]>
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 | ||
ddfeb486 DL |
23 | #ifdef HAVE_NETLINK |
24 | ||
1fdc9eae | 25 | #include "linklist.h" |
26 | #include "if.h" | |
27 | #include "log.h" | |
28 | #include "prefix.h" | |
29 | #include "connected.h" | |
30 | #include "table.h" | |
31 | #include "memory.h" | |
1fdc9eae | 32 | #include "rib.h" |
33 | #include "thread.h" | |
34 | #include "privs.h" | |
35 | #include "nexthop.h" | |
36 | #include "vrf.h" | |
37 | #include "mpls.h" | |
174482ef | 38 | #include "lib_errors.h" |
d4000d7b | 39 | #include "hash.h" |
1fdc9eae | 40 | |
3801e764 | 41 | #include "zebra/zebra_router.h" |
1fdc9eae | 42 | #include "zebra/zebra_ns.h" |
43 | #include "zebra/zebra_vrf.h" | |
05f7f5db | 44 | #include "zebra/rt.h" |
1fdc9eae | 45 | #include "zebra/debug.h" |
46 | #include "zebra/kernel_netlink.h" | |
47 | #include "zebra/rt_netlink.h" | |
48 | #include "zebra/if_netlink.h" | |
942bf97b | 49 | #include "zebra/rule_netlink.h" |
ebb61fca | 50 | #include "zebra/netconf_netlink.h" |
43e52561 | 51 | #include "zebra/zebra_errors.h" |
1fdc9eae | 52 | |
53 | #ifndef SO_RCVBUFFORCE | |
54 | #define SO_RCVBUFFORCE (33) | |
55 | #endif | |
56 | ||
57 | /* Hack for GNU libc version 2. */ | |
58 | #ifndef MSG_TRUNC | |
59 | #define MSG_TRUNC 0x20 | |
60 | #endif /* MSG_TRUNC */ | |
61 | ||
62 | #ifndef NLMSG_TAIL | |
d62a17ae | 63 | #define NLMSG_TAIL(nmsg) \ |
d7c0a89a QY |
64 | ((struct rtattr *)(((uint8_t *)(nmsg)) \ |
65 | + NLMSG_ALIGN((nmsg)->nlmsg_len))) | |
1fdc9eae | 66 | #endif |
67 | ||
68 | #ifndef RTA_TAIL | |
d62a17ae | 69 | #define RTA_TAIL(rta) \ |
d7c0a89a | 70 | ((struct rtattr *)(((uint8_t *)(rta)) + RTA_ALIGN((rta)->rta_len))) |
1fdc9eae | 71 | #endif |
72 | ||
f909c673 DS |
73 | #ifndef RTNL_FAMILY_IP6MR |
74 | #define RTNL_FAMILY_IP6MR 129 | |
75 | #endif | |
76 | ||
77 | #ifndef RTPROT_MROUTED | |
78 | #define RTPROT_MROUTED 17 | |
79 | #endif | |
80 | ||
531c92b8 | 81 | #define NL_DEFAULT_BATCH_BUFSIZE (16 * NL_PKT_BUF_SIZE) |
e63c7622 JU |
82 | |
83 | /* | |
84 | * We limit the batch's size to a number smaller than the length of the | |
85 | * underlying buffer since the last message that wouldn't fit the batch would go | |
86 | * over the upper boundary and then it would have to be encoded again into a new | |
87 | * buffer. If the difference between the limit and the length of the buffer is | |
88 | * big enough (bigger than the biggest Netlink message) then this situation | |
89 | * won't occur. | |
90 | */ | |
531c92b8 JU |
91 | #define NL_DEFAULT_BATCH_SEND_THRESHOLD (15 * NL_PKT_BUF_SIZE) |
92 | ||
d62a17ae | 93 | static const struct message nlmsg_str[] = {{RTM_NEWROUTE, "RTM_NEWROUTE"}, |
94 | {RTM_DELROUTE, "RTM_DELROUTE"}, | |
95 | {RTM_GETROUTE, "RTM_GETROUTE"}, | |
96 | {RTM_NEWLINK, "RTM_NEWLINK"}, | |
5d414138 | 97 | {RTM_SETLINK, "RTM_SETLINK"}, |
d62a17ae | 98 | {RTM_DELLINK, "RTM_DELLINK"}, |
99 | {RTM_GETLINK, "RTM_GETLINK"}, | |
100 | {RTM_NEWADDR, "RTM_NEWADDR"}, | |
101 | {RTM_DELADDR, "RTM_DELADDR"}, | |
102 | {RTM_GETADDR, "RTM_GETADDR"}, | |
103 | {RTM_NEWNEIGH, "RTM_NEWNEIGH"}, | |
104 | {RTM_DELNEIGH, "RTM_DELNEIGH"}, | |
105 | {RTM_GETNEIGH, "RTM_GETNEIGH"}, | |
942bf97b | 106 | {RTM_NEWRULE, "RTM_NEWRULE"}, |
107 | {RTM_DELRULE, "RTM_DELRULE"}, | |
108 | {RTM_GETRULE, "RTM_GETRULE"}, | |
79580b5a SW |
109 | {RTM_NEWNEXTHOP, "RTM_NEWNEXTHOP"}, |
110 | {RTM_DELNEXTHOP, "RTM_DELNEXTHOP"}, | |
111 | {RTM_GETNEXTHOP, "RTM_GETNEXTHOP"}, | |
ebb61fca DS |
112 | {RTM_NEWNETCONF, "RTM_NEWNETCONF"}, |
113 | {RTM_DELNETCONF, "RTM_DELNETCONF"}, | |
acc8e687 CS |
114 | {RTM_NEWTUNNEL, "RTM_NEWTUNNEL"}, |
115 | {RTM_DELTUNNEL, "RTM_DELTUNNEL"}, | |
116 | {RTM_GETTUNNEL, "RTM_GETTUNNEL"}, | |
d62a17ae | 117 | {0}}; |
1fdc9eae | 118 | |
119 | static const struct message rtproto_str[] = { | |
d62a17ae | 120 | {RTPROT_REDIRECT, "redirect"}, |
121 | {RTPROT_KERNEL, "kernel"}, | |
122 | {RTPROT_BOOT, "boot"}, | |
123 | {RTPROT_STATIC, "static"}, | |
124 | {RTPROT_GATED, "GateD"}, | |
125 | {RTPROT_RA, "router advertisement"}, | |
126 | {RTPROT_MRT, "MRT"}, | |
127 | {RTPROT_ZEBRA, "Zebra"}, | |
1fdc9eae | 128 | #ifdef RTPROT_BIRD |
d62a17ae | 129 | {RTPROT_BIRD, "BIRD"}, |
1fdc9eae | 130 | #endif /* RTPROT_BIRD */ |
d62a17ae | 131 | {RTPROT_MROUTED, "mroute"}, |
132 | {RTPROT_BGP, "BGP"}, | |
133 | {RTPROT_OSPF, "OSPF"}, | |
134 | {RTPROT_ISIS, "IS-IS"}, | |
135 | {RTPROT_RIP, "RIP"}, | |
136 | {RTPROT_RIPNG, "RIPNG"}, | |
d4d71f11 | 137 | {RTPROT_ZSTATIC, "static"}, |
d62a17ae | 138 | {0}}; |
139 | ||
140 | static const struct message family_str[] = {{AF_INET, "ipv4"}, | |
141 | {AF_INET6, "ipv6"}, | |
142 | {AF_BRIDGE, "bridge"}, | |
143 | {RTNL_FAMILY_IPMR, "ipv4MR"}, | |
144 | {RTNL_FAMILY_IP6MR, "ipv6MR"}, | |
145 | {0}}; | |
146 | ||
8c8f250b DS |
147 | static const struct message rttype_str[] = {{RTN_UNSPEC, "none"}, |
148 | {RTN_UNICAST, "unicast"}, | |
149 | {RTN_LOCAL, "local"}, | |
150 | {RTN_BROADCAST, "broadcast"}, | |
151 | {RTN_ANYCAST, "anycast"}, | |
d62a17ae | 152 | {RTN_MULTICAST, "multicast"}, |
8c8f250b DS |
153 | {RTN_BLACKHOLE, "blackhole"}, |
154 | {RTN_UNREACHABLE, "unreachable"}, | |
155 | {RTN_PROHIBIT, "prohibited"}, | |
156 | {RTN_THROW, "throw"}, | |
157 | {RTN_NAT, "nat"}, | |
158 | {RTN_XRESOLVE, "resolver"}, | |
d62a17ae | 159 | {0}}; |
b339bde7 | 160 | |
1fdc9eae | 161 | extern struct thread_master *master; |
1fdc9eae | 162 | |
163 | extern struct zebra_privs_t zserv_privs; | |
164 | ||
bf8d3d6a | 165 | DEFINE_MTYPE_STATIC(ZEBRA, NL_BUF, "Zebra Netlink buffers"); |
531c92b8 | 166 | |
34869809 MS |
167 | /* Hashtable and mutex to allow lookup of nlsock structs by socket/fd value. |
168 | * We have both the main and dplane pthreads using these structs, so we have | |
169 | * to protect the hash with a lock. | |
170 | */ | |
171 | static struct hash *nlsock_hash; | |
172 | pthread_mutex_t nlsock_mutex; | |
173 | ||
174 | /* Lock and unlock wrappers for nlsock hash */ | |
175 | #define NLSOCK_LOCK() pthread_mutex_lock(&nlsock_mutex) | |
176 | #define NLSOCK_UNLOCK() pthread_mutex_unlock(&nlsock_mutex) | |
177 | ||
531c92b8 JU |
178 | size_t nl_batch_tx_bufsize; |
179 | char *nl_batch_tx_buf; | |
180 | ||
531c92b8 JU |
181 | _Atomic uint32_t nl_batch_bufsize = NL_DEFAULT_BATCH_BUFSIZE; |
182 | _Atomic uint32_t nl_batch_send_threshold = NL_DEFAULT_BATCH_SEND_THRESHOLD; | |
183 | ||
e63c7622 JU |
184 | struct nl_batch { |
185 | void *buf; | |
186 | size_t bufsiz; | |
187 | size_t limit; | |
188 | ||
189 | void *buf_head; | |
190 | size_t curlen; | |
191 | size_t msgcnt; | |
192 | ||
193 | const struct zebra_dplane_info *zns; | |
e63c7622 | 194 | |
f6feb48b | 195 | struct dplane_ctx_q ctx_list; |
e63c7622 | 196 | |
f6feb48b JU |
197 | /* |
198 | * Pointer to the queue of completed contexts outbound back | |
199 | * towards the dataplane module. | |
200 | */ | |
201 | struct dplane_ctx_q *ctx_out_q; | |
e63c7622 JU |
202 | }; |
203 | ||
531c92b8 JU |
204 | int netlink_config_write_helper(struct vty *vty) |
205 | { | |
206 | uint32_t size = | |
207 | atomic_load_explicit(&nl_batch_bufsize, memory_order_relaxed); | |
208 | uint32_t threshold = atomic_load_explicit(&nl_batch_send_threshold, | |
209 | memory_order_relaxed); | |
210 | ||
211 | if (size != NL_DEFAULT_BATCH_BUFSIZE | |
212 | || threshold != NL_DEFAULT_BATCH_SEND_THRESHOLD) | |
213 | vty_out(vty, "zebra kernel netlink batch-tx-buf %u %u\n", size, | |
214 | threshold); | |
215 | ||
c40e1b1c SW |
216 | if (if_netlink_frr_protodown_r_bit_is_set()) |
217 | vty_out(vty, "zebra protodown reason-bit %u\n", | |
218 | if_netlink_get_frr_protodown_r_bit()); | |
219 | ||
531c92b8 JU |
220 | return 0; |
221 | } | |
222 | ||
223 | void netlink_set_batch_buffer_size(uint32_t size, uint32_t threshold, bool set) | |
224 | { | |
225 | if (!set) { | |
226 | size = NL_DEFAULT_BATCH_BUFSIZE; | |
227 | threshold = NL_DEFAULT_BATCH_SEND_THRESHOLD; | |
228 | } | |
229 | ||
230 | atomic_store_explicit(&nl_batch_bufsize, size, memory_order_relaxed); | |
231 | atomic_store_explicit(&nl_batch_send_threshold, threshold, | |
232 | memory_order_relaxed); | |
233 | } | |
234 | ||
2414abd3 | 235 | int netlink_talk_filter(struct nlmsghdr *h, ns_id_t ns_id, int startup) |
1fdc9eae | 236 | { |
3575d9e8 DS |
237 | /* |
238 | * This is an error condition that must be handled during | |
239 | * development. | |
240 | * | |
241 | * The netlink_talk_filter function is used for communication | |
242 | * down the netlink_cmd pipe and we are expecting | |
243 | * an ack being received. So if we get here | |
244 | * then we did not receive the ack and instead | |
245 | * received some other message in an unexpected | |
246 | * way. | |
247 | */ | |
43e52561 QY |
248 | zlog_debug("%s: ignoring message type 0x%04x(%s) NS %u", __func__, |
249 | h->nlmsg_type, nl_msg_type_to_str(h->nlmsg_type), ns_id); | |
d62a17ae | 250 | return 0; |
1fdc9eae | 251 | } |
252 | ||
d62a17ae | 253 | static int netlink_recvbuf(struct nlsock *nl, uint32_t newsize) |
1fdc9eae | 254 | { |
d7c0a89a | 255 | uint32_t oldsize; |
d62a17ae | 256 | socklen_t newlen = sizeof(newsize); |
