2 * lxc: linux Container library
4 * (C) Copyright IBM Corp. 2007, 2008
7 * Daniel Lezcano <daniel.lezcano at free.fr>
9 * This library is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
14 * This library is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with this library; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
33 #include <arpa/inet.h>
34 #include <linux/netlink.h>
35 #include <linux/rtnetlink.h>
36 #include <linux/sockios.h>
37 #include <net/ethernet.h>
39 #include <net/if_arp.h>
40 #include <netinet/in.h>
41 #include <sys/inotify.h>
42 #include <sys/ioctl.h>
43 #include <sys/param.h>
44 #include <sys/socket.h>
46 #include <sys/types.h>
60 #include <../include/ifaddrs.h>
64 #include "include/strlcpy.h"
67 lxc_log_define(network
, lxc
);
69 typedef int (*instantiate_cb
)(struct lxc_handler
*, struct lxc_netdev
*);
71 static int instantiate_veth(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
73 int bridge_index
, err
;
75 char veth1buf
[IFNAMSIZ
], veth2buf
[IFNAMSIZ
];
78 if (netdev
->priv
.veth_attr
.pair
[0] != '\0') {
79 veth1
= netdev
->priv
.veth_attr
.pair
;
80 if (handler
->conf
->reboot
)
81 lxc_netdev_delete_by_name(veth1
);
83 err
= snprintf(veth1buf
, sizeof(veth1buf
), "vethXXXXXX");
84 if (err
< 0 || (size_t)err
>= sizeof(veth1buf
))
87 veth1
= lxc_mkifname(veth1buf
);
91 /* store away for deconf */
92 memcpy(netdev
->priv
.veth_attr
.veth1
, veth1
, IFNAMSIZ
);
95 err
= snprintf(veth2buf
, sizeof(veth2buf
), "vethXXXXXX");
96 if (err
< 0 || (size_t)err
>= sizeof(veth2buf
))
99 veth2
= lxc_mkifname(veth2buf
);
103 err
= lxc_veth_create(veth1
, veth2
);
106 SYSERROR("Failed to create veth pair \"%s\" and \"%s\"", veth1
, veth2
);
110 /* changing the high byte of the mac address to 0xfe, the bridge interface
111 * will always keep the host's mac address and not take the mac address
113 err
= setup_private_host_hw_addr(veth1
);
116 SYSERROR("Failed to change mac address of host interface \"%s\"", veth1
);
120 /* Retrieve ifindex of the host's veth device. */
121 netdev
->priv
.veth_attr
.ifindex
= if_nametoindex(veth1
);
122 if (!netdev
->priv
.veth_attr
.ifindex
) {
123 ERROR("Failed to retrieve ifindex for \"%s\"", veth1
);
127 /* Note that we're retrieving the container's ifindex in the host's
128 * network namespace because we need it to move the device from the
129 * host's network namespace to the container's network namespace later
132 netdev
->ifindex
= if_nametoindex(veth2
);
133 if (!netdev
->ifindex
) {
134 ERROR("Failed to retrieve ifindex for \"%s\"", veth2
);
139 if (lxc_safe_uint(netdev
->mtu
, &mtu
) < 0)
140 WARN("Failed to parse mtu");
142 INFO("Retrieved mtu %d", mtu
);
143 } else if (netdev
->link
[0] != '\0') {
144 bridge_index
= if_nametoindex(netdev
->link
);
146 mtu
= netdev_get_mtu(bridge_index
);
147 INFO("Retrieved mtu %d from %s", mtu
, netdev
->link
);
149 mtu
= netdev_get_mtu(netdev
->ifindex
);
150 INFO("Retrieved mtu %d from %s", mtu
, veth2
);
155 err
= lxc_netdev_set_mtu(veth1
, mtu
);
157 err
= lxc_netdev_set_mtu(veth2
, mtu
);
161 SYSERROR("Failed to set mtu \"%d\" for veth pair \"%s\" "
162 "and \"%s\"", mtu
, veth1
, veth2
);
167 if (netdev
->link
[0] != '\0') {
168 err
= lxc_bridge_attach(netdev
->link
, veth1
);
171 SYSERROR("Failed to attach \"%s\" to bridge \"%s\"",
172 veth1
, netdev
->link
);
175 INFO("Attached \"%s\" to bridge \"%s\"", veth1
, netdev
->link
);
178 err
= lxc_netdev_up(veth1
);
181 SYSERROR("Failed to set \"%s\" up", veth1
);
185 if (netdev
->upscript
) {
193 err
= run_script_argv(handler
->name
,
194 handler
->conf
->hooks_version
, "net",
195 netdev
->upscript
, "up", argv
);
200 DEBUG("Instantiated veth \"%s/%s\", index is \"%d\"", veth1
, veth2
,
206 if (netdev
->ifindex
!= 0)
207 lxc_netdev_delete_by_name(veth1
);
211 static int instantiate_macvlan(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
213 char peerbuf
[IFNAMSIZ
], *peer
;
216 if (netdev
->link
[0] == '\0') {
217 ERROR("No link for macvlan network device specified");
221 err
= snprintf(peerbuf
, sizeof(peerbuf
), "mcXXXXXX");
222 if (err
< 0 || (size_t)err
>= sizeof(peerbuf
))
225 peer
= lxc_mkifname(peerbuf
);
229 err
= lxc_macvlan_create(netdev
->link
, peer
,
230 netdev
->priv
.macvlan_attr
.mode
);
233 SYSERROR("Failed to create macvlan interface \"%s\" on \"%s\"",
238 netdev
->ifindex
= if_nametoindex(peer
);
239 if (!netdev
->ifindex
) {
240 ERROR("Failed to retrieve ifindex for \"%s\"", peer
);
244 if (netdev
->upscript
) {
251 err
= run_script_argv(handler
->name
,
252 handler
->conf
->hooks_version
, "net",
253 netdev
->upscript
, "up", argv
);
258 DEBUG("Instantiated macvlan \"%s\" with ifindex is %d and mode %d",
259 peer
, netdev
->ifindex
, netdev
->priv
.macvlan_attr
.mode
);
264 lxc_netdev_delete_by_name(peer
);
268 static int instantiate_vlan(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
272 static uint16_t vlan_cntr
= 0;
273 unsigned int mtu
= 0;
275 if (netdev
->link
[0] == '\0') {
276 ERROR("No link for vlan network device specified");
