 | __func__); | |
743 | zlog_hexdump(buf, status); | |
744 | } | |
745 | ||
81a2f870 SW |
746 | #if defined(HANDLE_ZAPI_FUZZING) |
747 | if (!netlink_read) { | |
748 | zlog_debug("Writing incoming netlink message"); | |
749 | netlink_write_incoming(buf, status, | |
750 | netlink_file_counter++); | |
751 | } | |
752 | #endif | |
753 | ||
d62a17ae | 754 | read_in++; |
755 | for (h = (struct nlmsghdr *)buf; | |
756 | NLMSG_OK(h, (unsigned int)status); | |
757 | h = NLMSG_NEXT(h, status)) { | |
758 | /* Finish of reading. */ | |
759 | if (h->nlmsg_type == NLMSG_DONE) | |
760 | return ret; | |
761 | ||
762 | /* Error handling. */ | |
763 | if (h->nlmsg_type == NLMSG_ERROR) { | |
764 | struct nlmsgerr *err = | |
765 | (struct nlmsgerr *)NLMSG_DATA(h); | |
766 | int errnum = err->error; | |
767 | int msg_type = err->msg.nlmsg_type; | |
768 | ||
5d307d5d DS |
769 | if (h->nlmsg_len |
770 | < NLMSG_LENGTH(sizeof(struct nlmsgerr))) { | |
771 | zlog_err("%s error: message truncated", | |
772 | nl->name); | |
773 | return -1; | |
774 | } | |
775 | ||
776 | /* | |
777 | * Parse the extended information before | |
778 | * we actually handle it. | |
779 | * At this point in time we do not | |
780 | * do anything other than report the | |
781 | * issue. | |
782 | */ | |
783 | if (h->nlmsg_flags & NLM_F_ACK_TLVS) | |
784 | netlink_parse_extended_ack(h); | |
785 | ||
d62a17ae | 786 | /* If the error field is zero, then this is an |
787 | * ACK */ | |
788 | if (err->error == 0) { | |
789 | if (IS_ZEBRA_DEBUG_KERNEL) { | |
790 | zlog_debug( | |
791 | "%s: %s ACK: type=%s(%u), seq=%u, pid=%u", | |
792 | __FUNCTION__, nl->name, | |
793 | nl_msg_type_to_str( | |
794 | err->msg.nlmsg_type), | |
795 | err->msg.nlmsg_type, | |
796 | err->msg.nlmsg_seq, | |
797 | err->msg.nlmsg_pid); | |
798 | } | |
799 | ||
800 | /* return if not a multipart message, | |
801 | * otherwise continue */ | |
802 | if (!(h->nlmsg_flags & NLM_F_MULTI)) | |
803 | return 0; | |
804 | continue; | |
805 | } | |
806 | ||
d62a17ae | 807 | /* Deal with errors that occur because of races |
808 | * in link handling */ | |
809 | if (nl == &zns->netlink_cmd | |
810 | && ((msg_type == RTM_DELROUTE | |
811 | && (-errnum == ENODEV | |
812 | || -errnum == ESRCH)) | |
813 | || (msg_type == RTM_NEWROUTE | |
814 | && (-errnum == ENETDOWN | |
815 | || -errnum == EEXIST)))) { | |
816 | if (IS_ZEBRA_DEBUG_KERNEL) | |
817 | zlog_debug( | |
818 | "%s: error: %s type=%s(%u), seq=%u, pid=%u", | |
819 | nl->name, | |
820 | safe_strerror(-errnum), | |
821 | nl_msg_type_to_str( | |
822 | msg_type), | |
823 | msg_type, | |
824 | err->msg.nlmsg_seq, | |
825 | err->msg.nlmsg_pid); | |
826 | return 0; | |
827 | } | |
828 | ||
829 | /* We see RTM_DELNEIGH when shutting down an | |
830 | * interface with an IPv4 | |
831 | * link-local. The kernel should have already | |
832 | * deleted the neighbor | |
833 | * so do not log these as an error. | |
834 | */ | |
835 | if (msg_type == RTM_DELNEIGH | |
836 | || (nl == &zns->netlink_cmd | |
837 | && msg_type == RTM_NEWROUTE | |
838 | && (-errnum == ESRCH | |
839 | || -errnum == ENETUNREACH))) { | |
840 | /* This is known to happen in some | |
841 | * situations, don't log | |
842 | * as error. | |
843 | */ | |
844 | if (IS_ZEBRA_DEBUG_KERNEL) | |
845 | zlog_debug( | |