257 | socklen_t oldlen = sizeof(oldsize); | |
258 | int ret; | |
259 | ||
260 | ret = getsockopt(nl->sock, SOL_SOCKET, SO_RCVBUF, &oldsize, &oldlen); | |
261 | if (ret < 0) { | |
450971aa | 262 | flog_err_sys(EC_LIB_SOCKET, |
09c866e3 QY |
263 | "Can't get %s receive buffer size: %s", nl->name, |
264 | safe_strerror(errno)); | |
d62a17ae | 265 | return -1; |
266 | } | |
267 | ||
268 | /* Try force option (linux >= 2.6.14) and fall back to normal set */ | |
0cf6db21 | 269 | frr_with_privs(&zserv_privs) { |
01b9e3fd | 270 | ret = setsockopt(nl->sock, SOL_SOCKET, SO_RCVBUFFORCE, |
9fb83b55 | 271 | &rcvbufsize, sizeof(rcvbufsize)); |
01b9e3fd | 272 | } |
d62a17ae | 273 | if (ret < 0) |
9fb83b55 DS |
274 | ret = setsockopt(nl->sock, SOL_SOCKET, SO_RCVBUF, &rcvbufsize, |
275 | sizeof(rcvbufsize)); | |
d62a17ae | 276 | if (ret < 0) { |
450971aa | 277 | flog_err_sys(EC_LIB_SOCKET, |
09c866e3 QY |
278 | "Can't set %s receive buffer size: %s", nl->name, |
279 | safe_strerror(errno)); | |
d62a17ae | 280 | return -1; |
281 | } | |
282 | ||
283 | ret = getsockopt(nl->sock, SOL_SOCKET, SO_RCVBUF, &newsize, &newlen); | |
284 | if (ret < 0) { | |
450971aa | 285 | flog_err_sys(EC_LIB_SOCKET, |
09c866e3 QY |
286 | "Can't get %s receive buffer size: %s", nl->name, |
287 | safe_strerror(errno)); | |
d62a17ae | 288 | return -1; |
289 | } | |
d62a17ae | 290 | return 0; |
1fdc9eae | 291 | } |
292 | ||
fe953d7c DS |
293 | static const char *group2str(uint32_t group) |
294 | { | |
295 | switch (group) { | |
296 | case RTNLGRP_TUNNEL: | |
297 | return "RTNLGRP_TUNNEL"; | |
298 | default: | |
299 | return "UNKNOWN"; | |
300 | } | |
301 | } | |
302 | ||
1fdc9eae | 303 | /* Make socket for Linux netlink interface. */ |
d62a17ae | 304 | static int netlink_socket(struct nlsock *nl, unsigned long groups, |
fe953d7c DS |
305 | uint32_t ext_groups[], uint8_t ext_group_size, |
306 | ns_id_t ns_id) | |
1fdc9eae | 307 | { |
d62a17ae | 308 | int ret; |
309 | struct sockaddr_nl snl; | |
310 | int sock; | |
311 | int namelen; | |
d62a17ae | 312 | |
0cf6db21 | 313 | frr_with_privs(&zserv_privs) { |
6bb30c2c DL |
314 | sock = ns_socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE, ns_id); |
315 | if (sock < 0) { | |
316 | zlog_err("Can't open %s socket: %s", nl->name, | |
317 | safe_strerror(errno)); | |
318 | return -1; | |
319 | } | |
d62a17ae | 320 | |
0d6f7fd6 | 321 | memset(&snl, 0, sizeof(snl)); |
6bb30c2c DL |
322 | snl.nl_family = AF_NETLINK; |
323 | snl.nl_groups = groups; | |
d62a17ae | 324 | |
fe953d7c DS |
325 | if (ext_group_size) { |
326 | uint8_t i; | |
327 | ||
328 | for (i = 0; i < ext_group_size; i++) { | |
cc408c06 | 329 | #if defined SOL_NETLINK |
fe953d7c DS |
330 | ret = setsockopt(sock, SOL_NETLINK, |
331 | NETLINK_ADD_MEMBERSHIP, | |
332 | &ext_groups[i], | |
333 | sizeof(ext_groups[i])); | |
334 | if (ret < 0) { | |
335 | zlog_notice( | |
336 | "can't setsockopt NETLINK_ADD_MEMBERSHIP for group %s(%u), this linux kernel does not support it: %s(%d)", | |
337 | group2str(ext_groups[i]), | |
338 | ext_groups[i], | |
339 | safe_strerror(errno), errno); | |
340 | } | |
cc408c06 DS |
341 | #else |
342 | zlog_notice( | |
343 | "Unable to use NETLINK_ADD_MEMBERSHIP via SOL_NETLINK for %s(%u) since the linux kernel does not support the socket option", | |
344 | group2str(ext_groups[i]), | |
345 | ext_groups[i]); | |
346 | #endif | |
e5b1de8a CS |
347 | } |
348 | } | |
47e2eb27 | 349 | |
6bb30c2c | 350 | /* Bind the socket to the netlink structure for anything. */ |
0d6f7fd6 | 351 | ret = bind(sock, (struct sockaddr *)&snl, sizeof(snl)); |
6bb30c2c | 352 | } |
d62a17ae | 353 | |
354 | if (ret < 0) { | |
6bb30c2c DL |
355 | zlog_err("Can't bind %s socket to group 0x%x: %s", nl->name, |
356 | snl.nl_groups, safe_strerror(errno)); | |
d62a17ae | 357 | close(sock); |
358 | return -1; | |
359 | } | |
360 | ||
361 | /* multiple netlink sockets will have different nl_pid */ | |
0d6f7fd6 | 362 | namelen = sizeof(snl); |
d62a17ae | 363 | ret = getsockname(sock, (struct sockaddr *)&snl, (socklen_t *)&namelen); |
0d6f7fd6 | 364 | if (ret < 0 || namelen != sizeof(snl)) { |
450971aa | 365 | flog_err_sys(EC_LIB_SOCKET, "Can't get %s socket name: %s", |
09c866e3 | 366 | nl->name, safe_strerror(errno)); |
d62a17ae | 367 | close(sock); |
368 | return -1; | |
369 | } | |
370 | ||
371 | nl->snl = snl; | |
372 | nl->sock = sock; | |
2cf7651f DS |
373 | nl->buflen = NL_RCV_PKT_BUF_SIZE; |
374 | nl->buf = XMALLOC(MTYPE_NL_BUF, nl->buflen); | |
375 | ||
d62a17ae | 376 | return ret; |
1fdc9eae | 377 | } |
378 | ||
d166308b MS |
379 | /* |
380 | * Dispatch an incoming netlink message; used by the zebra main pthread's | |
381 | * netlink event reader. | |
382 | */ | |
2414abd3 | 383 | static int netlink_information_fetch(struct nlmsghdr *h, ns_id_t ns_id, |
d62a17ae | 384 | int startup) |
1fdc9eae | 385 | { |
3575d9e8 DS |
386 | /* |
387 | * When we handle new message types here | |
388 | * because we are starting to install them | |
389 | * then lets check the netlink_install_filter | |
390 | * and see if we should add the corresponding | |
391 | * allow through entry there. | |
392 | * Probably not needed to do but please | |
393 | * think about it. | |
394 | */ | |
d62a17ae | 395 | switch (h->nlmsg_type) { |
396 | case RTM_NEWROUTE: | |
2414abd3 | 397 | return netlink_route_change(h, ns_id, startup); |
d62a17ae | 398 | case RTM_DELROUTE: |
2414abd3 | 399 | return netlink_route_change(h, ns_id, startup); |
d62a17ae | 400 | case RTM_NEWLINK: |
2414abd3 | 401 | return netlink_link_change(h, ns_id, startup); |
d62a17ae | 402 | case RTM_DELLINK: |
2414abd3 | 403 | return netlink_link_change(h, ns_id, startup); |
d62a17ae | 404 | case RTM_NEWNEIGH: |
d62a17ae | 405 | case RTM_DELNEIGH: |
951f8bcb | 406 | case RTM_GETNEIGH: |
7a52f27e | 407 | return netlink_neigh_change(h, ns_id); |
942bf97b | 408 | case RTM_NEWRULE: |
2414abd3 | 409 | return netlink_rule_change(h, ns_id, startup); |
942bf97b | 410 | case RTM_DELRULE: |
2414abd3 | 411 | return netlink_rule_change(h, ns_id, startup); |
79580b5a | 412 | case RTM_NEWNEXTHOP: |
d9f5b2f5 | 413 | return netlink_nexthop_change(h, ns_id, startup); |
79580b5a | 414 | case RTM_DELNEXTHOP: |
d9f5b2f5 | 415 | return netlink_nexthop_change(h, ns_id, startup); |
d166308b MS |
416 | |
417 | /* Messages handled in the dplane thread */ | |
418 | case RTM_NEWADDR: | |
419 | case RTM_DELADDR: | |
cd787a8a MS |
420 | case RTM_NEWNETCONF: |
421 | case RTM_DELNETCONF: | |
acc8e687 CS |
422 | case RTM_NEWTUNNEL: |
423 | case RTM_DELTUNNEL: | |
424 | case RTM_GETTUNNEL: | |
d166308b | 425 | return 0; |
d62a17ae | 426 | default: |
3575d9e8 DS |
427 | /* |
428 | * If we have received this message then | |
429 | * we have made a mistake during development | |
430 | * and we need to write some code to handle | |
431 | * this message type or not ask for | |
432 | * it to be sent up to us | |
433 | */ | |
e914ccbe | 434 | flog_err(EC_ZEBRA_UNKNOWN_NLMSG, |
1d5453d6 | 435 | "Unknown netlink nlmsg_type %s(%d) vrf %u", |
1c50c1c0 QY |
436 | nl_msg_type_to_str(h->nlmsg_type), h->nlmsg_type, |
437 | ns_id); | |
d62a17ae | 438 | break; |
439 | } | |
440 | return 0; | |
1fdc9eae | 441 | } |
442 | ||
d166308b MS |
443 | /* |
444 | * Dispatch an incoming netlink message; used by the dataplane pthread's | |
445 | * netlink event reader code. | |
446 | */ | |
447 | static int dplane_netlink_information_fetch(struct nlmsghdr *h, ns_id_t ns_id, | |
448 | int startup) | |
449 | { | |
450 | /* | |
451 | * Dispatch the incoming messages that the dplane pthread handles | |
452 | */ | |
453 | switch (h->nlmsg_type) { | |
454 | case RTM_NEWADDR: | |
455 | case RTM_DELADDR: | |
456 | return netlink_interface_addr_dplane(h, ns_id, startup); | |
457 | ||
9f3f1486 MS |
458 | case RTM_NEWNETCONF: |
459 | case RTM_DELNETCONF: | |
460 | return netlink_netconf_change(h, ns_id, startup); | |
461 | ||
462 | /* TODO -- other messages for the dplane socket and pthread */ | |
463 | ||
d166308b MS |
464 | case RTM_NEWLINK: |
465 | case RTM_DELLINK: | |
466 | ||
467 | default: | |
468 | break; | |
469 | } | |
470 | ||
471 | return 0; | |
472 | } | |
473 | ||
cc9f21da | 474 | static void kernel_read(struct thread *thread) |
1fdc9eae | 475 | { |
d62a17ae | 476 | struct zebra_ns *zns = (struct zebra_ns *)THREAD_ARG(thread); |
85a75f1e MS |
477 | struct zebra_dplane_info dp_info; |
478 | ||
479 | /* Capture key info from ns struct */ | |
480 | zebra_dplane_info_from_zns(&dp_info, zns, false); | |
481 | ||
482 | netlink_parse_info(netlink_information_fetch, &zns->netlink, &dp_info, | |
9bfadae8 | 483 | 5, false); |
d166308b | 484 | |
3801e764 | 485 | thread_add_read(zrouter.master, kernel_read, zns, zns->netlink.sock, |
d62a17ae | 486 | &zns->t_netlink); |
1fdc9eae | 487 | } |
488 | ||
d166308b MS |
489 | /* |
490 | * Called by the dplane pthread to read incoming OS messages and dispatch them. | |
491 | */ | |
492 | int kernel_dplane_read(struct zebra_dplane_info *info) | |
493 | { | |
d4000d7b DS |
494 | struct nlsock *nl = kernel_netlink_nlsock_lookup(info->sock); |
495 | ||
496 | netlink_parse_info(dplane_netlink_information_fetch, nl, info, 5, | |