280 err
= snprintf(peer
, sizeof(peer
), "vlan%d-%d", netdev
->priv
.vlan_attr
.vid
, vlan_cntr
++);
281 if (err
< 0 || (size_t)err
>= sizeof(peer
))
284 err
= lxc_vlan_create(netdev
->link
, peer
, netdev
->priv
.vlan_attr
.vid
);
287 SYSERROR("Failed to create vlan interface \"%s\" on \"%s\"",
292 netdev
->ifindex
= if_nametoindex(peer
);
293 if (!netdev
->ifindex
) {
294 ERROR("Failed to retrieve ifindex for \"%s\"", peer
);
295 lxc_netdev_delete_by_name(peer
);
299 DEBUG("Instantiated vlan \"%s\" with ifindex is \"%d\" (vlan1000)",
300 peer
, netdev
->ifindex
);
302 if (lxc_safe_uint(netdev
->mtu
, &mtu
) < 0) {
303 ERROR("Failed to retrieve mtu from \"%d\"/\"%s\".",
305 netdev
->name
[0] != '\0' ? netdev
->name
: "(null)");
309 err
= lxc_netdev_set_mtu(peer
, mtu
);
312 SYSERROR("Failed to set mtu \"%s\" for \"%s\"",
314 lxc_netdev_delete_by_name(peer
);
322 static int instantiate_phys(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
331 if (netdev
->link
[0] == '\0') {
332 ERROR("No link for physical interface specified");
336 /* Note that we're retrieving the container's ifindex in the host's
337 * network namespace because we need it to move the device from the
338 * host's network namespace to the container's network namespace later
340 * Note that netdev->link will contain the name of the physical network
341 * device in the host's namespace.
343 netdev
->ifindex
= if_nametoindex(netdev
->link
);
344 if (!netdev
->ifindex
) {
345 ERROR("Failed to retrieve ifindex for \"%s\"", netdev
->link
);
349 /* Store the ifindex of the host's network device in the host's
352 netdev
->priv
.phys_attr
.ifindex
= netdev
->ifindex
;
354 if (!netdev
->upscript
)
357 ret
= run_script_argv(handler
->name
, handler
->conf
->hooks_version
,
358 "net", netdev
->upscript
, "up", argv
);
365 static int instantiate_empty(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
374 if (!netdev
->upscript
)
377 ret
= run_script_argv(handler
->name
, handler
->conf
->hooks_version
,
378 "net", netdev
->upscript
, "up", argv
);
385 static int instantiate_none(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
391 static instantiate_cb netdev_conf
[LXC_NET_MAXCONFTYPE
+ 1] = {
392 [LXC_NET_VETH
] = instantiate_veth
,
393 [LXC_NET_MACVLAN
] = instantiate_macvlan
,
394 [LXC_NET_VLAN
] = instantiate_vlan
,
395 [LXC_NET_PHYS
] = instantiate_phys
,
396 [LXC_NET_EMPTY
] = instantiate_empty
,
397 [LXC_NET_NONE
] = instantiate_none
,
400 static int shutdown_veth(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
410 if (!netdev
->downscript
)
413 if (netdev
->priv
.veth_attr
.pair
[0] != '\0')
414 argv
[2] = netdev
->priv
.veth_attr
.pair
;
416 argv
[2] = netdev
->priv
.veth_attr
.veth1
;
418 ret
= run_script_argv(handler
->name
,
419 handler
->conf
->hooks_version
, "net",
420 netdev
->downscript
, "down", argv
);
427 static int shutdown_macvlan(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
436 if (!netdev
->downscript
)
439 ret
= run_script_argv(handler
->name
, handler
->conf
->hooks_version
,
440 "net", netdev
->downscript
, "down", argv
);
447 static int shutdown_vlan(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
452 static int shutdown_phys(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
461 if (!netdev
->downscript
)
464 ret
= run_script_argv(handler
->name
, handler
->conf
->hooks_version
,
465 "net", netdev
->downscript
, "down", argv
);
472 static int shutdown_empty(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
480 if (!netdev
->downscript
)
483 ret
= run_script_argv(handler
->name
, handler
->conf
->hooks_version
,
484 "net", netdev
->downscript
, "down", argv
);
491 static int shutdown_none(struct lxc_handler
*handler
, struct lxc_netdev
*netdev
)
496 static instantiate_cb netdev_deconf
[LXC_NET_MAXCONFTYPE
+ 1] = {
497 [LXC_NET_VETH
] = shutdown_veth
,
498 [LXC_NET_MACVLAN
] = shutdown_macvlan
,
499 [LXC_NET_VLAN
] = shutdown_vlan
,
500 [LXC_NET_PHYS
] = shutdown_phys
,
501 [LXC_NET_EMPTY
] = shutdown_empty
,
502 [LXC_NET_NONE
] = shutdown_none
,
505 int lxc_netdev_move_by_index(int ifindex
, pid_t pid
, const char *ifname
)
508 struct nl_handler nlh
;
509 struct ifinfomsg
*ifi
;
510 struct nlmsg
*nlmsg
= NULL
;
512 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
517 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
521 nlmsg
->nlmsghdr
->nlmsg_flags
= NLM_F_REQUEST
| NLM_F_ACK
;
522 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWLINK
;
524 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
527 ifi
->ifi_family
= AF_UNSPEC
;
528 ifi
->ifi_index
= ifindex
;
530 if (nla_put_u32(nlmsg
, IFLA_NET_NS_PID
, pid
))
533 if (ifname
!= NULL
) {
534 if (nla_put_string(nlmsg
, IFLA_IFNAME
, ifname
))
538 err
= netlink_transaction(&nlh
, nlmsg
, nlmsg
);
545 /* If we are asked to move a wireless interface, then we must actually move its
546 * phyN device. Detect that condition and return the physname here. The physname
547 * will be passed to lxc_netdev_move_wlan() which will free it when done.