846 | "%s error: %s, type=%s(%u), seq=%u, pid=%u", | |
847 | nl->name, | |
848 | safe_strerror(-errnum), | |
849 | nl_msg_type_to_str( | |
850 | msg_type), | |
851 | msg_type, | |
852 | err->msg.nlmsg_seq, | |
853 | err->msg.nlmsg_pid); | |
854 | } else | |
855 | zlog_err( | |
856 | "%s error: %s, type=%s(%u), seq=%u, pid=%u", | |
857 | nl->name, | |
858 | safe_strerror(-errnum), | |
859 | nl_msg_type_to_str(msg_type), | |
860 | msg_type, err->msg.nlmsg_seq, | |
861 | err->msg.nlmsg_pid); | |
862 | ||
863 | return -1; | |
864 | } | |
865 | ||
866 | /* OK we got netlink message. */ | |
867 | if (IS_ZEBRA_DEBUG_KERNEL) | |
868 | zlog_debug( | |
869 | "netlink_parse_info: %s type %s(%u), len=%d, seq=%u, pid=%u", | |
870 | nl->name, | |
871 | nl_msg_type_to_str(h->nlmsg_type), | |
872 | h->nlmsg_type, h->nlmsg_len, | |
873 | h->nlmsg_seq, h->nlmsg_pid); | |
874 | ||
783827ae DS |
875 | |
876 | /* | |
877 | * Ignore messages that maybe sent from | |
878 | * other actors besides the kernel | |
879 | */ | |
880 | if (snl.nl_pid != 0) { | |
881 | zlog_err("Ignoring message from pid %u", | |
882 | snl.nl_pid); | |
d62a17ae | 883 | continue; |
884 | } | |
885 | ||
2414abd3 | 886 | error = (*filter)(h, zns->ns_id, startup); |
d62a17ae | 887 | if (error < 0) { |
888 | zlog_err("%s filter function error", nl->name); | |
6ca7b664 | 889 | zlog_backtrace(LOG_ERR); |
d62a17ae | 890 | ret = error; |
891 | } | |
892 | } | |
893 | ||
894 | /* After error care. */ | |
895 | if (msg.msg_flags & MSG_TRUNC) { | |
896 | zlog_err("%s error: message truncated", nl->name); | |
897 | continue; | |
898 | } | |
899 | if (status) { | |
900 | zlog_err("%s error: data remnant size %d", nl->name, | |
901 | status); | |
902 | return -1; | |
903 | } | |
904 | } | |
905 | return ret; | |
1fdc9eae | 906 | } |
907 | ||
936ebf0a DS |
908 | /* |
909 | * netlink_talk | |
910 | * | |
911 | * sendmsg() to netlink socket then recvmsg(). | |
912 | * Calls netlink_parse_info to parse returned data | |
913 | * | |
914 | * filter -> The filter to read final results from kernel | |
915 | * nlmsghdr -> The data to send to the kernel | |
916 | * nl -> The netlink socket information | |
917 | * zns -> The zebra namespace information | |
918 | * startup -> Are we reading in under startup conditions | |
919 | * This is passed through eventually to filter. | |
920 | */ | |
2414abd3 | 921 | int netlink_talk(int (*filter)(struct nlmsghdr *, ns_id_t, int startup), |
d62a17ae | 922 | struct nlmsghdr *n, struct nlsock *nl, struct zebra_ns *zns, |
923 | int startup) | |
1fdc9eae | 924 | { |
d62a17ae | 925 | int status; |
926 | struct sockaddr_nl snl; | |
cbaca6a1 DS |
927 | struct iovec iov; |
928 | struct msghdr msg; | |
d62a17ae | 929 | int save_errno; |
930 | ||
931 | memset(&snl, 0, sizeof snl); | |
cbaca6a1 DS |
932 | memset(&iov, 0, sizeof iov); |
933 | memset(&msg, 0, sizeof msg); | |
934 | ||
935 | iov.iov_base = n; | |
936 | iov.iov_len = n->nlmsg_len; | |
937 | msg.msg_name = (void *)&snl; | |
938 | msg.msg_namelen = sizeof snl; | |
939 | msg.msg_iov = &iov; | |
940 | msg.msg_iovlen = 1; | |
941 | ||
d62a17ae | 942 | snl.nl_family = AF_NETLINK; |
943 | ||
944 | n->nlmsg_seq = ++nl->seq; | |
945 | n->nlmsg_pid = nl->snl.nl_pid; | |
946 | ||
d62a17ae | 947 | if (IS_ZEBRA_DEBUG_KERNEL) |
948 | zlog_debug( | |