497 | false); | |
d166308b MS |
498 | |
499 | return 0; | |
500 | } | |
501 | ||
3575d9e8 DS |
502 | /* |
503 | * Filter out messages from self that occur on listener socket, | |
62b8bb7a | 504 | * caused by our actions on the command socket(s) |
3575d9e8 DS |
505 | * |
506 | * When we add new Netlink message types we probably | |
507 | * do not need to add them here as that we are filtering | |
508 | * on the routes we actually care to receive( which is rarer | |
509 | * then the normal course of operations). We are intentionally | |
510 | * allowing some messages from ourselves through | |
511 | * ( I'm looking at you Interface based netlink messages ) | |
9f3f1486 MS |
512 | * so that we only have to write one way to handle incoming |
513 | * address add/delete and xxxNETCONF changes. | |
1fdc9eae | 514 | */ |
ff45112c | 515 | static void netlink_install_filter(int sock, uint32_t pid, uint32_t dplane_pid) |
1fdc9eae | 516 | { |
3575d9e8 DS |
517 | /* |
518 | * BPF_JUMP instructions and where you jump to are based upon | |
519 | * 0 as being the next statement. So count from 0. Writing | |
520 | * this down because every time I look at this I have to | |
521 | * re-remember it. | |
522 | */ | |
d62a17ae | 523 | struct sock_filter filter[] = { |
3575d9e8 DS |
524 | /* |
525 | * Logic: | |
62b8bb7a MS |
526 | * if (nlmsg_pid == pid || |
527 | * nlmsg_pid == dplane_pid) { | |
3575d9e8 | 528 | * if (the incoming nlmsg_type == |
9f3f1486 MS |
529 | * RTM_NEWADDR || RTM_DELADDR || RTM_NEWNETCONF || |
530 | * RTM_DELNETCONF) | |
3575d9e8 DS |
531 | * keep this message |
532 | * else | |
533 | * skip this message | |
534 | * } else | |
535 | * keep this netlink message | |
536 | */ | |
537 | /* | |
538 | * 0: Load the nlmsg_pid into the BPF register | |
539 | */ | |
d62a17ae | 540 | BPF_STMT(BPF_LD | BPF_ABS | BPF_W, |
541 | offsetof(struct nlmsghdr, nlmsg_pid)), | |
3575d9e8 DS |
542 | /* |
543 | * 1: Compare to pid | |
544 | */ | |
62b8bb7a | 545 | BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htonl(pid), 1, 0), |
3575d9e8 | 546 | /* |
62b8bb7a MS |
547 | * 2: Compare to dplane pid |
548 | */ | |
9f3f1486 | 549 | BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htonl(dplane_pid), 0, 6), |
62b8bb7a MS |
550 | /* |
551 | * 3: Load the nlmsg_type into BPF register | |
3575d9e8 DS |
552 | */ |
553 | BPF_STMT(BPF_LD | BPF_ABS | BPF_H, | |
554 | offsetof(struct nlmsghdr, nlmsg_type)), | |
555 | /* | |
62b8bb7a | 556 | * 4: Compare to RTM_NEWADDR |
3575d9e8 | 557 | */ |
9f3f1486 | 558 | BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_NEWADDR), 4, 0), |
3575d9e8 | 559 | /* |
62b8bb7a | 560 | * 5: Compare to RTM_DELADDR |
3575d9e8 | 561 | */ |
9f3f1486 MS |
562 | BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_DELADDR), 3, 0), |
563 | /* | |
564 | * 6: Compare to RTM_NEWNETCONF | |
565 | */ | |
566 | BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_NEWNETCONF), 2, | |
567 | 0), | |
568 | /* | |
569 | * 7: Compare to RTM_DELNETCONF | |
570 | */ | |
571 | BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_DELNETCONF), 1, | |
572 | 0), | |
3575d9e8 | 573 | /* |
9f3f1486 | 574 | * 8: This is the end state of we want to skip the |
3575d9e8 DS |
575 | * message |
576 | */ | |
d62a17ae | 577 | BPF_STMT(BPF_RET | BPF_K, 0), |
9f3f1486 | 578 | /* 9: This is the end state of we want to keep |
3575d9e8 DS |
579 | * the message |
580 | */ | |
d62a17ae | 581 | BPF_STMT(BPF_RET | BPF_K, 0xffff), |
582 | }; | |
583 | ||
584 | struct sock_fprog prog = { | |
9d303b37 | 585 | .len = array_size(filter), .filter = filter, |
d62a17ae | 586 | }; |
587 | ||
588 | if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER, &prog, sizeof(prog)) | |
589 | < 0) | |
1d5453d6 | 590 | flog_err_sys(EC_LIB_SOCKET, "Can't install socket filter: %s", |
9df414fe | 591 | safe_strerror(errno)); |
1fdc9eae | 592 | } |
593 | ||
d166308b MS |
594 | void netlink_parse_rtattr_flags(struct rtattr **tb, int max, struct rtattr *rta, |
595 | int len, unsigned short flags) | |
4bcdb608 NA |
596 | { |
597 | unsigned short type; | |
598 | ||
269b69d7 | 599 | memset(tb, 0, sizeof(struct rtattr *) * (max + 1)); |
4bcdb608 NA |
600 | while (RTA_OK(rta, len)) { |
601 | type = rta->rta_type & ~flags; | |
602 | if ((type <= max) && (!tb[type])) | |
603 | tb[type] = rta; | |
604 | rta = RTA_NEXT(rta, len); | |
605 | } | |
606 | } | |
607 | ||
d62a17ae | 608 | void netlink_parse_rtattr(struct rtattr **tb, int max, struct rtattr *rta, |
609 | int len) | |
1fdc9eae | 610 | { |
269b69d7 | 611 | memset(tb, 0, sizeof(struct rtattr *) * (max + 1)); |
d62a17ae | 612 | while (RTA_OK(rta, len)) { |
613 | if (rta->rta_type <= max) | |
614 | tb[rta->rta_type] = rta; | |
615 | rta = RTA_NEXT(rta, len); | |
616 | } | |
1fdc9eae | 617 | } |
618 | ||
87da6a60 SW |
619 | /** |
620 | * netlink_parse_rtattr_nested() - Parses a nested route attribute | |
621 | * @tb: Pointer to array for storing rtattr in. | |
622 | * @max: Max number to store. | |
623 | * @rta: Pointer to rtattr to look for nested items in. | |
624 | */ | |
625 | void netlink_parse_rtattr_nested(struct rtattr **tb, int max, | |
626 | struct rtattr *rta) | |
627 | { | |
628 | netlink_parse_rtattr(tb, max, RTA_DATA(rta), RTA_PAYLOAD(rta)); | |
629 | } | |
630 | ||
f8f3e484 CS |
631 | bool nl_addraw_l(struct nlmsghdr *n, unsigned int maxlen, const void *data, |
632 | unsigned int len) | |
633 | { | |
634 | if (NLMSG_ALIGN(n->nlmsg_len) + NLMSG_ALIGN(len) > maxlen) { | |
1b3cf91b | 635 | zlog_err("ERROR message exceeded bound of %d", maxlen); |
f8f3e484 CS |
636 | return false; |
637 | } | |
638 | ||
639 | memcpy(NLMSG_TAIL(n), data, len); | |
640 | memset((uint8_t *)NLMSG_TAIL(n) + len, 0, NLMSG_ALIGN(len) - len); | |
641 | n->nlmsg_len = NLMSG_ALIGN(n->nlmsg_len) + NLMSG_ALIGN(len); | |
642 | ||
643 | return true; | |
644 | } | |
645 | ||
312a6bee JU |
646 | bool nl_attr_put(struct nlmsghdr *n, unsigned int maxlen, int type, |
647 | const void *data, unsigned int alen) | |
1fdc9eae | 648 | { |
d62a17ae | 649 | int len; |
650 | struct rtattr *rta; | |
1fdc9eae | 651 | |
d62a17ae | 652 | len = RTA_LENGTH(alen); |
1fdc9eae | 653 | |
d62a17ae | 654 | if (NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len) > maxlen) |
312a6bee | 655 | return false; |
1fdc9eae | 656 | |
d62a17ae | 657 | rta = (struct rtattr *)(((char *)n) + NLMSG_ALIGN(n->nlmsg_len)); |
658 | rta->rta_type = type; | |
659 | rta->rta_len = len; | |
4b2792b5 | 660 | |
d62a17ae | 661 | if (data) |
662 | memcpy(RTA_DATA(rta), data, alen); | |
663 | else | |
664 | assert(alen == 0); | |
4b2792b5 | 665 | |
d62a17ae | 666 | n->nlmsg_len = NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len); |
1fdc9eae | 667 | |
312a6bee | 668 | return true; |
1fdc9eae | 669 | } |
670 | ||
94d70a65 DS |
671 | bool nl_attr_put8(struct nlmsghdr *n, unsigned int maxlen, int type, |
672 | uint8_t data) | |
673 | { | |
674 | return nl_attr_put(n, maxlen, type, &data, sizeof(uint8_t)); | |
675 | } | |
676 | ||
312a6bee JU |
677 | bool nl_attr_put16(struct nlmsghdr *n, unsigned int maxlen, int type, |
678 | uint16_t data) | |
bbc16902 | 679 | { |
312a6bee | 680 | return nl_attr_put(n, maxlen, type, &data, sizeof(uint16_t)); |
bbc16902 | 681 | } |
682 | ||
312a6bee JU |
683 | bool nl_attr_put32(struct nlmsghdr *n, unsigned int maxlen, int type, |
684 | uint32_t data) | |
1fdc9eae | 685 | { |
312a6bee | 686 | return nl_attr_put(n, maxlen, type, &data, sizeof(uint32_t)); |
1fdc9eae | 687 | } |
688 | ||
312a6bee | 689 | struct rtattr *nl_attr_nest(struct nlmsghdr *n, unsigned int maxlen, int type) |
1fdc9eae | 690 | { |
d62a17ae | 691 | struct rtattr *nest = NLMSG_TAIL(n); |
1fdc9eae | 692 | |
312a6bee JU |
693 | if (!nl_attr_put(n, maxlen, type, NULL, 0)) |
694 | return NULL; | |
695 | ||
40d86eba | 696 | nest->rta_type |= NLA_F_NESTED; |
d62a17ae | 697 | return nest; |
1fdc9eae | 698 | } |
699 | ||
312a6bee | 700 | int nl_attr_nest_end(struct nlmsghdr *n, struct rtattr *nest) |
1fdc9eae | 701 | { |
d7c0a89a | 702 | nest->rta_len = (uint8_t *)NLMSG_TAIL(n) - (uint8_t *)nest; |
d62a17ae | 703 | return n->nlmsg_len; |
1fdc9eae | 704 | } |
705 | ||
312a6bee | 706 | struct rtnexthop *nl_attr_rtnh(struct nlmsghdr *n, unsigned int maxlen) |
1fdc9eae | 707 | { |
312a6bee | 708 | struct rtnexthop *rtnh = (struct rtnexthop *)NLMSG_TAIL(n); |
1fdc9eae | 709 | |
312a6bee JU |
710 | if (NLMSG_ALIGN(n->nlmsg_len) + RTNH_ALIGN(sizeof(struct rtnexthop)) |
711 | > maxlen) | |
712 | return NULL; | |
713 | ||
714 | memset(rtnh, 0, sizeof(struct rtnexthop)); | |
715 | n->nlmsg_len = | |
716 | NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(sizeof(struct rtnexthop)); | |
717 | ||
718 | return rtnh; | |
1fdc9eae | 719 | } |
720 | ||
312a6bee | 721 | void nl_attr_rtnh_end(struct nlmsghdr *n, struct rtnexthop *rtnh) |
1fdc9eae | 722 | { |
312a6bee | 723 | rtnh->rtnh_len = (uint8_t *)NLMSG_TAIL(n) - (uint8_t *)rtnh; |
1fdc9eae | 724 | } |
725 | ||
f8f3e484 CS |
726 | bool nl_rta_put(struct rtattr *rta, unsigned int maxlen, int type, |
727 | const void *data, int alen) | |
728 | { | |
729 | struct rtattr *subrta; | |