549 #define PHYSNAME "/sys/class/net/%s/phy80211/name"
550 static char *is_wlan(const char *ifname
)
557 char *physname
= NULL
;
559 len
= strlen(ifname
) + strlen(PHYSNAME
) - 1;
560 path
= alloca(len
+ 1);
561 ret
= snprintf(path
, len
, PHYSNAME
, ifname
);
562 if (ret
< 0 || (size_t)ret
>= len
)
565 f
= fopen(path
, "r");
569 /* Feh - sb.st_size is always 4096. */
570 fseek(f
, 0, SEEK_END
);
572 fseek(f
, 0, SEEK_SET
);
578 physname
= malloc(physlen
+ 1);
584 memset(physname
, 0, physlen
+ 1);
585 ret
= fread(physname
, 1, physlen
, f
);
590 for (i
= 0; i
< physlen
; i
++) {
591 if (physname
[i
] == '\n')
594 if (physname
[i
] == '\0')
605 static int lxc_netdev_rename_by_name_in_netns(pid_t pid
, const char *old
,
615 return wait_for_pid(fpid
);
617 if (!switch_to_ns(pid
, "net"))
620 _exit(lxc_netdev_rename_by_name(old
, new));
623 static int lxc_netdev_move_wlan(char *physname
, const char *ifname
, pid_t pid
,
630 /* Move phyN into the container. TODO - do this using netlink.
631 * However, IIUC this involves a bit more complicated work to talk to
632 * the 80211 module, so for now just call out to iw.
634 cmd
= on_path("iw", NULL
);
645 sprintf(pidstr
, "%d", pid
);
646 execlp("iw", "iw", "phy", physname
, "set", "netns", pidstr
,
651 if (wait_for_pid(fpid
))
656 err
= lxc_netdev_rename_by_name_in_netns(pid
, ifname
, newname
);
663 int lxc_netdev_move_by_name(const char *ifname
, pid_t pid
, const char* newname
)
671 index
= if_nametoindex(ifname
);
675 physname
= is_wlan(ifname
);
677 return lxc_netdev_move_wlan(physname
, ifname
, pid
, newname
);
679 return lxc_netdev_move_by_index(index
, pid
, newname
);
682 int lxc_netdev_delete_by_index(int ifindex
)
685 struct ifinfomsg
*ifi
;
686 struct nl_handler nlh
;
687 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
689 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
694 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
698 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
702 nlmsg
->nlmsghdr
->nlmsg_flags
= NLM_F_ACK
| NLM_F_REQUEST
;
703 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_DELLINK
;
705 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
708 ifi
->ifi_family
= AF_UNSPEC
;
709 ifi
->ifi_index
= ifindex
;
711 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
719 int lxc_netdev_delete_by_name(const char *name
)
723 index
= if_nametoindex(name
);
727 return lxc_netdev_delete_by_index(index
);
730 int lxc_netdev_rename_by_index(int ifindex
, const char *newname
)
733 struct ifinfomsg
*ifi
;
734 struct nl_handler nlh
;
735 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
737 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
741 len
= strlen(newname
);
742 if (len
== 1 || len
>= IFNAMSIZ
)
746 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
750 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
754 nlmsg
->nlmsghdr
->nlmsg_flags
= NLM_F_ACK
| NLM_F_REQUEST
;
755 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWLINK
;
757 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
760 ifi
->ifi_family
= AF_UNSPEC
;
761 ifi
->ifi_index
= ifindex
;
763 if (nla_put_string(nlmsg
, IFLA_IFNAME
, newname
))
766 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
774 int lxc_netdev_rename_by_name(const char *oldname
, const char *newname
)
778 len
= strlen(oldname
);
779 if (len
== 1 || len
>= IFNAMSIZ
)
782 index
= if_nametoindex(oldname
);
786 return lxc_netdev_rename_by_index(index
, newname
);
789 int netdev_set_flag(const char *name
, int flag
)
792 struct ifinfomsg
*ifi
;
793 struct nl_handler nlh
;
794 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
796 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
802 if (len
== 1 || len
>= IFNAMSIZ
)
806 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
810 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
815 index
= if_nametoindex(name
);
819 nlmsg
->nlmsghdr
->nlmsg_flags
= NLM_F_REQUEST
| NLM_F_ACK
;
820 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWLINK
;
822 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
827 ifi
->ifi_family
= AF_UNSPEC
;
828 ifi
->ifi_index
= index
;
829 ifi
->ifi_change
|= IFF_UP
;
830 ifi
->ifi_flags
|= flag
;
832 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
840 int netdev_get_flag(const char *name
, int *flag
)
843 struct ifinfomsg
*ifi
;
844 struct nl_handler nlh
;
845 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
850 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
856 if (len
== 1 || len
>= IFNAMSIZ
)
860 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
864 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
869 index
= if_nametoindex(name
);
873 nlmsg
->nlmsghdr
->nlmsg_flags
= NLM_F_REQUEST
;
874 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_GETLINK
;
876 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
881 ifi
->ifi_family
= AF_UNSPEC
;
882 ifi
->ifi_index
= index
;
884 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
888 ifi
= NLMSG_DATA(answer
->nlmsghdr
);
890 *flag
= ifi
->ifi_flags
;
899 * \brief Check a interface is up or not.
901 * \param name: name for the interface.
904 * 0 means interface is down.
905 * 1 means interface is up.
906 * Others means error happened, and ret-value is the error number.
908 int lxc_netdev_isup(const char *name
)
912 err
= netdev_get_flag(name
, &flag
);
922 int netdev_get_mtu(int ifindex
)
924 int answer_len
, err
, res
;
925 struct nl_handler nlh
;
926 struct ifinfomsg
*ifi
;
927 struct nlmsghdr
*msg
;
928 int readmore
= 0, recv_len
= 0;
929 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
931 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
936 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
940 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
944 /* Save the answer buffer length, since it will be overwritten
945 * on the first receive (and we might need to receive more than
948 answer_len
= answer
->nlmsghdr
->nlmsg_len
;
950 nlmsg
->nlmsghdr
->nlmsg_flags
= NLM_F_REQUEST
| NLM_F_DUMP
;
951 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_GETLINK
;
953 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
956 ifi
->ifi_family
= AF_UNSPEC
;
958 /* Send the request for addresses, which returns all addresses
959 * on all interfaces. */
960 err
= netlink_send(&nlh
, nlmsg
);
965 /* Restore the answer buffer length, it might have been
966 * overwritten by a previous receive.