949 | "netlink_talk: %s type %s(%u), len=%d seq=%u flags 0x%x", | |
950 | nl->name, nl_msg_type_to_str(n->nlmsg_type), | |
951 | n->nlmsg_type, n->nlmsg_len, n->nlmsg_seq, | |
952 | n->nlmsg_flags); | |
953 | ||
954 | /* Send message to netlink interface. */ | |
955 | if (zserv_privs.change(ZPRIVS_RAISE)) | |
956 | zlog_err("Can't raise privileges"); | |
957 | status = sendmsg(nl->sock, &msg, 0); | |
958 | save_errno = errno; | |
959 | if (zserv_privs.change(ZPRIVS_LOWER)) | |
960 | zlog_err("Can't lower privileges"); | |
961 | ||
962 | if (IS_ZEBRA_DEBUG_KERNEL_MSGDUMP_SEND) { | |
963 | zlog_debug("%s: >> netlink message dump [sent]", __func__); | |
964 | zlog_hexdump(n, n->nlmsg_len); | |
965 | } | |
966 | ||
967 | if (status < 0) { | |
968 | zlog_err("netlink_talk sendmsg() error: %s", | |
969 | safe_strerror(save_errno)); | |
970 | return -1; | |
971 | } | |
972 | ||
973 | ||
974 | /* | |
975 | * Get reply from netlink socket. | |
976 | * The reply should either be an acknowlegement or an error. | |
977 | */ | |
978 | return netlink_parse_info(filter, nl, zns, 0, startup); | |
1fdc9eae | 979 | } |
980 | ||
289602d7 | 981 | /* Issue request message to kernel via netlink socket. GET messages |
982 | * are issued through this interface. | |
983 | */ | |
d62a17ae | 984 | int netlink_request(struct nlsock *nl, struct nlmsghdr *n) |
1fdc9eae | 985 | { |
d62a17ae | 986 | int ret; |
987 | struct sockaddr_nl snl; | |
988 | int save_errno; | |
989 | ||
990 | /* Check netlink socket. */ | |
991 | if (nl->sock < 0) { | |
992 | zlog_err("%s socket isn't active.", nl->name); | |
993 | return -1; | |
994 | } | |
995 | ||
996 | /* Fill common fields for all requests. */ | |
997 | n->nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST; | |
998 | n->nlmsg_pid = nl->snl.nl_pid; | |
999 | n->nlmsg_seq = ++nl->seq; | |
1000 | ||
1001 | memset(&snl, 0, sizeof snl); | |
1002 | snl.nl_family = AF_NETLINK; | |
1003 | ||
1004 | /* Raise capabilities and send message, then lower capabilities. */ | |
1005 | if (zserv_privs.change(ZPRIVS_RAISE)) { | |
1006 | zlog_err("Can't raise privileges"); | |
1007 | return -1; | |
1008 | } | |
1009 | ||
1010 | ret = sendto(nl->sock, (void *)n, n->nlmsg_len, 0, | |
1011 | (struct sockaddr *)&snl, sizeof snl); | |
1012 | save_errno = errno; | |
1013 | ||
1014 | if (zserv_privs.change(ZPRIVS_LOWER)) | |
1015 | zlog_err("Can't lower privileges"); | |
1016 | ||
1017 | if (ret < 0) { | |
1018 | zlog_err("%s sendto failed: %s", nl->name, | |
1019 | safe_strerror(save_errno)); | |
1020 | return -1; | |
1021 | } | |
1022 | ||
1023 | return 0; | |
1fdc9eae | 1024 | } |
1025 | ||
1026 | /* Exported interface function. This function simply calls | |
1027 | netlink_socket (). */ | |
d62a17ae | 1028 | void kernel_init(struct zebra_ns *zns) |
1fdc9eae | 1029 | { |
d62a17ae | 1030 | unsigned long groups; |
5d307d5d DS |
1031 | #if defined SOL_NETLINK |
1032 | int one, ret; | |
1033 | #endif | |
d62a17ae | 1034 | |
026a316f DS |
1035 | /* |
1036 | * Initialize netlink sockets | |
1037 | * | |
1038 | * If RTMGRP_XXX exists use that, but at some point | |
1039 | * I think the kernel developers realized that | |
1040 | * keeping track of all the different values would | |
1041 | * lead to confusion, so we need to convert the | |