730 | int len = RTA_LENGTH(alen); | |
731 | ||
732 | if (RTA_ALIGN(rta->rta_len) + RTA_ALIGN(len) > maxlen) { | |
733 | zlog_err("ERROR max allowed bound %d exceeded for rtattr", | |
734 | maxlen); | |
735 | return false; | |
736 | } | |
737 | subrta = (struct rtattr *)(((char *)rta) + RTA_ALIGN(rta->rta_len)); | |
738 | subrta->rta_type = type; | |
739 | subrta->rta_len = len; | |
740 | if (alen) | |
741 | memcpy(RTA_DATA(subrta), data, alen); | |
742 | rta->rta_len = NLMSG_ALIGN(rta->rta_len) + RTA_ALIGN(len); | |
743 | ||
744 | return true; | |
745 | } | |
746 | ||
747 | bool nl_rta_put16(struct rtattr *rta, unsigned int maxlen, int type, | |
748 | uint16_t data) | |
749 | { | |
750 | return nl_rta_put(rta, maxlen, type, &data, sizeof(uint16_t)); | |
751 | } | |
752 | ||
753 | bool nl_rta_put64(struct rtattr *rta, unsigned int maxlen, int type, | |
754 | uint64_t data) | |
755 | { | |
756 | return nl_rta_put(rta, maxlen, type, &data, sizeof(uint64_t)); | |
757 | } | |
758 | ||
759 | struct rtattr *nl_rta_nest(struct rtattr *rta, unsigned int maxlen, int type) | |
760 | { | |
761 | struct rtattr *nest = RTA_TAIL(rta); | |
762 | ||
763 | if (nl_rta_put(rta, maxlen, type, NULL, 0)) | |
764 | return NULL; | |
765 | ||
766 | nest->rta_type |= NLA_F_NESTED; | |
767 | ||
768 | return nest; | |
769 | } | |
770 | ||
771 | int nl_rta_nest_end(struct rtattr *rta, struct rtattr *nest) | |
772 | { | |
773 | nest->rta_len = (uint8_t *)RTA_TAIL(rta) - (uint8_t *)nest; | |
774 | ||
775 | return rta->rta_len; | |
776 | } | |
777 | ||
d62a17ae | 778 | const char *nl_msg_type_to_str(uint16_t msg_type) |
1fdc9eae | 779 | { |
d62a17ae | 780 | return lookup_msg(nlmsg_str, msg_type, ""); |
1fdc9eae | 781 | } |
782 | ||
d7c0a89a | 783 | const char *nl_rtproto_to_str(uint8_t rtproto) |
1fdc9eae | 784 | { |
d62a17ae | 785 | return lookup_msg(rtproto_str, rtproto, ""); |
1fdc9eae | 786 | } |
b339bde7 | 787 | |
d7c0a89a | 788 | const char *nl_family_to_str(uint8_t family) |
b339bde7 | 789 | { |
d62a17ae | 790 | return lookup_msg(family_str, family, ""); |
b339bde7 DS |
791 | } |
792 | ||
d7c0a89a | 793 | const char *nl_rttype_to_str(uint8_t rttype) |
b339bde7 | 794 | { |
d62a17ae | 795 | return lookup_msg(rttype_str, rttype, ""); |
b339bde7 DS |
796 | } |
797 | ||
4cebb2b6 | 798 | #define NLA_OK(nla, len) \ |
5d307d5d DS |
799 | ((len) >= (int)sizeof(struct nlattr) \ |
800 | && (nla)->nla_len >= sizeof(struct nlattr) \ | |
801 | && (nla)->nla_len <= (len)) | |
4cebb2b6 SW |
802 | #define NLA_NEXT(nla, attrlen) \ |
803 | ((attrlen) -= NLA_ALIGN((nla)->nla_len), \ | |
804 | (struct nlattr *)(((char *)(nla)) + NLA_ALIGN((nla)->nla_len))) | |
805 | #define NLA_LENGTH(len) (NLA_ALIGN(sizeof(struct nlattr)) + (len)) | |
806 | #define NLA_DATA(nla) ((struct nlattr *)(((char *)(nla)) + NLA_LENGTH(0))) | |
807 | ||
808 | #define ERR_NLA(err, inner_len) \ | |
809 | ((struct nlattr *)(((char *)(err)) \ | |
810 | + NLMSG_ALIGN(sizeof(struct nlmsgerr)) \ | |
811 | + NLMSG_ALIGN((inner_len)))) | |
5d307d5d DS |
812 | |
813 | static void netlink_parse_nlattr(struct nlattr **tb, int max, | |
814 | struct nlattr *nla, int len) | |
815 | { | |
4cebb2b6 | 816 | while (NLA_OK(nla, len)) { |
5d307d5d DS |
817 | if (nla->nla_type <= max) |
818 | tb[nla->nla_type] = nla; | |
4cebb2b6 | 819 | nla = NLA_NEXT(nla, len); |
5d307d5d DS |
820 | } |
821 | } | |
822 | ||
823 | static void netlink_parse_extended_ack(struct nlmsghdr *h) | |
824 | { | |
4cebb2b6 SW |
825 | struct nlattr *tb[NLMSGERR_ATTR_MAX + 1] = {}; |
826 | const struct nlmsgerr *err = (const struct nlmsgerr *)NLMSG_DATA(h); | |
5d307d5d | 827 | const struct nlmsghdr *err_nlh = NULL; |
4cebb2b6 SW |
828 | /* Length not including nlmsghdr */ |
829 | uint32_t len = 0; | |
830 | /* Inner error netlink message length */ | |
831 | uint32_t inner_len = 0; | |
5d307d5d DS |
832 | const char *msg = NULL; |
833 | uint32_t off = 0; | |
834 | ||
835 | if (!(h->nlmsg_flags & NLM_F_CAPPED)) | |
4cebb2b6 SW |
836 | inner_len = (uint32_t)NLMSG_PAYLOAD(&err->msg, 0); |
837 | ||
838 | len = (uint32_t)(NLMSG_PAYLOAD(h, sizeof(struct nlmsgerr)) - inner_len); | |
5d307d5d | 839 | |
4cebb2b6 SW |
840 | netlink_parse_nlattr(tb, NLMSGERR_ATTR_MAX, ERR_NLA(err, inner_len), |
841 | len); | |
5d307d5d DS |
842 | |
843 | if (tb[NLMSGERR_ATTR_MSG]) | |
4cebb2b6 | 844 | msg = (const char *)NLA_DATA(tb[NLMSGERR_ATTR_MSG]); |
5d307d5d DS |
845 | |
846 | if (tb[NLMSGERR_ATTR_OFFS]) { | |
4cebb2b6 | 847 | off = *(uint32_t *)NLA_DATA(tb[NLMSGERR_ATTR_OFFS]); |
5d307d5d DS |
848 | |
849 | if (off > h->nlmsg_len) { | |
9165c5f5 | 850 | zlog_err("Invalid offset for NLMSGERR_ATTR_OFFS"); |
5d307d5d DS |
851 | } else if (!(h->nlmsg_flags & NLM_F_CAPPED)) { |
852 | /* | |
853 | * Header of failed message | |
854 | * we are not doing anything currently with it | |
855 | * but noticing it for later. | |
856 | */ | |
857 | err_nlh = &err->msg; | |
15569c58 | 858 | zlog_debug("%s: Received %s extended Ack", __func__, |
87b5d1b0 | 859 | nl_msg_type_to_str(err_nlh->nlmsg_type)); |
5d307d5d DS |
860 | } |
861 | } | |
862 | ||
863 | if (msg && *msg != '\0') { | |
864 | bool is_err = !!err->error; | |
865 | ||
866 | if (is_err) | |
867 | zlog_err("Extended Error: %s", msg); | |
868 | else | |
e914ccbe | 869 | flog_warn(EC_ZEBRA_NETLINK_EXTENDED_WARNING, |
9df414fe | 870 | "Extended Warning: %s", msg); |
5d307d5d DS |
871 | } |
872 | } | |
873 | ||
ae6138bf JU |
874 | /* |
875 | * netlink_send_msg - send a netlink message of a certain size. | |
876 | * | |
877 | * Returns -1 on error. Otherwise, it returns the number of bytes sent. | |
878 | */ | |
f8653393 JU |
879 | static ssize_t netlink_send_msg(const struct nlsock *nl, void *buf, |
880 | size_t buflen) | |
ae6138bf | 881 | { |
f8653393 JU |
882 | struct sockaddr_nl snl = {}; |
883 | struct iovec iov = {}; | |
884 | struct msghdr msg = {}; | |
885 | ssize_t status; | |
886 | int save_errno = 0; | |
ae6138bf JU |
887 | |
888 | iov.iov_base = buf; | |
889 | iov.iov_len = buflen; | |
f8653393 | 890 | msg.msg_name = &snl; |
ae6138bf JU |
891 | msg.msg_namelen = sizeof(snl); |
892 | msg.msg_iov = &iov; | |
893 | msg.msg_iovlen = 1; | |
894 | ||
895 | snl.nl_family = AF_NETLINK; | |
896 | ||
897 | /* Send message to netlink interface. */ | |
898 | frr_with_privs(&zserv_privs) { | |
899 | status = sendmsg(nl->sock, &msg, 0); | |
900 | save_errno = errno; | |
901 | } | |
902 | ||
903 | if (IS_ZEBRA_DEBUG_KERNEL_MSGDUMP_SEND) { | |
904 | zlog_debug("%s: >> netlink message dump [sent]", __func__); | |
eead0bc4 RZ |
905 | #ifdef NETLINK_DEBUG |
906 | nl_dump(buf, buflen); | |
907 | #else | |
ae6138bf | 908 | zlog_hexdump(buf, buflen); |
eead0bc4 | 909 | #endif /* NETLINK_DEBUG */ |
ae6138bf JU |
910 | } |
911 | ||
f8653393 | 912 | if (status == -1) { |
ae6138bf JU |
913 | flog_err_sys(EC_LIB_SOCKET, "%s error: %s", __func__, |
914 | safe_strerror(save_errno)); | |
915 | return -1; | |
916 | } | |
917 | ||
918 | return status; | |
919 | } | |
920 | ||
921 | /* | |
922 | * netlink_recv_msg - receive a netlink message. | |
923 | * | |
924 | * Returns -1 on error, 0 if read would block or the number of bytes received. | |
925 | */ | |
2cf7651f | 926 | static int netlink_recv_msg(struct nlsock *nl, struct msghdr *msg) |
ae6138bf JU |
927 | { |
928 | struct iovec iov; | |
929 | int status; | |
930 | ||
2cf7651f DS |
931 | iov.iov_base = nl->buf; |
932 | iov.iov_len = nl->buflen; | |
933 | msg->msg_iov = &iov; | |
934 | msg->msg_iovlen = 1; | |
ae6138bf JU |
935 | |
936 | do { | |
2cf7651f DS |
937 | int bytes; |
938 | ||
939 | bytes = recv(nl->sock, NULL, 0, MSG_PEEK | MSG_TRUNC); | |
940 | ||
941 | if (bytes >= 0 && (size_t)bytes > nl->buflen) { | |
942 | nl->buf = XREALLOC(MTYPE_NL_BUF, nl->buf, bytes); | |
943 | nl->buflen = bytes; | |
944 | iov.iov_base = nl->buf; | |
945 | iov.iov_len = nl->buflen; | |
946 | } | |
947 | ||
948 | status = recvmsg(nl->sock, msg, 0); | |
f8653393 | 949 | } while (status == -1 && errno == EINTR); |
ae6138bf | 950 | |
f8653393 | 951 | if (status == -1) { |
ae6138bf JU |
952 | if (errno == EWOULDBLOCK || errno == EAGAIN) |
953 | return 0; | |
954 | flog_err(EC_ZEBRA_RECVMSG_OVERRUN, "%s recvmsg overrun: %s", | |
955 | nl->name, safe_strerror(errno)); | |
956 | /* | |
957 | * In this case we are screwed. There is no good way to recover | |
958 | * zebra at this point. | |
959 | */ | |
960 | exit(-1); | |
961 | } | |
962 | ||
963 | if (status == 0) { | |
964 | flog_err_sys(EC_LIB_SOCKET, "%s EOF", nl->name); | |
965 | return -1; | |
966 | } | |
967 | ||
2cf7651f | 968 | if (msg->msg_namelen != sizeof(struct sockaddr_nl)) { |
ae6138bf JU |
969 | flog_err(EC_ZEBRA_NETLINK_LENGTH_ERROR, |
970 | "%s sender address length error: length %d", nl->name, | |
2cf7651f | 971 | msg->msg_namelen); |
ae6138bf JU |
972 | return -1; |
973 | } | |
974 | ||
975 | if (IS_ZEBRA_DEBUG_KERNEL_MSGDUMP_RECV) { | |
976 | zlog_debug("%s: << netlink message dump [recv]", __func__); | |
eead0bc4 | 977 | #ifdef NETLINK_DEBUG |
2cf7651f | 978 | nl_dump(nl->buf, status); |
eead0bc4 | 979 | #else |
2cf7651f | 980 | zlog_hexdump(nl->buf, status); |