968 answer
->nlmsghdr
->nlmsg_len
= answer_len
;
970 /* Get the (next) batch of reply messages */
971 err
= netlink_rcv(&nlh
, answer
);
977 /* Satisfy the typing for the netlink macros */
978 msg
= answer
->nlmsghdr
;
980 while (NLMSG_OK(msg
, recv_len
)) {
982 /* Stop reading if we see an error message */
983 if (msg
->nlmsg_type
== NLMSG_ERROR
) {
984 struct nlmsgerr
*errmsg
=
985 (struct nlmsgerr
*)NLMSG_DATA(msg
);
990 /* Stop reading if we see a NLMSG_DONE message */
991 if (msg
->nlmsg_type
== NLMSG_DONE
) {
996 ifi
= NLMSG_DATA(msg
);
997 if (ifi
->ifi_index
== ifindex
) {
998 struct rtattr
*rta
= IFLA_RTA(ifi
);
1000 msg
->nlmsg_len
- NLMSG_LENGTH(sizeof(*ifi
));
1002 while (RTA_OK(rta
, attr_len
)) {
1003 /* Found a local address for the
1004 * requested interface, return it.
1006 if (rta
->rta_type
== IFLA_MTU
) {
1007 memcpy(&res
, RTA_DATA(rta
),
1012 rta
= RTA_NEXT(rta
, attr_len
);
1016 /* Keep reading more data from the socket if the last
1017 * message had the NLF_F_MULTI flag set.
1019 readmore
= (msg
->nlmsg_flags
& NLM_F_MULTI
);
1021 /* Look at the next message received in this buffer. */
1022 msg
= NLMSG_NEXT(msg
, recv_len
);
1026 /* If we end up here, we didn't find any result, so signal an error. */
1030 netlink_close(&nlh
);
1036 int lxc_netdev_set_mtu(const char *name
, int mtu
)
1038 int err
, index
, len
;
1039 struct ifinfomsg
*ifi
;
1040 struct nl_handler nlh
;
1041 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
1043 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
1049 if (len
== 1 || len
>= IFNAMSIZ
)
1053 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
1057 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
1062 index
= if_nametoindex(name
);
1066 nlmsg
->nlmsghdr
->nlmsg_flags
= NLM_F_REQUEST
| NLM_F_ACK
;
1067 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWLINK
;
1069 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
1074 ifi
->ifi_family
= AF_UNSPEC
;
1075 ifi
->ifi_index
= index
;
1077 if (nla_put_u32(nlmsg
, IFLA_MTU
, mtu
))
1080 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
1082 netlink_close(&nlh
);
1088 int lxc_netdev_up(const char *name
)
1090 return netdev_set_flag(name
, IFF_UP
);
1093 int lxc_netdev_down(const char *name
)
1095 return netdev_set_flag(name
, 0);
1098 int lxc_veth_create(const char *name1
, const char *name2
)
1101 struct ifinfomsg
*ifi
;
1102 struct nl_handler nlh
;
1103 struct rtattr
*nest1
, *nest2
, *nest3
;
1104 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
1106 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
1111 len
= strlen(name1
);
1112 if (len
== 1 || len
>= IFNAMSIZ
)
1115 len
= strlen(name2
);
1116 if (len
== 1 || len
>= IFNAMSIZ
)
1120 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
1124 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
1128 nlmsg
->nlmsghdr
->nlmsg_flags
=
1129 NLM_F_REQUEST
| NLM_F_CREATE
| NLM_F_EXCL
| NLM_F_ACK
;
1130 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWLINK
;
1132 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
1135 ifi
->ifi_family
= AF_UNSPEC
;
1138 nest1
= nla_begin_nested(nlmsg
, IFLA_LINKINFO
);
1142 if (nla_put_string(nlmsg
, IFLA_INFO_KIND
, "veth"))
1145 nest2
= nla_begin_nested(nlmsg
, IFLA_INFO_DATA
);
1149 nest3
= nla_begin_nested(nlmsg
, VETH_INFO_PEER
);
1153 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
1159 if (nla_put_string(nlmsg
, IFLA_IFNAME
, name2
))
1162 nla_end_nested(nlmsg
, nest3
);
1163 nla_end_nested(nlmsg
, nest2
);
1164 nla_end_nested(nlmsg
, nest1
);
1166 if (nla_put_string(nlmsg
, IFLA_IFNAME
, name1
))
1169 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
1171 netlink_close(&nlh
);
1177 /* TODO: merge with lxc_macvlan_create */
1178 int lxc_vlan_create(const char *master
, const char *name
, unsigned short vlanid
)
1180 int err
, len
, lindex
;
1181 struct ifinfomsg
*ifi
;
1182 struct nl_handler nlh
;
1183 struct rtattr
*nest
, *nest2
;
1184 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
1186 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
1191 len
= strlen(master
);
1192 if (len
== 1 || len
>= IFNAMSIZ
)
1196 if (len
== 1 || len
>= IFNAMSIZ
)
1200 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
1204 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
1209 lindex
= if_nametoindex(master
);
1213 nlmsg
->nlmsghdr
->nlmsg_flags
=
1214 NLM_F_REQUEST
| NLM_F_CREATE
| NLM_F_EXCL
| NLM_F_ACK
;
1215 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWLINK
;
1217 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
1222 ifi
->ifi_family
= AF_UNSPEC
;
1224 nest
= nla_begin_nested(nlmsg
, IFLA_LINKINFO
);
1228 if (nla_put_string(nlmsg
, IFLA_INFO_KIND
, "vlan"))
1231 nest2
= nla_begin_nested(nlmsg
, IFLA_INFO_DATA
);
1235 if (nla_put_u16(nlmsg
, IFLA_VLAN_ID
, vlanid
))
1238 nla_end_nested(nlmsg
, nest2
);
1239 nla_end_nested(nlmsg