1042 | * RTNLGRP_XXX to a bit position for ourself | |
1043 | */ | |
1044 | groups = RTMGRP_LINK | | |
1045 | RTMGRP_IPV4_ROUTE | | |
1046 | RTMGRP_IPV4_IFADDR | | |
1047 | RTMGRP_IPV6_ROUTE | | |
1048 | RTMGRP_IPV6_IFADDR | | |
1049 | RTMGRP_IPV4_MROUTE | | |
1050 | RTMGRP_NEIGH | | |
1051 | (1 << (RTNLGRP_IPV4_RULE - 1)) | | |
1052 | (1 << (RTNLGRP_IPV6_RULE - 1)); | |
d62a17ae | 1053 | |
1054 | snprintf(zns->netlink.name, sizeof(zns->netlink.name), | |
1055 | "netlink-listen (NS %u)", zns->ns_id); | |
1056 | zns->netlink.sock = -1; | |
19d5a4fe DS |
1057 | if (netlink_socket(&zns->netlink, groups, zns->ns_id) < 0) { |
1058 | zlog_err("Failure to create %s socket", | |
1059 | zns->netlink.name); | |
1060 | exit(-1); | |
1061 | } | |
d62a17ae | 1062 | |
1063 | snprintf(zns->netlink_cmd.name, sizeof(zns->netlink_cmd.name), | |
1064 | "netlink-cmd (NS %u)", zns->ns_id); | |
1065 | zns->netlink_cmd.sock = -1; | |
19d5a4fe DS |
1066 | if (netlink_socket(&zns->netlink_cmd, 0, zns->ns_id) < 0) { |
1067 | zlog_err("Failure to create %s socket", | |
1068 | zns->netlink_cmd.name); | |
1069 | exit(-1); | |
1070 | } | |
d62a17ae | 1071 | |
5d307d5d DS |
1072 | /* |
1073 | * SOL_NETLINK is not available on all platforms yet | |
1074 | * apparently. It's in bits/socket.h which I am not | |
1075 | * sure that we want to pull into our build system. | |
1076 | */ | |
1077 | #if defined SOL_NETLINK | |
1078 | /* | |
1079 | * Let's tell the kernel that we want to receive extended | |
1080 | * ACKS over our command socket | |
1081 | */ | |
1082 | one = 1; | |
1083 | ret = setsockopt(zns->netlink_cmd.sock, SOL_NETLINK, NETLINK_EXT_ACK, | |
1084 | &one, sizeof(one)); | |
1085 | ||
1086 | if (ret < 0) | |
1087 | zlog_notice("Registration for extended ACK failed : %d %s", | |
1088 | errno, safe_strerror(errno)); | |
1089 | #endif | |
1090 | ||
d62a17ae | 1091 | /* Register kernel socket. */ |
19d5a4fe | 1092 | if (fcntl(zns->netlink.sock, F_SETFL, O_NONBLOCK) < 0) |
8c85e8ea DS |
1093 | zlog_err("Can't set %s socket error: %s(%d)", |
1094 | zns->netlink.name, safe_strerror(errno), errno); | |
1095 | ||
1096 | if (fcntl(zns->netlink_cmd.sock, F_SETFL, O_NONBLOCK) < 0) | |
1097 | zlog_err("Can't set %s socket error: %s(%d)", | |
1098 | zns->netlink_cmd.name, safe_strerror(errno), errno); | |
19d5a4fe DS |
1099 | |
1100 | /* Set receive buffer size if it's set from command line */ | |
1101 | if (nl_rcvbufsize) | |
1102 | netlink_recvbuf(&zns->netlink, nl_rcvbufsize); | |
1103 | ||
1104 | netlink_install_filter(zns->netlink.sock, | |
1105 | zns->netlink_cmd.snl.nl_pid); | |
1106 | zns->t_netlink = NULL; | |
1107 | ||
1108 | thread_add_read(zebrad.master, kernel_read, zns, | |
1109 | zns->netlink.sock, &zns->t_netlink); | |
d62a17ae | 1110 | |
1111 | rt_netlink_init(); | |
1fdc9eae | 1112 | } |
1113 | ||
d62a17ae | 1114 | void kernel_terminate(struct zebra_ns *zns) |
1fdc9eae | 1115 | { |
d62a17ae | 1116 | THREAD_READ_OFF(zns->t_netlink); |
1117 | ||
1118 | if (zns->netlink.sock >= 0) { | |
1119 | close(zns->netlink.sock); | |
1120 | zns->netlink.sock = -1; | |
1121 | } | |
1122 | ||
1123 | if (zns->netlink_cmd.sock >= 0) { | |
1124 | close(zns->netlink_cmd.sock); | |
1125 | zns->netlink_cmd.sock = -1; | |
1126 | } | |
1fdc9eae | 1127 | } |
ddfeb486 DL |
1128 | |
1129 | #endif /* HAVE_NETLINK */ |