eead0bc4 | 981 | #endif /* NETLINK_DEBUG */ |
ae6138bf JU |
982 | } |
983 | ||
ae6138bf JU |
984 | return status; |
985 | } | |
986 | ||
987 | /* | |
988 | * netlink_parse_error - parse a netlink error message | |
989 | * | |
990 | * Returns 1 if this message is acknowledgement, 0 if this error should be | |
991 | * ignored, -1 otherwise. | |
992 | */ | |
993 | static int netlink_parse_error(const struct nlsock *nl, struct nlmsghdr *h, | |
d166308b | 994 | bool is_cmd, bool startup) |
ae6138bf JU |
995 | { |
996 | struct nlmsgerr *err = (struct nlmsgerr *)NLMSG_DATA(h); | |
997 | int errnum = err->error; | |
998 | int msg_type = err->msg.nlmsg_type; | |
999 | ||
1000 | if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) { | |
1001 | flog_err(EC_ZEBRA_NETLINK_LENGTH_ERROR, | |
1002 | "%s error: message truncated", nl->name); | |
1003 | return -1; | |
1004 | } | |
1005 | ||
1006 | /* | |
1007 | * Parse the extended information before we actually handle it. At this | |
1008 | * point in time we do not do anything other than report the issue. | |
1009 | */ | |
1010 | if (h->nlmsg_flags & NLM_F_ACK_TLVS) | |
1011 | netlink_parse_extended_ack(h); | |
1012 | ||
1013 | /* If the error field is zero, then this is an ACK. */ | |
1014 | if (err->error == 0) { | |
1015 | if (IS_ZEBRA_DEBUG_KERNEL) { | |
1016 | zlog_debug("%s: %s ACK: type=%s(%u), seq=%u, pid=%u", | |
1017 | __func__, nl->name, | |
1018 | nl_msg_type_to_str(err->msg.nlmsg_type), | |
1019 | err->msg.nlmsg_type, err->msg.nlmsg_seq, | |
1020 | err->msg.nlmsg_pid); | |
1021 | } | |
1022 | ||
1023 | return 1; | |
1024 | } | |
1025 | ||
1026 | /* Deal with errors that occur because of races in link handling. */ | |
d166308b | 1027 | if (is_cmd |
ae6138bf JU |
1028 | && ((msg_type == RTM_DELROUTE |
1029 | && (-errnum == ENODEV || -errnum == ESRCH)) | |
1030 | || (msg_type == RTM_NEWROUTE | |
1031 | && (-errnum == ENETDOWN || -errnum == EEXIST)))) { | |
1032 | if (IS_ZEBRA_DEBUG_KERNEL) | |
1033 | zlog_debug("%s: error: %s type=%s(%u), seq=%u, pid=%u", | |
1034 | nl->name, safe_strerror(-errnum), | |
1035 | nl_msg_type_to_str(msg_type), msg_type, | |
1036 | err->msg.nlmsg_seq, err->msg.nlmsg_pid); | |
1037 | return 0; | |
1038 | } | |
1039 | ||
1040 | /* | |
1041 | * We see RTM_DELNEIGH when shutting down an interface with an IPv4 | |
1042 | * link-local. The kernel should have already deleted the neighbor so | |
1043 | * do not log these as an error. | |
1044 | */ | |
1045 | if (msg_type == RTM_DELNEIGH | |
d166308b | 1046 | || (is_cmd && msg_type == RTM_NEWROUTE |
ae6138bf JU |
1047 | && (-errnum == ESRCH || -errnum == ENETUNREACH))) { |
1048 | /* | |
1049 | * This is known to happen in some situations, don't log as | |
1050 | * error. | |
1051 | */ | |
1052 | if (IS_ZEBRA_DEBUG_KERNEL) | |
1053 | zlog_debug("%s error: %s, type=%s(%u), seq=%u, pid=%u", | |
1054 | nl->name, safe_strerror(-errnum), | |
1055 | nl_msg_type_to_str(msg_type), msg_type, | |
1056 | err->msg.nlmsg_seq, err->msg.nlmsg_pid); | |
1057 | } else { | |
1058 | if ((msg_type != RTM_GETNEXTHOP) || !startup) | |
1059 | flog_err(EC_ZEBRA_UNEXPECTED_MESSAGE, | |
1060 | "%s error: %s, type=%s(%u), seq=%u, pid=%u", | |
1061 | nl->name, safe_strerror(-errnum), | |
1062 | nl_msg_type_to_str(msg_type), msg_type, | |
1063 | err->msg.nlmsg_seq, err->msg.nlmsg_pid); | |
1064 | } | |
1065 | ||
1066 | return -1; | |
1067 | } | |
1068 | ||
936ebf0a DS |
1069 | /* |
1070 | * netlink_parse_info | |
1071 | * | |
1072 | * Receive message from netlink interface and pass those information | |
1073 | * to the given function. | |
1074 | * | |
1075 | * filter -> Function to call to read the results | |
1076 | * nl -> netlink socket information | |
1077 | * zns -> The zebra namespace data | |
1078 | * count -> How many we should read in, 0 means as much as possible | |
1079 | * startup -> Are we reading in under startup conditions? passed to | |
1080 | * the filter. | |
1081 | */ | |
2414abd3 | 1082 | int netlink_parse_info(int (*filter)(struct nlmsghdr *, ns_id_t, int), |
2cf7651f | 1083 | struct nlsock *nl, const struct zebra_dplane_info *zns, |
9bfadae8 | 1084 | int count, bool startup) |
1fdc9eae | 1085 | { |
d62a17ae | 1086 | int status; |
1087 | int ret = 0; | |
1088 | int error; | |
1089 | int read_in = 0; | |
1090 | ||
1091 | while (1) { | |
d62a17ae | 1092 | struct sockaddr_nl snl; |
1093 | struct msghdr msg = {.msg_name = (void *)&snl, | |
ae6138bf | 1094 | .msg_namelen = sizeof(snl)}; |
d62a17ae | 1095 | struct nlmsghdr *h; |
1096 | ||
1097 | if (count && read_in >= count) | |
1098 | return 0; | |
1099 | ||
2cf7651f | 1100 | status = netlink_recv_msg(nl, &msg); |
ae6138bf | 1101 | if (status == -1) |
d62a17ae | 1102 | return -1; |
ae6138bf JU |
1103 | else if (status == 0) |
1104 | break; | |
81a2f870 | 1105 | |
d62a17ae | 1106 | read_in++; |
2cf7651f | 1107 | for (h = (struct nlmsghdr *)nl->buf; |
e6a0e0d1 | 1108 | (status >= 0 && NLMSG_OK(h, (unsigned int)status)); |
d62a17ae | 1109 | h = NLMSG_NEXT(h, status)) { |
1110 | /* Finish of reading. */ | |
1111 | if (h->nlmsg_type == NLMSG_DONE) | |
1112 | return ret; | |
1113 | ||
1114 | /* Error handling. */ | |
1115 | if (h->nlmsg_type == NLMSG_ERROR) { | |
d166308b MS |
1116 | int err = netlink_parse_error( |
1117 | nl, h, zns->is_cmd, startup); | |
1118 | ||
ae6138bf | 1119 | if (err == 1) { |
d62a17ae | 1120 | if (!(h->nlmsg_flags & NLM_F_MULTI)) |
1121 | return 0; | |
1122 | continue; | |
ae6138bf JU |
1123 | } else |
1124 | return err; | |
d62a17ae | 1125 | } |
1126 | ||
2f71996a DS |
1127 | /* |
1128 | * What is the right thing to do? The kernel | |
1129 | * is telling us that the dump request was interrupted | |
1130 | * and we more than likely are out of luck and have | |
1131 | * missed data from the kernel. At this point in time | |
1132 | * lets just note that this is happening. | |
1133 | */ | |
1134 | if (h->nlmsg_flags & NLM_F_DUMP_INTR) | |
1135 | flog_err( | |
1136 | EC_ZEBRA_NETLINK_BAD_SEQUENCE, | |
1137 | "netlink recvmsg: The Dump request was interrupted"); | |
1138 | ||
d62a17ae | 1139 | /* OK we got netlink message. */ |
1140 | if (IS_ZEBRA_DEBUG_KERNEL) | |
1141 | zlog_debug( | |
d166308b MS |
1142 | "%s: %s type %s(%u), len=%d, seq=%u, pid=%u", |
1143 | __func__, nl->name, | |
d62a17ae | 1144 | nl_msg_type_to_str(h->nlmsg_type), |
1145 | h->nlmsg_type, h->nlmsg_len, | |
1146 | h->nlmsg_seq, h->nlmsg_pid); | |
1147 | ||
783827ae DS |
1148 | |
1149 | /* | |
1150 | * Ignore messages that maybe sent from | |
1151 | * other actors besides the kernel | |
1152 | */ | |
1153 | if (snl.nl_pid != 0) { | |
43e52561 QY |
1154 | zlog_debug("Ignoring message from pid %u", |
1155 | snl.nl_pid); | |
d62a17ae | 1156 | continue; |
1157 | } | |
1158 | ||
2414abd3 | 1159 | error = (*filter)(h, zns->ns_id, startup); |
d62a17ae | 1160 | if (error < 0) { |
9df414fe QY |
1161 | zlog_debug("%s filter function error", |
1162 | nl->name); | |
d62a17ae | 1163 | ret = error; |
1164 | } | |
1165 | } | |
1166 | ||
1167 | /* After error care. */ | |
1168 | if (msg.msg_flags & MSG_TRUNC) { | |
e914ccbe | 1169 | flog_err(EC_ZEBRA_NETLINK_LENGTH_ERROR, |
1c50c1c0 | 1170 | "%s error: message truncated", nl->name); |
d62a17ae | 1171 | continue; |
1172 | } | |
1173 | if (status) { | |
e914ccbe | 1174 | flog_err(EC_ZEBRA_NETLINK_LENGTH_ERROR, |
1c50c1c0 QY |
1175 | "%s error: data remnant size %d", nl->name, |
1176 | status); | |
d62a17ae | 1177 | return -1; |
1178 | } | |
1179 | } | |
1180 | return ret; | |
1fdc9eae | 1181 | } |
1182 | ||
936ebf0a | 1183 | /* |
7cdb1a84 | 1184 | * netlink_talk_info |
936ebf0a DS |
1185 | * |
1186 | * sendmsg() to netlink socket then recvmsg(). | |
1187 | * Calls netlink_parse_info to parse returned data | |
1188 | * | |
1189 | * filter -> The filter to read final results from kernel | |
1190 | * nlmsghdr -> The data to send to the kernel | |
8b962e77 | 1191 | * dp_info -> The dataplane and netlink socket information |
936ebf0a DS |
1192 | * startup -> Are we reading in under startup conditions |
1193 | * This is passed through eventually to filter. | |
1194 | */ | |
2cf7651f DS |
1195 | static int netlink_talk_info(int (*filter)(struct nlmsghdr *, ns_id_t, |
1196 | int startup), | |
1197 | struct nlmsghdr *n, | |
1198 | struct zebra_dplane_info *dp_info, bool startup) | |
1fdc9eae | 1199 | { |
d4000d7b | 1200 | struct nlsock *nl; |
d62a17ae | 1201 | |
d4000d7b | 1202 | nl = kernel_netlink_nlsock_lookup(dp_info->sock); |
3670f504 | 1203 | n->nlmsg_seq = dp_info->seq; |
d62a17ae | 1204 | n->nlmsg_pid = nl->snl.nl_pid; |
1205 | ||
d62a17ae | 1206 | if (IS_ZEBRA_DEBUG_KERNEL) |
1207 | zlog_debug( | |
1208 | "netlink_talk: %s type %s(%u), len=%d seq=%u flags 0x%x", | |
1209 | nl->name, nl_msg_type_to_str(n->nlmsg_type), | |
1210 | n->nlmsg_type, n->nlmsg_len, n->nlmsg_seq, | |
1211 | n->nlmsg_flags); | |
1212 | ||
f8653393 | 1213 | if (netlink_send_msg(nl, n, n->nlmsg_len) == -1) |
d62a17ae | 1214 | return -1; |
d62a17ae | 1215 | |
d62a17ae | 1216 | /* |
1217 | * Get reply from netlink socket. | |
1218 | * The reply should either be an acknowlegement or an error. | |
1219 | */ | |
7cdb1a84 MS |