, nest
);
1241 if (nla_put_u32(nlmsg
, IFLA_LINK
, lindex
))
1244 if (nla_put_string(nlmsg
, IFLA_IFNAME
, name
))
1247 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
1253 netlink_close(&nlh
);
1257 int lxc_macvlan_create(const char *master
, const char *name
, int mode
)
1259 int err
, index
, len
;
1260 struct ifinfomsg
*ifi
;
1261 struct nl_handler nlh
;
1262 struct rtattr
*nest
, *nest2
;
1263 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
1265 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
1270 len
= strlen(master
);
1271 if (len
== 1 || len
>= IFNAMSIZ
)
1275 if (len
== 1 || len
>= IFNAMSIZ
)
1279 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
1283 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
1288 index
= if_nametoindex(master
);
1292 nlmsg
->nlmsghdr
->nlmsg_flags
=
1293 NLM_F_REQUEST
| NLM_F_CREATE
| NLM_F_EXCL
| NLM_F_ACK
;
1294 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWLINK
;
1296 ifi
= nlmsg_reserve(nlmsg
, sizeof(struct ifinfomsg
));
1301 ifi
->ifi_family
= AF_UNSPEC
;
1303 nest
= nla_begin_nested(nlmsg
, IFLA_LINKINFO
);
1307 if (nla_put_string(nlmsg
, IFLA_INFO_KIND
, "macvlan"))
1311 nest2
= nla_begin_nested(nlmsg
, IFLA_INFO_DATA
);
1315 if (nla_put_u32(nlmsg
, IFLA_MACVLAN_MODE
, mode
))
1318 nla_end_nested(nlmsg
, nest2
);
1321 nla_end_nested(nlmsg
, nest
);
1323 if (nla_put_u32(nlmsg
, IFLA_LINK
, index
))
1326 if (nla_put_string(nlmsg
, IFLA_IFNAME
, name
))
1329 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
1331 netlink_close(&nlh
);
1337 static int proc_sys_net_write(const char *path
, const char *value
)
1342 fd
= open(path
, O_WRONLY
);
1346 if (lxc_write_nointr(fd
, value
, strlen(value
)) < 0)
1353 static int neigh_proxy_set(const char *ifname
, int family
, int flag
)
1356 char path
[MAXPATHLEN
];
1358 if (family
!= AF_INET
&& family
!= AF_INET6
)
1361 ret
= snprintf(path
, MAXPATHLEN
, "/proc/sys/net/%s/conf/%s/%s",
1362 family
== AF_INET
? "ipv4" : "ipv6", ifname
,
1363 family
== AF_INET
? "proxy_arp" : "proxy_ndp");
1364 if (ret
< 0 || (size_t)ret
>= MAXPATHLEN
)
1367 return proc_sys_net_write(path
, flag
? "1" : "0");
1370 int lxc_neigh_proxy_on(const char *name
, int family
)
1372 return neigh_proxy_set(name
, family
, 1);
1375 int lxc_neigh_proxy_off(const char *name
, int family
)
1377 return neigh_proxy_set(name
, family
, 0);
1380 int lxc_convert_mac(char *macaddr
, struct sockaddr
*sockaddr
)
1385 unsigned char *data
;
1387 sockaddr
->sa_family
= ARPHRD_ETHER
;
1388 data
= (unsigned char *)sockaddr
->sa_data
;
1390 while ((*macaddr
!= '\0') && (i
< ETH_ALEN
)) {
1394 else if (c
>= 'a' && c
<= 'f')
1396 else if (c
>= 'A' && c
<= 'F')
1405 else if (c
>= 'a' && c
<= 'f')
1406 val
|= c
- 'a' + 10;
1407 else if (c
>= 'A' && c
<= 'F')
1408 val
|= c
- 'A' + 10;
1409 else if (c
== ':' || c
== 0)
1415 *data
++ = (unsigned char)(val
& 0377);
1418 if (*macaddr
== ':')
1425 static int ip_addr_add(int family
, int ifindex
, void *addr
, void *bcast
,
1426 void *acast
, int prefix
)
1429 struct ifaddrmsg
*ifa
;
1430 struct nl_handler nlh
;
1431 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
1433 addrlen
= family
== AF_INET
? sizeof(struct in_addr
)
1434 : sizeof(struct in6_addr
);
1436 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
1441 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
1445 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
1449 nlmsg
->nlmsghdr
->nlmsg_flags
=
1450 NLM_F_ACK
| NLM_F_REQUEST
| NLM_F_CREATE
| NLM_F_EXCL
;
1451 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWADDR
;
1453 ifa
= nlmsg_reserve(nlmsg
, sizeof(struct ifaddrmsg
));
1456 ifa
->ifa_prefixlen
= prefix
;
1457 ifa
->ifa_index
= ifindex
;
1458 ifa
->ifa_family
= family
;
1462 if (nla_put_buffer(nlmsg
, IFA_LOCAL
, addr
, addrlen
))
1465 if (nla_put_buffer(nlmsg
, IFA_ADDRESS
, addr
, addrlen
))
1468 if (nla_put_buffer(nlmsg
, IFA_BROADCAST
, bcast
, addrlen
))
1471 /* TODO: multicast, anycast with ipv6 */
1472 err
= -EPROTONOSUPPORT
;
1473 if (family
== AF_INET6
&&
1474 (memcmp(bcast
, &in6addr_any
, sizeof(in6addr_any
)) ||
1475 memcmp(acast
, &in6addr_any
, sizeof(in6addr_any
))))
1478 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
1480 netlink_close(&nlh
);
1486 int lxc_ipv6_addr_add(int ifindex
, struct in6_addr
*addr
,
1487 struct in6_addr
*mcast
, struct in6_addr
*acast
,
1490 return ip_addr_add(AF_INET6
, ifindex
, addr
, mcast
, acast
, prefix
);
1493 int lxc_ipv4_addr_add(int ifindex
, struct in_addr
*addr
, struct in_addr
*bcast
,
1496 return ip_addr_add(AF_INET
, ifindex
, addr
, bcast
, NULL
, prefix
);
1499 /* Find an IFA_LOCAL (or IFA_ADDRESS if not IFA_LOCAL is present) address from
1500 * the given RTM_NEWADDR message. Allocates memory for the address and stores
1501 * that pointer in *res (so res should be an in_addr** or in6_addr**).