1220 | return netlink_parse_info(filter, nl, dp_info, 0, startup); |
1221 | } | |
1222 | ||
1223 | /* | |
1224 | * Synchronous version of netlink_talk_info. Converts args to suit the | |
1225 | * common version, which is suitable for both sync and async use. | |
7cdb1a84 MS |
1226 | */ |
1227 | int netlink_talk(int (*filter)(struct nlmsghdr *, ns_id_t, int startup), | |
1228 | struct nlmsghdr *n, struct nlsock *nl, struct zebra_ns *zns, | |
9bfadae8 | 1229 | bool startup) |
7cdb1a84 MS |
1230 | { |
1231 | struct zebra_dplane_info dp_info; | |
1232 | ||
1233 | /* Increment sequence number before capturing snapshot of ns socket | |
1234 | * info. | |
1235 | */ | |
1236 | nl->seq++; | |
1237 | ||
1238 | /* Capture info in intermediate info struct */ | |
85a75f1e | 1239 | zebra_dplane_info_from_zns(&dp_info, zns, (nl == &(zns->netlink_cmd))); |
7cdb1a84 | 1240 | |
5709131c | 1241 | return netlink_talk_info(filter, n, &dp_info, startup); |
1fdc9eae | 1242 | } |
1243 | ||
289602d7 | 1244 | /* Issue request message to kernel via netlink socket. GET messages |
1245 | * are issued through this interface. | |
1246 | */ | |
fd3f8e52 | 1247 | int netlink_request(struct nlsock *nl, void *req) |
1fdc9eae | 1248 | { |
fd3f8e52 | 1249 | struct nlmsghdr *n = (struct nlmsghdr *)req; |
d62a17ae | 1250 | |
1251 | /* Check netlink socket. */ | |
1252 | if (nl->sock < 0) { | |
450971aa | 1253 | flog_err_sys(EC_LIB_SOCKET, "%s socket isn't active.", |
09c866e3 | 1254 | nl->name); |
d62a17ae | 1255 | return -1; |
1256 | } | |
1257 | ||
1258 | /* Fill common fields for all requests. */ | |
d62a17ae | 1259 | n->nlmsg_pid = nl->snl.nl_pid; |
1260 | n->nlmsg_seq = ++nl->seq; | |
1261 | ||
f8653393 | 1262 | if (netlink_send_msg(nl, req, n->nlmsg_len) == -1) |
d62a17ae | 1263 | return -1; |
d62a17ae | 1264 | |
1265 | return 0; | |
1fdc9eae | 1266 | } |
1267 | ||
e63c7622 JU |
1268 | static int nl_batch_read_resp(struct nl_batch *bth) |
1269 | { | |
1270 | struct nlmsghdr *h; | |
1271 | struct sockaddr_nl snl; | |
9d06e121 | 1272 | struct msghdr msg = {}; |
f6feb48b | 1273 | int status, seq; |
d4000d7b | 1274 | struct nlsock *nl; |
f6feb48b JU |
1275 | struct zebra_dplane_ctx *ctx; |
1276 | bool ignore_msg; | |
e63c7622 | 1277 | |
d4000d7b | 1278 | nl = kernel_netlink_nlsock_lookup(bth->zns->sock); |
e63c7622 JU |
1279 | |
1280 | msg.msg_name = (void *)&snl; | |
1281 | msg.msg_namelen = sizeof(snl); | |
1282 | ||
2f9dbd3a JU |
1283 | /* |
1284 | * The responses are not batched, so we need to read and process one | |
1285 | * message at a time. | |
1286 | */ | |
1287 | while (true) { | |
2cf7651f | 1288 | status = netlink_recv_msg(nl, &msg); |
00249e25 DS |
1289 | /* |
1290 | * status == -1 is a full on failure somewhere | |
1291 | * since we don't know where the problem happened | |
1292 | * we must mark all as failed | |
1293 | * | |
1294 | * Else we mark everything as worked | |
1295 | * | |
1296 | */ | |
1297 | if (status == -1 || status == 0) { | |
1298 | while ((ctx = dplane_ctx_dequeue(&(bth->ctx_list))) != | |
1299 | NULL) { | |
1300 | if (status == -1) | |
1301 | dplane_ctx_set_status( | |
1302 | ctx, | |
1303 | ZEBRA_DPLANE_REQUEST_FAILURE); | |
1304 | dplane_ctx_enqueue_tail(bth->ctx_out_q, ctx); | |
1305 | } | |
2f9dbd3a | 1306 | return status; |
00249e25 | 1307 | } |
e63c7622 | 1308 | |
2cf7651f | 1309 | h = (struct nlmsghdr *)nl->buf; |
f6feb48b JU |
1310 | ignore_msg = false; |
1311 | seq = h->nlmsg_seq; | |
e63c7622 | 1312 | /* |
f6feb48b JU |
1313 | * Find the corresponding context object. Received responses are |
1314 | * in the same order as requests we sent, so we can simply | |
1315 | * iterate over the context list and match responses with | |
1316 | * requests at same time. | |
e63c7622 | 1317 | */ |
f6feb48b | 1318 | while (true) { |
c8453cd7 DS |
1319 | ctx = dplane_ctx_get_head(&(bth->ctx_list)); |
1320 | if (ctx == NULL) { | |
1321 | /* | |
1322 | * This is a situation where we have gotten | |
1323 | * into a bad spot. We need to know that | |
1324 | * this happens( does it? ) | |
1325 | */ | |
1326 | zlog_err( | |
1327 | "%s:WARNING Received netlink Response for an error and no Contexts to associate with it", | |
1328 | __func__); | |
e63c7622 | 1329 | break; |
c8453cd7 | 1330 | } |
f6feb48b JU |
1331 | |
1332 | /* | |
1333 | * 'update' context objects take two consecutive | |
1334 | * sequence numbers. | |
1335 | */ | |
3670f504 DS |
1336 | if (dplane_ctx_is_update(ctx) && |
1337 | dplane_ctx_get_ns(ctx)->seq + 1 == seq) { | |
f6feb48b JU |
1338 | /* |
1339 | * This is the situation where we get a response | |
1340 | * to a message that should be ignored. | |
1341 | */ | |
1342 | ignore_msg = true; | |
1343 | break; | |
1344 | } | |
c8453cd7 DS |
1345 | |
1346 | ctx = dplane_ctx_dequeue(&(bth->ctx_list)); | |
1347 | dplane_ctx_enqueue_tail(bth->ctx_out_q, ctx); | |
1348 | ||
1349 | /* We have found corresponding context object. */ | |
3670f504 | 1350 | if (dplane_ctx_get_ns(ctx)->seq == seq) |
c8453cd7 DS |
1351 | break; |
1352 | ||
3670f504 | 1353 | if (dplane_ctx_get_ns(ctx)->seq > seq) |
c8453cd7 | 1354 | zlog_warn( |
b9d95135 | 1355 | "%s:WARNING Received %u is less than any context on the queue ctx->seq %u", |
c8453cd7 | 1356 | __func__, seq, |
3670f504 | 1357 | dplane_ctx_get_ns(ctx)->seq); |
e63c7622 JU |
1358 | } |
1359 | ||
c8453cd7 DS |
1360 | if (ignore_msg) { |
1361 | /* | |
1362 | * If we ignore the message due to an update | |
1363 | * above we should still fricking decode the | |
1364 | * message for our operator to understand | |
1365 | * what is going on | |
1366 | */ | |
1367 | int err = netlink_parse_error(nl, h, bth->zns->is_cmd, | |
1368 | false); | |
1369 | ||
1370 | zlog_debug("%s: netlink error message seq=%d %d", | |
1371 | __func__, h->nlmsg_seq, err); | |
f6feb48b | 1372 | continue; |
c8453cd7 | 1373 | } |
f6feb48b | 1374 | |
e63c7622 JU |
1375 | /* |
1376 | * We received a message with the sequence number that isn't | |
1377 | * associated with any dplane context object. | |
1378 | */ | |
f6feb48b | 1379 | if (ctx == NULL) { |
4c99d413 MS |
1380 | if (IS_ZEBRA_DEBUG_KERNEL) |
1381 | zlog_debug( | |
1382 | "%s: skipping unassociated response, seq number %d NS %u", | |
1383 | __func__, h->nlmsg_seq, | |
1384 | bth->zns->ns_id); | |
e63c7622 JU |
1385 | continue; |
1386 | } | |
1387 | ||
1388 | if (h->nlmsg_type == NLMSG_ERROR) { | |
d166308b MS |
1389 | int err = netlink_parse_error(nl, h, bth->zns->is_cmd, |
1390 | false); | |
e63c7622 JU |
1391 | |
1392 | if (err == -1) | |
f6feb48b JU |
1393 | dplane_ctx_set_status( |
1394 | ctx, ZEBRA_DPLANE_REQUEST_FAILURE); | |
e63c7622 | 1395 | |
4c99d413 MS |
1396 | if (IS_ZEBRA_DEBUG_KERNEL) |
1397 | zlog_debug("%s: netlink error message seq=%d ", | |
1398 | __func__, h->nlmsg_seq); | |
e63c7622 JU |
1399 | continue; |
1400 | } | |
1401 | ||
1402 | /* | |
1403 | * If we get here then we did not receive neither the ack nor | |
1404 | * the error and instead received some other message in an | |
1405 | * unexpected way. | |
1406 | */ | |
4c99d413 MS |
1407 | if (IS_ZEBRA_DEBUG_KERNEL) |
1408 | zlog_debug("%s: ignoring message type 0x%04x(%s) NS %u", | |
1409 | __func__, h->nlmsg_type, | |
1410 | nl_msg_type_to_str(h->nlmsg_type), | |
1411 | bth->zns->ns_id); | |
e63c7622 JU |
1412 | } |
1413 | ||
1414 | return 0; | |
1415 | } | |
1416 | ||
1417 | static void nl_batch_reset(struct nl_batch *bth) | |
1418 | { | |
e63c7622 JU |
1419 | bth->buf_head = bth->buf; |
1420 | bth->curlen = 0; | |
1421 | bth->msgcnt = 0; | |
1422 | bth->zns = NULL; | |
1423 | ||
f6feb48b | 1424 | TAILQ_INIT(&(bth->ctx_list)); |
e63c7622 JU |
1425 | } |
1426 | ||
f6feb48b | 1427 | static void nl_batch_init(struct nl_batch *bth, struct dplane_ctx_q *ctx_out_q) |
e63c7622 | 1428 | { |
531c92b8 JU |
1429 | /* |
1430 | * If the size of the buffer has changed, free and then allocate a new | |
1431 | * one. | |
1432 | */ | |
1433 | size_t bufsize = | |
1434 | atomic_load_explicit(&nl_batch_bufsize, memory_order_relaxed); | |
1435 | if (bufsize != nl_batch_tx_bufsize) { | |
1436 | if (nl_batch_tx_buf) | |
1437 | XFREE(MTYPE_NL_BUF, nl_batch_tx_buf); | |
1438 | ||
1439 | nl_batch_tx_buf = XCALLOC(MTYPE_NL_BUF, bufsize); | |
1440 | nl_batch_tx_bufsize = bufsize; | |
1441 | } | |
1442 | ||
f6feb48b | 1443 | bth->buf = nl_batch_tx_buf; |
531c92b8 JU |
1444 | bth->bufsiz = bufsize; |
1445 | bth->limit = atomic_load_explicit(&nl_batch_send_threshold, | |
1446 | memory_order_relaxed); | |
e63c7622 | 1447 | |
f6feb48b | 1448 | bth->ctx_out_q = ctx_out_q; |
e63c7622 | 1449 | |
f6feb48b JU |
1450 | nl_batch_reset(bth); |
1451 | } | |
1452 | ||
1453 | static void nl_batch_send(struct nl_batch *bth) | |
1454 | { | |
1455 | struct zebra_dplane_ctx *ctx; | |
1456 | bool err = false; | |
e63c7622 | 1457 | |
f6feb48b | 1458 | if (bth->curlen != 0 && bth->zns != NULL) { |
d4000d7b DS |
1459 | struct nlsock *nl = |
1460 | kernel_netlink_nlsock_lookup(bth->zns->sock); | |
1461 | ||
f6feb48b JU |
1462 | if (IS_ZEBRA_DEBUG_KERNEL) |
1463 | zlog_debug("%s: %s, batch size=%zu, msg cnt=%zu", | |
d4000d7b | 1464 | __func__, nl->name, bth->curlen, |
f6feb48b | 1465 | bth->msgcnt); |
e63c7622 | 1466 | |
d4000d7b | 1467 | if (netlink_send_msg(nl, bth->buf, bth->curlen) == -1) |
e63c7622 | 1468 | err = true; |
e63c7622 | 1469 | |
f6feb48b JU |
1470 | if (!err) { |
1471 | if (nl_batch_read_resp(bth) == -1) | |
1472 | err = true; | |
1473 | } | |
1474 | } | |
e63c7622 | 1475 | |
f6feb48b JU |
1476 | /* Move remaining contexts to the outbound queue. */ |
1477 | while (true) { | |
1478 | ctx = dplane_ctx_dequeue(&(bth->ctx_list)); | |
1479 | if (ctx == NULL) | |
1480 | break; | |
e63c7622 | 1481 | |
f6feb48b JU |
1482 | if (err) |
1483 | dplane_ctx_set_status(ctx, | |
1484 | ZEBRA_DPLANE_REQUEST_FAILURE); | |
e63c7622 | 1485 | |
f6feb48b | 1486 | dplane_ctx_enqueue_tail(bth->ctx_out_q, ctx); |
e63c7622 JU |
1487 | } |
1488 | ||
1489 | nl_batch_reset(bth); | |
1490 | } | |
1491 | ||
e63c7622 JU |
1492 | enum netlink_msg_status netlink_batch_add_msg( |
1493 | struct nl_batch *bth, struct zebra_dplane_ctx *ctx, | |
1494 | ssize_t (*msg_encoder)(struct zebra_dplane_ctx *, void *, size_t), | |
f6feb48b | 1495 | bool ignore_res) |
e63c7622 JU |
1496 | { |
1497 | int seq; | |
1498 | ssize_t size; | |
1499 | struct nlmsghdr *msgh; | |
d4000d7b | 1500 | struct nlsock *nl; |
e63c7622 | 1501 | |
e63c7622 JU |
1502 | size = (*msg_encoder)(ctx, bth->buf_head, bth->bufsiz - bth->curlen); |
1503 | ||
1504 | /* | |
1505 | * If there was an error while encoding the message (other than buffer | |
1506 | * overflow) then return an error. | |
1507 | */ | |
1508 | if (size < 0) | |
1509 | return FRR_NETLINK_ERROR; | |
1510 | ||
1511 | /* | |
1512 | * If the message doesn't fit entirely in the buffer then send the batch | |
1513 | * and retry. | |
1514 | */ | |
1515 | if (size == 0) { | |
1516 | nl_batch_send(bth); | |
1517 | size = (*msg_encoder)(ctx, bth->buf_head, | |
1518 | bth->bufsiz - bth->curlen); | |
1519 | /* | |
1520 | * If the message doesn't fit in the empty buffer then just | |
1521 | * return an error. | |
1522 | */ | |
1523 | if (size <= 0) | |
1524 | return FRR_NETLINK_ERROR; | |
1525 | } | |
1526 | ||
3670f504 | 1527 | seq = dplane_ctx_get_ns(ctx)->seq; |
d4000d7b DS |
1528 | nl = kernel_netlink_nlsock_lookup(dplane_ctx_get_ns_sock(ctx)); |
1529 | ||
f6feb48b | 1530 | if (ignore_res) |
e63c7622 JU |
1531 | seq++; |
1532 | ||
1533 | msgh = (struct nlmsghdr *)bth->buf_head; | |
1534 | msgh->nlmsg_seq = seq; | |
d4000d7b | 1535 | msgh->nlmsg_pid = nl->snl.nl_pid; |
e63c7622 | 1536 | |
e63c7622 JU |
1537 | bth->zns = dplane_ctx_get_ns(ctx); |
1538 | bth->buf_head = ((char *)bth->buf_head) + size; | |
1539 | bth->curlen += size; | |
1540 | bth->msgcnt++; | |
1541 | ||
e63c7622 JU |
1542 | return FRR_NETLINK_QUEUED; |
1543 | } | |
1544 | ||
67e3369e JU |
1545 | static enum netlink_msg_status nl_put_msg(struct nl_batch *bth, |
1546 | struct zebra_dplane_ctx *ctx) | |
1547 | { | |
1548 | if (dplane_ctx_is_skip_kernel(ctx)) | |
1549 | return FRR_NETLINK_SUCCESS; | |
1550 | ||
1551 | switch (dplane_ctx_get_op(ctx)) { | |
1552 | ||
1553 | case DPLANE_OP_ROUTE_INSTALL: | |
1554 | case DPLANE_OP_ROUTE_UPDATE: | |
1555 | case DPLANE_OP_ROUTE_DELETE: | |
1556 | return netlink_put_route_update_msg(bth, ctx); | |
1557 | ||
1558 | case DPLANE_OP_NH_INSTALL: | |
1559 | case DPLANE_OP_NH_UPDATE: | |
1560 | case DPLANE_OP_NH_DELETE: | |
1561 | return netlink_put_nexthop_update_msg(bth, ctx); | |
1562 | ||
1563 | case DPLANE_OP_LSP_INSTALL: | |
1564 | case DPLANE_OP_LSP_UPDATE: | |
1565 | case DPLANE_OP_LSP_DELETE: | |
1566 | return netlink_put_lsp_update_msg(bth, ctx); | |
1567 | ||
1568 | case DPLANE_OP_PW_INSTALL: | |
1569 | case DPLANE_OP_PW_UNINSTALL: | |
1570 | return netlink_put_pw_update_msg(bth, ctx); | |
1571 | ||
1572 | case DPLANE_OP_ADDR_INSTALL: | |
1573 | case DPLANE_OP_ADDR_UNINSTALL: | |
1574 | return netlink_put_address_update_msg(bth, ctx); | |
1575 | ||
1576 | case DPLANE_OP_MAC_INSTALL: | |
1577 | case DPLANE_OP_MAC_DELETE: | |
1578 | return netlink_put_mac_update_msg(bth, ctx); | |
1579 | ||
1580 | case DPLANE_OP_NEIGH_INSTALL: | |
1581 | case DPLANE_OP_NEIGH_UPDATE: | |
1582 | case DPLANE_OP_NEIGH_DELETE: | |
1583 | case DPLANE_OP_VTEP_ADD: | |
1584 | case DPLANE_OP_VTEP_DELETE: | |
d68e74b4 | 1585 | case DPLANE_OP_NEIGH_DISCOVER: |
0a27a2fe PG |
1586 | case DPLANE_OP_NEIGH_IP_INSTALL: |
1587 | case DPLANE_OP_NEIGH_IP_DELETE: | |
e18747a9 | 1588 | case DPLANE_OP_NEIGH_TABLE_UPDATE: |
67e3369e JU |
1589 | return netlink_put_neigh_update_msg(bth, ctx); |
1590 | ||
1591 | case DPLANE_OP_RULE_ADD: | |
1592 | case DPLANE_OP_RULE_DELETE: | |
1593 | case DPLANE_OP_RULE_UPDATE: | |
1594 | return netlink_put_rule_update_msg(bth, ctx); | |
1595 | ||
1596 | case DPLANE_OP_SYS_ROUTE_ADD: | |
1597 | case DPLANE_OP_SYS_ROUTE_DELETE: | |
1598 | case DPLANE_OP_ROUTE_NOTIFY: | |
1599 | case DPLANE_OP_LSP_NOTIFY: | |
c60522f7 | 1600 | case DPLANE_OP_BR_PORT_UPDATE: |
67e3369e JU |
1601 | return FRR_NETLINK_SUCCESS; |
1602 | ||
5162e000 PG |
1603 | case DPLANE_OP_IPTABLE_ADD: |
1604 | case DPLANE_OP_IPTABLE_DELETE: | |
ef524230 PG |
1605 | case DPLANE_OP_IPSET_ADD: |
1606 | case DPLANE_OP_IPSET_DELETE: | |
1607 | case DPLANE_OP_IPSET_ENTRY_ADD: | |
1608 | case DPLANE_OP_IPSET_ENTRY_DELETE: | |
5162e000 PG |
1609 | return FRR_NETLINK_ERROR; |
1610 | ||
62b4b7e4 PG |
1611 | case DPLANE_OP_GRE_SET: |
1612 | return netlink_put_gre_set_msg(bth, ctx); | |
1613 | ||
9d59df63 MS |
1614 | case DPLANE_OP_INTF_ADDR_ADD: |
1615 | case DPLANE_OP_INTF_ADDR_DEL: | |
728f2017 | 1616 | case DPLANE_OP_INTF_NETCONFIG: |
67e3369e JU |
1617 | case DPLANE_OP_NONE: |
1618 | return FRR_NETLINK_ERROR; | |
5d414138 SW |
1619 | |
1620 | case DPLANE_OP_INTF_INSTALL: | |
1621 | case DPLANE_OP_INTF_UPDATE: | |
1622 | case DPLANE_OP_INTF_DELETE: | |
1623 | return netlink_put_intf_update_msg(bth, ctx); | |
67e3369e JU |
1624 | } |
1625 | ||
1626 | return FRR_NETLINK_ERROR; | |
1627 | } | |
1628 | ||
fef24b03 JU |
1629 | void kernel_update_multi(struct dplane_ctx_q *ctx_list) |
1630 | { | |
67e3369e JU |
1631 | struct nl_batch batch; |
1632 | struct zebra_dplane_ctx *ctx; | |
1633 | struct dplane_ctx_q handled_list; | |
1634 | enum netlink_msg_status res; | |
1635 | ||
67e3369e | 1636 | TAILQ_INIT(&handled_list); |
f6feb48b | 1637 | nl_batch_init(&batch, &handled_list); |
67e3369e JU |
1638 | |
1639 | while (true) { | |
1640 | ctx = dplane_ctx_dequeue(ctx_list); | |
1641 | if (ctx == NULL) | |
1642 | break; | |
1643 | ||
f6feb48b JU |
1644 | if (batch.zns != NULL |
1645 | && batch.zns->ns_id != dplane_ctx_get_ns(ctx)->ns_id) | |
1646 | nl_batch_send(&batch); | |
67e3369e JU |
1647 | |
1648 | /* | |
f6feb48b JU |
1649 | * Assume all messages will succeed and then mark only the ones |
1650 | * that failed. | |
67e3369e | 1651 | */ |
f6feb48b JU |
1652 | dplane_ctx_set_status(ctx, ZEBRA_DPLANE_REQUEST_SUCCESS); |
1653 | ||
1654 | res = nl_put_msg(&batch, ctx); | |
1655 | ||
1656 | dplane_ctx_enqueue_tail(&(batch.ctx_list), ctx); | |
1657 | if (res == FRR_NETLINK_ERROR) | |
67e3369e JU |
1658 | dplane_ctx_set_status(ctx, |
1659 | ZEBRA_DPLANE_REQUEST_FAILURE); | |
1660 | ||
f6feb48b JU |
1661 | if (batch.curlen > batch.limit) |
1662 | nl_batch_send(&batch); | |
67e3369e JU |
1663 | } |
1664 | ||
1665 | nl_batch_send(&batch); | |
1666 | ||
1667 | TAILQ_INIT(ctx_list); | |
1668 | dplane_ctx_list_append(ctx_list, &handled_list); | |
fef24b03 JU |
1669 | } |
1670 | ||
d4000d7b DS |
1671 | struct nlsock *kernel_netlink_nlsock_lookup(int sock) |
1672 | { | |
34869809 | 1673 | struct nlsock lookup, *retval; |
d4000d7b DS |
1674 | |
1675 | lookup.sock = sock; | |
1676 | ||
34869809 MS |
1677 | NLSOCK_LOCK(); |
1678 | retval = hash_lookup(nlsock_hash, &lookup); | |
1679 | NLSOCK_UNLOCK(); | |
1680 | ||
1681 | return retval; | |
1682 | } | |
1683 | ||
1684 | /* Insert nlsock entry into hash */ | |
1685 | static void kernel_netlink_nlsock_insert(struct nlsock *nls) | |
1686 | { | |
1687 | NLSOCK_LOCK(); | |
1688 | (void)hash_get(nlsock_hash, nls, hash_alloc_intern); | |
1689 | NLSOCK_UNLOCK(); | |
1690 | } | |
1691 | ||
1692 | /* Remove nlsock entry from hash */ | |
1693 | static void kernel_netlink_nlsock_remove(struct nlsock *nls) | |
1694 | { | |
1695 | NLSOCK_LOCK(); | |
1696 | (void)hash_release(nlsock_hash, nls); | |
1697 | NLSOCK_UNLOCK(); | |
d4000d7b DS |
1698 | } |
1699 | ||
1700 | static uint32_t kernel_netlink_nlsock_key(const void *arg) | |
1701 | { | |
1702 | const struct nlsock *nl = arg; | |
1703 | ||
1704 | return nl->sock; | |
1705 | } | |
1706 | ||
1707 | static bool kernel_netlink_nlsock_hash_equal(const void *arg1, const void *arg2) | |
1708 | { | |
1709 | const struct nlsock *nl1 = arg1; | |
1710 | const struct nlsock *nl2 = arg2; | |
1711 | ||
1712 | if (nl1->sock == nl2->sock) | |
1713 | return true; | |
1714 | ||
1715 | return false; | |
1716 | } | |
1717 | ||
1fdc9eae | 1718 | /* Exported interface function. This function simply calls |
1719 | netlink_socket (). */ | |
d62a17ae | 1720 | void kernel_init(struct zebra_ns *zns) |
1fdc9eae | 1721 | { |
47e2eb27 | 1722 | uint32_t groups, dplane_groups, ext_groups; |
5d307d5d DS |
1723 | #if defined SOL_NETLINK |
1724 | int one, ret; | |