1503 static int ifa_get_local_ip(int family
, struct nlmsghdr
*msg
, void **res
)
1506 struct ifaddrmsg
*ifa
= NLMSG_DATA(msg
);
1507 struct rtattr
*rta
= IFA_RTA(ifa
);
1508 int attr_len
= NLMSG_PAYLOAD(msg
, sizeof(struct ifaddrmsg
));
1510 if (ifa
->ifa_family
!= family
)
1513 addrlen
= family
== AF_INET
? sizeof(struct in_addr
)
1514 : sizeof(struct in6_addr
);
1516 /* Loop over the rtattr's in this message */
1517 while (RTA_OK(rta
, attr_len
)) {
1518 /* Found a local address for the requested interface,
1521 if (rta
->rta_type
== IFA_LOCAL
||
1522 rta
->rta_type
== IFA_ADDRESS
) {
1523 /* Sanity check. The family check above should make sure
1524 * the address length is correct, but check here just in
1527 if (RTA_PAYLOAD(rta
) != addrlen
)
1530 /* We might have found an IFA_ADDRESS before, which we
1531 * now overwrite with an IFA_LOCAL.
1534 *res
= malloc(addrlen
);
1539 memcpy(*res
, RTA_DATA(rta
), addrlen
);
1540 if (rta
->rta_type
== IFA_LOCAL
)
1543 rta
= RTA_NEXT(rta
, attr_len
);
1548 static int ip_addr_get(int family
, int ifindex
, void **res
)
1550 int answer_len
, err
;
1551 struct ifaddrmsg
*ifa
;
1552 struct nl_handler nlh
;
1553 struct nlmsghdr
*msg
;
1554 int readmore
= 0, recv_len
= 0;
1555 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
1557 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
1562 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
1566 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
1570 /* Save the answer buffer length, since it will be overwritten on the
1571 * first receive (and we might need to receive more than once).
1573 answer_len
= answer
->nlmsghdr
->nlmsg_len
;
1575 nlmsg
->nlmsghdr
->nlmsg_flags
= NLM_F_REQUEST
| NLM_F_ROOT
;
1576 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_GETADDR
;
1578 ifa
= nlmsg_reserve(nlmsg
, sizeof(struct ifaddrmsg
));
1581 ifa
->ifa_family
= family
;
1583 /* Send the request for addresses, which returns all addresses on all
1586 err
= netlink_send(&nlh
, nlmsg
);
1591 /* Restore the answer buffer length, it might have been
1592 * overwritten by a previous receive.
1594 answer
->nlmsghdr
->nlmsg_len
= answer_len
;
1596 /* Get the (next) batch of reply messages. */
1597 err
= netlink_rcv(&nlh
, answer
);
1604 /* Satisfy the typing for the netlink macros. */
1605 msg
= answer
->nlmsghdr
;
1607 while (NLMSG_OK(msg
, recv_len
)) {
1608 /* Stop reading if we see an error message. */
1609 if (msg
->nlmsg_type
== NLMSG_ERROR
) {
1610 struct nlmsgerr
*errmsg
=
1611 (struct nlmsgerr
*)NLMSG_DATA(msg
);
1612 err
= errmsg
->error
;
1616 /* Stop reading if we see a NLMSG_DONE message. */
1617 if (msg
->nlmsg_type
== NLMSG_DONE
) {
1622 if (msg
->nlmsg_type
!= RTM_NEWADDR
) {
1627 ifa
= (struct ifaddrmsg
*)NLMSG_DATA(msg
);
1628 if (ifa
->ifa_index
== ifindex
) {
1629 if (ifa_get_local_ip(family
, msg
, res
) < 0) {
1634 /* Found a result, stop searching. */
1639 /* Keep reading more data from the socket if the last
1640 * message had the NLF_F_MULTI flag set.
1642 readmore
= (msg
->nlmsg_flags
& NLM_F_MULTI
);
1644 /* Look at the next message received in this buffer. */
1645 msg
= NLMSG_NEXT(msg
, recv_len
);
1649 /* If we end up here, we didn't find any result, so signal an
1655 netlink_close(&nlh
);
1661 int lxc_ipv6_addr_get(int ifindex
, struct in6_addr
**res
)
1663 return ip_addr_get(AF_INET6
, ifindex
, (void **)res
);
1666 int lxc_ipv4_addr_get(int ifindex
, struct in_addr
**res
)
1668 return ip_addr_get(AF_INET
, ifindex
, (void **)res
);
1671 static int ip_gateway_add(int family
, int ifindex
, void *gw
)
1674 struct nl_handler nlh
;
1676 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
1678 addrlen
= family
== AF_INET
? sizeof(struct in_addr
)
1679 : sizeof(struct in6_addr
);
1681 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
1686 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
1690 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
1694 nlmsg
->nlmsghdr
->nlmsg_flags
=
1695 NLM_F_ACK
| NLM_F_REQUEST
| NLM_F_CREATE
| NLM_F_EXCL
;
1696 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWROUTE
;
1698 rt
= nlmsg_reserve(nlmsg
, sizeof(struct rtmsg
));
1701 rt
->rtm_family
= family
;
1702 rt
->rtm_table
= RT_TABLE_MAIN
;
1703 rt
->rtm_scope
= RT_SCOPE_UNIVERSE
;
1704 rt
->rtm_protocol
= RTPROT_BOOT
;
1705 rt
->rtm_type
= RTN_UNICAST
;
1706 /* "default" destination */
1707 rt
->rtm_dst_len
= 0;
1710 if (nla_put_buffer(nlmsg
, RTA_GATEWAY
, gw
, addrlen
))
1713 /* Adding the interface index enables the use of link-local
1714 * addresses for the gateway.
1716 if (nla_put_u32(nlmsg
, RTA_OIF
, ifindex
))
1719 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
1721 netlink_close(&nlh
);
1727 int lxc_ipv4_gateway_add(int ifindex
, struct in_addr
*gw
)
1729 return ip_gateway_add(AF_INET
, ifindex
, gw
);
1732 int lxc_ipv6_gateway_add(int ifindex
, struct in6_addr
*gw
)
1734 return ip_gateway_add(AF_INET6
, ifindex
, gw
);
1737 static int ip_route_dest_add(int family
, int ifindex
, void *dest
)
1740 struct nl_handler nlh
;
1742 struct nlmsg
*answer
= NULL
, *nlmsg
= NULL
;
1744 addrlen
= family
== AF_INET
? sizeof(struct in_addr
)
1745 : sizeof(struct in6_addr
);
1747 err
= netlink_open(&nlh
, NETLINK_ROUTE
);
1752 nlmsg
= nlmsg_alloc(NLMSG_GOOD_SIZE
);
1756 answer
= nlmsg_alloc_reserve(NLMSG_GOOD_SIZE
);
1760 nlmsg
->nlmsghdr
->nlmsg_flags
=
1761 NLM_F_ACK
| NLM_F_REQUEST
| NLM_F_CREATE
| NLM_F_EXCL
;
1762 nlmsg
->nlmsghdr
->nlmsg_type
= RTM_NEWROUTE
;
1764 rt
= nlmsg_reserve(nlmsg
, sizeof(struct rtmsg
));
1767 rt
->rtm_family
= family
;
1768 rt
->rtm_table
= RT_TABLE_MAIN
;
1769 rt
->rtm_scope
= RT_SCOPE_LINK
;
1770 rt
->rtm_protocol
= RTPROT_BOOT
;
1771 rt
->rtm_type
= RTN_UNICAST
;
1772 rt
->rtm_dst_len
= addrlen
* 8;
1775 if (nla_put_buffer(nlmsg
, RTA_DST
, dest
, addrlen
))
1777 if (nla_put_u32(nlmsg
, RTA_OIF
, ifindex
))
1779 err
= netlink_transaction(&nlh
, nlmsg
, answer
);
1781 netlink_close(&nlh
);
1787 int lxc_ipv4_dest_add(int ifindex
, struct in_addr
*dest
)
1789 return ip_route_dest_add(AF_INET
, ifindex
, dest
);
1792 int lxc_ipv6_dest_add(int ifindex
, struct in6_addr
*dest
)
1794 return ip_route_dest_add(AF_INET6
, ifindex
, dest
);
1797 bool is_ovs_bridge(const char *bridge
)
1801 char brdirname
[22 + IFNAMSIZ
+ 1] = {0};
1803 ret
= snprintf(brdirname
, 22 + IFNAMSIZ
+ 1, "/sys/class/net/%s/bridge",
1805 if (ret
< 0 || (size_t)ret
>= 22 + IFNAMSIZ
+ 1)
1808 ret
= stat(brdirname
, &sb
);
1809 if (ret
< 0 && errno
== ENOENT
)
1815 struct ovs_veth_args
{
1820 /* Called from a background thread - when nic goes away, remove it from the
1823 static int lxc_ovs_delete_port_exec(void *data
)
1825 struct ovs_veth_args
*args
= data
;
1827 execlp("ovs-vsctl", "ovs-vsctl", "del-port", args
->bridge
, args
->nic
,
1832 int lxc_ovs_delete_port(const char *bridge
, const char *nic
)
1835 char cmd_output
[MAXPATHLEN
];
1836 struct ovs_veth_args args
;
1838 args
.bridge
= bridge
;
1840 ret
= run_command(cmd_output
, sizeof(cmd_output
),
1841 lxc_ovs_delete_port_exec
, (void *)&args
);
1843 ERROR("Failed to delete \"%s\" from openvswitch bridge \"%s\": "
1844 "%s", bridge
, nic
, cmd_output
);
1851 static int lxc_ovs_attach_bridge_exec(void *data
)
1853 struct ovs_veth_args
*args
= data
;
1855 execlp("ovs-vsctl", "ovs-vsctl", "add-port", args
->bridge
, args
->nic
,
1860 static int lxc_ovs_attach_bridge(const char *bridge
, const char *nic
)
1863 char cmd_output
[MAXPATHLEN
];
1864 struct ovs_veth_args args
;
1866 args
.bridge
= bridge
;
1868 ret
= run_command(cmd_output
, sizeof(cmd_output
),
1869 lxc_ovs_attach_bridge_exec
, (void *)&args
);
1871 ERROR("Failed to attach \"%s\" to openvswitch bridge \"%s\": %s",
1872 bridge
, nic
, cmd_output
);
1879 int lxc_bridge_attach(const char *bridge
, const char *ifname
)
1885 if (strlen(ifname
) >= IFNAMSIZ
)
1888 index
= if_nametoindex(ifname
);
1892 if (is_ovs_bridge(bridge
))
1893 return lxc_ovs_attach_bridge(bridge
, ifname
);
1895 fd
= socket(AF_INET
, SOCK_STREAM
, 0);
1899 retlen
= strlcpy(ifr
.ifr_name
, bridge
, IFNAMSIZ
);
1900 if (retlen
>= IFNAMSIZ
) {
1905 ifr
.ifr_name
[IFNAMSIZ
- 1] = '\0';
1906 ifr
.ifr_ifindex
= index
;
1907 err
= ioctl(fd
, SIOCBRADDIF
, &ifr
);
1915 static const char *const lxc_network_types
[LXC_NET_MAXCONFTYPE
+ 1] = {
1916 [LXC_NET_EMPTY
] = "empty",
1917 [LXC_NET_VETH
] = "veth",
1918 [LXC_NET_MACVLAN
] = "macvlan",
1919 [LXC_NET_PHYS
] = "phys",
1920 [LXC_NET_VLAN
] = "vlan",
1921 [LXC_NET_NONE
] = "none",
1924 const char *lxc_net_type_to_str(int type
)
1926 if (type
< 0 || type
> LXC_NET_MAXCONFTYPE
)
1929 return lxc_network_types
[type
];
1932 static const char padchar
[] = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ";
1934 char *lxc_mkifname(char *template)
1937 struct ifaddrs
*ifa
, *ifaddr
;
1938 char name
[IFNAMSIZ
];
1939 bool exists
= false;
1944 seed
= randseed(false);
1947 (void)randseed(true);
1950 if (strlen(template) >= IFNAMSIZ
)
1953 /* Get all the network interfaces. */
1954 ret