1725 | #endif | |
d62a17ae | 1726 | |
026a316f DS |
1727 | /* |
1728 | * Initialize netlink sockets | |
1729 | * | |
1730 | * If RTMGRP_XXX exists use that, but at some point | |
1731 | * I think the kernel developers realized that | |
1732 | * keeping track of all the different values would | |
1733 | * lead to confusion, so we need to convert the | |
1734 | * RTNLGRP_XXX to a bit position for ourself | |
1735 | */ | |
1736 | groups = RTMGRP_LINK | | |
1737 | RTMGRP_IPV4_ROUTE | | |
1738 | RTMGRP_IPV4_IFADDR | | |
1739 | RTMGRP_IPV6_ROUTE | | |
1740 | RTMGRP_IPV6_IFADDR | | |
1741 | RTMGRP_IPV4_MROUTE | | |
1742 | RTMGRP_NEIGH | | |
67188ca2 QY |
1743 | ((uint32_t) 1 << (RTNLGRP_IPV4_RULE - 1)) | |
1744 | ((uint32_t) 1 << (RTNLGRP_IPV6_RULE - 1)) | | |
cd787a8a | 1745 | ((uint32_t) 1 << (RTNLGRP_NEXTHOP - 1)); |
d62a17ae | 1746 | |
ceab66b7 MS |
1747 | dplane_groups = (RTMGRP_LINK | |
1748 | RTMGRP_IPV4_IFADDR | | |
cd787a8a MS |
1749 | RTMGRP_IPV6_IFADDR | |
1750 | ((uint32_t) 1 << (RTNLGRP_IPV4_NETCONF - 1)) | | |
1751 | ((uint32_t) 1 << (RTNLGRP_IPV6_NETCONF - 1)) | | |
1752 | ((uint32_t) 1 << (RTNLGRP_MPLS_NETCONF - 1))); | |
1753 | ||
47e2eb27 CS |
1754 | /* Use setsockopt for > 31 group */ |
1755 | ext_groups = RTNLGRP_TUNNEL; | |
ceab66b7 | 1756 | |
d62a17ae | 1757 | snprintf(zns->netlink.name, sizeof(zns->netlink.name), |
1758 | "netlink-listen (NS %u)", zns->ns_id); | |
1759 | zns->netlink.sock = -1; | |
fe953d7c DS |
1760 | if (netlink_socket(&zns->netlink, groups, &ext_groups, 1, zns->ns_id) < |
1761 | 0) { | |
19d5a4fe DS |
1762 | zlog_err("Failure to create %s socket", |
1763 | zns->netlink.name); | |
1764 | exit(-1); | |
1765 | } | |
34869809 MS |
1766 | |
1767 | kernel_netlink_nlsock_insert(&zns->netlink); | |
d62a17ae | 1768 | |
1769 | snprintf(zns->netlink_cmd.name, sizeof(zns->netlink_cmd.name), | |
1770 | "netlink-cmd (NS %u)", zns->ns_id); | |
1771 | zns->netlink_cmd.sock = -1; | |
fe953d7c | 1772 | if (netlink_socket(&zns->netlink_cmd, 0, 0, 0, zns->ns_id) < 0) { |
19d5a4fe DS |
1773 | zlog_err("Failure to create %s socket", |
1774 | zns->netlink_cmd.name); | |
1775 | exit(-1); | |
1776 | } | |
34869809 MS |
1777 | |
1778 | kernel_netlink_nlsock_insert(&zns->netlink_cmd); | |
d62a17ae | 1779 | |
80dcc388 MS |
1780 | /* Outbound socket for dplane programming of the host OS. */ |
1781 | snprintf(zns->netlink_dplane_out.name, | |
1782 | sizeof(zns->netlink_dplane_out.name), "netlink-dp (NS %u)", | |
1783 | zns->ns_id); | |
1784 | zns->netlink_dplane_out.sock = -1; | |
fe953d7c | 1785 | if (netlink_socket(&zns->netlink_dplane_out, 0, 0, 0, zns->ns_id) < 0) { |
62b8bb7a | 1786 | zlog_err("Failure to create %s socket", |
80dcc388 MS |
1787 | zns->netlink_dplane_out.name); |
1788 | exit(-1); | |
1789 | } | |
34869809 MS |
1790 | |
1791 | kernel_netlink_nlsock_insert(&zns->netlink_dplane_out); | |
80dcc388 MS |
1792 | |
1793 | /* Inbound socket for OS events coming to the dplane. */ | |
1794 | snprintf(zns->netlink_dplane_in.name, | |
1795 | sizeof(zns->netlink_dplane_in.name), "netlink-dp-in (NS %u)", | |
1796 | zns->ns_id); | |
1797 | zns->netlink_dplane_in.sock = -1; | |
fe953d7c | 1798 | if (netlink_socket(&zns->netlink_dplane_in, dplane_groups, 0, 0, |
ceab66b7 | 1799 | zns->ns_id) < 0) { |
80dcc388 MS |
1800 | zlog_err("Failure to create %s socket", |
1801 | zns->netlink_dplane_in.name); | |
62b8bb7a MS |
1802 | exit(-1); |
1803 | } | |
34869809 MS |
1804 | |
1805 | kernel_netlink_nlsock_insert(&zns->netlink_dplane_in); | |
62b8bb7a | 1806 | |
5d307d5d DS |
1807 | /* |
1808 | * SOL_NETLINK is not available on all platforms yet | |
1809 | * apparently. It's in bits/socket.h which I am not | |
1810 | * sure that we want to pull into our build system. | |
1811 | */ | |
1812 | #if defined SOL_NETLINK | |
1813 | /* | |
1814 | * Let's tell the kernel that we want to receive extended | |
62b8bb7a | 1815 | * ACKS over our command socket(s) |
5d307d5d DS |
1816 | */ |
1817 | one = 1; | |
1818 | ret = setsockopt(zns->netlink_cmd.sock, SOL_NETLINK, NETLINK_EXT_ACK, | |
1819 | &one, sizeof(one)); | |
1820 | ||
1821 | if (ret < 0) | |
62b8bb7a MS |
1822 | zlog_notice("Registration for extended cmd ACK failed : %d %s", |
1823 | errno, safe_strerror(errno)); | |
1824 | ||
1825 | one = 1; | |
80dcc388 MS |
1826 | ret = setsockopt(zns->netlink_dplane_out.sock, SOL_NETLINK, |
1827 | NETLINK_EXT_ACK, &one, sizeof(one)); | |
62b8bb7a MS |
1828 | |
1829 | if (ret < 0) | |
1830 | zlog_notice("Registration for extended dp ACK failed : %d %s", | |
5d307d5d | 1831 | errno, safe_strerror(errno)); |
97f85144 JU |
1832 | |
1833 | /* | |
1834 | * Trim off the payload of the original netlink message in the | |
1835 | * acknowledgment. This option is available since Linux 4.2, so if | |
1836 | * setsockopt fails, ignore the error. | |
1837 | */ | |
1838 | one = 1; | |
80dcc388 MS |
1839 | ret = setsockopt(zns->netlink_dplane_out.sock, SOL_NETLINK, |
1840 | NETLINK_CAP_ACK, &one, sizeof(one)); | |
9781e6a0 DS |
1841 | if (ret < 0) |
1842 | zlog_notice( | |
1843 | "Registration for reduced ACK packet size failed, probably running an early kernel"); | |
5d307d5d DS |
1844 | #endif |
1845 | ||
d62a17ae | 1846 | /* Register kernel socket. */ |
19d5a4fe | 1847 | if (fcntl(zns->netlink.sock, F_SETFL, O_NONBLOCK) < 0) |
450971aa | 1848 | flog_err_sys(EC_LIB_SOCKET, "Can't set %s socket flags: %s", |
09c866e3 | 1849 | zns->netlink.name, safe_strerror(errno)); |
8c85e8ea DS |
1850 | |
1851 | if (fcntl(zns->netlink_cmd.sock, F_SETFL, O_NONBLOCK) < 0) | |
1852 | zlog_err("Can't set %s socket error: %s(%d)", | |
1853 | zns->netlink_cmd.name, safe_strerror(errno), errno); | |
19d5a4fe | 1854 | |
80dcc388 | 1855 | if (fcntl(zns->netlink_dplane_out.sock, F_SETFL, O_NONBLOCK) < 0) |
62b8bb7a | 1856 | zlog_err("Can't set %s socket error: %s(%d)", |
80dcc388 MS |
1857 | zns->netlink_dplane_out.name, safe_strerror(errno), |
1858 | errno); | |
1859 | ||
1860 | if (fcntl(zns->netlink_dplane_in.sock, F_SETFL, O_NONBLOCK) < 0) | |
1861 | zlog_err("Can't set %s socket error: %s(%d)", | |
1862 | zns->netlink_dplane_in.name, safe_strerror(errno), | |
1863 | errno); | |
62b8bb7a | 1864 | |
19d5a4fe | 1865 | /* Set receive buffer size if it's set from command line */ |
9fb83b55 DS |
1866 | if (rcvbufsize) { |
1867 | netlink_recvbuf(&zns->netlink, rcvbufsize); | |
1868 | netlink_recvbuf(&zns->netlink_cmd, rcvbufsize); | |
1869 | netlink_recvbuf(&zns->netlink_dplane_out, rcvbufsize); | |
1870 | netlink_recvbuf(&zns->netlink_dplane_in, rcvbufsize); | |
97f85144 | 1871 | } |
19d5a4fe | 1872 | |
80dcc388 MS |
1873 | /* Set filter for inbound sockets, to exclude events we've generated |
1874 | * ourselves. | |
1875 | */ | |
1876 | netlink_install_filter(zns->netlink.sock, zns->netlink_cmd.snl.nl_pid, | |
1877 | zns->netlink_dplane_out.snl.nl_pid); | |
1878 | ||
1879 | netlink_install_filter(zns->netlink_dplane_in.sock, | |
62b8bb7a | 1880 | zns->netlink_cmd.snl.nl_pid, |
80dcc388 | 1881 | zns->netlink_dplane_out.snl.nl_pid); |
62b8bb7a | 1882 | |
19d5a4fe DS |
1883 | zns->t_netlink = NULL; |
1884 | ||
3801e764 | 1885 | thread_add_read(zrouter.master, kernel_read, zns, |
19d5a4fe | 1886 | zns->netlink.sock, &zns->t_netlink); |
d62a17ae | 1887 | |
1888 | rt_netlink_init(); | |
1fdc9eae | 1889 | } |
1890 | ||
34869809 MS |
1891 | /* Helper to clean up an nlsock */ |
1892 | static void kernel_nlsock_fini(struct nlsock *nls) | |
1893 | { | |
1894 | if (nls && nls->sock >= 0) { | |
1895 | kernel_netlink_nlsock_remove(nls); | |
1896 | close(nls->sock); | |
1897 | nls->sock = -1; | |
1898 | XFREE(MTYPE_NL_BUF, nls->buf); | |
1899 | nls->buflen = 0; | |
1900 | } | |
1901 | } | |
1902 | ||
62b8bb7a | 1903 | void kernel_terminate(struct zebra_ns *zns, bool complete) |
1fdc9eae | 1904 | { |
50478845 | 1905 | thread_cancel(&zns->t_netlink); |
d62a17ae | 1906 | |
34869809 | 1907 | kernel_nlsock_fini(&zns->netlink); |
d62a17ae | 1908 | |
34869809 | 1909 | kernel_nlsock_fini(&zns->netlink_cmd); |
ddfeb486 | 1910 | |
34869809 | 1911 | kernel_nlsock_fini(&zns->netlink_dplane_in); |
80dcc388 | 1912 | |
62b8bb7a MS |
1913 | /* During zebra shutdown, we need to leave the dataplane socket |
1914 | * around until all work is done. | |
1915 | */ | |
34869809 MS |
1916 | if (complete) |
1917 | kernel_nlsock_fini(&zns->netlink_dplane_out); | |
1918 | } | |
d4000d7b | 1919 | |
34869809 MS |
1920 | /* |
1921 | * Global init for platform-/OS-specific things | |
1922 | */ | |
1923 | void kernel_router_init(void) | |
1924 | { | |
1925 | /* Init nlsock hash and lock */ | |
1926 | pthread_mutex_init(&nlsock_mutex, NULL); | |
1927 | nlsock_hash = hash_create_size(8, kernel_netlink_nlsock_key, | |
1928 | kernel_netlink_nlsock_hash_equal, | |
1929 | "Netlink Socket Hash"); | |
1930 | } | |
1931 | ||
1932 | /* | |
1933 | * Global deinit for platform-/OS-specific things | |
1934 | */ | |
1935 | void kernel_router_terminate(void) | |
1936 | { | |
1937 | pthread_mutex_destroy(&nlsock_mutex); | |
1938 | ||
1939 | hash_free(nlsock_hash); | |
1940 | nlsock_hash = NULL; | |
62b8bb7a | 1941 | } |
34869809 | 1942 | |
ddfeb486 | 1943 | #endif /* HAVE_NETLINK */ |