= getifaddrs(&ifaddr
);
1956 SYSERROR("Failed to get network interfaces");
1960 /* Generate random names until we find one that doesn't exist. */
1963 (void)strlcpy(name
, template, IFNAMSIZ
);
1967 for (i
= 0; i
< strlen(name
); i
++) {
1968 if (name
[i
] == 'X') {
1970 name
[i
] = padchar
[rand_r(&seed
) % (strlen(padchar
) - 1)];
1972 name
[i
] = padchar
[rand() % (strlen(padchar
) - 1)];
1977 for (ifa
= ifaddr
; ifa
!= NULL
; ifa
= ifa
->ifa_next
) {
1978 if (!strcmp(ifa
->ifa_name
, name
)) {
1988 freeifaddrs(ifaddr
);
1989 (void)strlcpy(template, name
, strlen(template) + 1);
1994 int setup_private_host_hw_addr(char *veth1
)
1999 sockfd
= socket(AF_INET
, SOCK_DGRAM
, 0);
2003 err
= snprintf((char *)ifr
.ifr_name
, IFNAMSIZ
, "%s", veth1
);
2004 if (err
< 0 || (size_t)err
>= IFNAMSIZ
) {
2009 err
= ioctl(sockfd
, SIOCGIFHWADDR
, &ifr
);
2015 ifr
.ifr_hwaddr
.sa_data
[0] = 0xfe;
2016 err
= ioctl(sockfd
, SIOCSIFHWADDR
, &ifr
);
2024 int lxc_find_gateway_addresses(struct lxc_handler
*handler
)
2026 struct lxc_list
*network
= &handler
->conf
->network
;
2027 struct lxc_list
*iterator
;
2028 struct lxc_netdev
*netdev
;
2031 lxc_list_for_each(iterator
, network
) {
2032 netdev
= iterator
->elem
;
2034 if (!netdev
->ipv4_gateway_auto
&& !netdev
->ipv6_gateway_auto
)
2037 if (netdev
->type
!= LXC_NET_VETH
&& netdev
->type
!= LXC_NET_MACVLAN
) {
2038 ERROR("Automatic gateway detection is only supported "
2039 "for veth and macvlan");
2043 if (netdev
->link
[0] == '\0') {
2044 ERROR("Automatic gateway detection needs a link interface");
2048 link_index
= if_nametoindex(netdev
->link
);
2052 if (netdev
->ipv4_gateway_auto
) {
2053 if (lxc_ipv4_addr_get(link_index
, &netdev
->ipv4_gateway
)) {
2054 ERROR("Failed to automatically find ipv4 gateway "
2055 "address from link interface \"%s\"", netdev
->link
);
2060 if (netdev
->ipv6_gateway_auto
) {
2061 if (lxc_ipv6_addr_get(link_index
, &netdev
->ipv6_gateway
)) {
2062 ERROR("Failed to automatically find ipv6 gateway "
2063 "address from link interface \"%s\"", netdev
->link
);
2072 #define LXC_USERNIC_PATH LIBEXECDIR "/lxc/lxc-user-nic"
2073 static int lxc_create_network_unpriv_exec(const char *lxcpath
, const char *lxcname
,
2074 struct lxc_netdev
*netdev
, pid_t pid
, unsigned int hooks_version
)
2078 int bytes
, pipefd
[2];
2079 char *token
, *saveptr
= NULL
;
2080 char netdev_link
[IFNAMSIZ
];
2081 char buffer
[MAXPATHLEN
] = {0};
2084 if (netdev
->type
!= LXC_NET_VETH
) {
2085 ERROR("Network type %d not support for unprivileged use", netdev
->type
);
2091 SYSERROR("Failed to create pipe");
2097 SYSERROR("Failed to create new process");
2106 char pidstr
[INTTYPE_TO_STRLEN(pid_t
)];
2110 ret
= dup2(pipefd
[1], STDOUT_FILENO
);
2112 ret
= dup2(pipefd
[1], STDERR_FILENO
);
2115 SYSERROR("Failed to duplicate std{err,out} file descriptor");
2116 _exit(EXIT_FAILURE
);
2119 if (netdev
->link
[0] != '\0')
2120 retlen
= strlcpy(netdev_link
, netdev
->link
, IFNAMSIZ
);
2122 retlen
= strlcpy(netdev_link
, "none", IFNAMSIZ
);
2123 if (retlen
>= IFNAMSIZ
) {
2124 SYSERROR("Invalid network device name");
2125 _exit(EXIT_FAILURE
);
2128 ret
= snprintf(pidstr
, sizeof(pidstr
), "%d", pid
);
2129 if (ret
< 0 || ret
>= sizeof(pidstr
))
2130 _exit(EXIT_FAILURE
);
2131 pidstr
[sizeof(pidstr
) - 1] = '\0';
2133 INFO("Execing lxc-user-nic create %s %s %s veth %s %s", lxcpath
,
2134 lxcname
, pidstr
, netdev_link
,
2135 netdev
->name
[0] != '\0' ? netdev
->name
: "(null)");
2136 if (netdev
->name
[0] != '\0')
2137 execlp(LXC_USERNIC_PATH
, LXC_USERNIC_PATH
, "create",
2138 lxcpath
, lxcname
, pidstr
, "veth", netdev_link
,
2139 netdev
->name
, (char *)NULL
);
2141 execlp(LXC_USERNIC_PATH
, LXC_USERNIC_PATH
, "create",
2142 lxcpath
, lxcname
, pidstr
, "veth", netdev_link
,
2144 SYSERROR("Failed to execute lxc-user-nic");
2145 _exit(EXIT_FAILURE
);
2148 /* close the write-end of the pipe */
2151 bytes
= lxc_read_nointr(pipefd
[0], &buffer
, MAXPATHLEN
);
2153 SYSERROR("Failed to read from pipe file descriptor");
2156 buffer
[bytes
- 1] = '\0';
2159 ret
= wait_for_pid(child
);
2161 if (ret
!= 0 || bytes
< 0) {
2162 ERROR("lxc-user-nic failed to configure requested network: %s",
2163 buffer
[0] != '\0' ? buffer
: "(null)");
2166 TRACE("Received output \"%s\" from lxc-user-nic", buffer
);
2169 token
= strtok_r(buffer
, ":", &saveptr
);
2171 ERROR("Failed to parse lxc-user-nic output");
2175 memset(netdev
->name
, 0, IFNAMSIZ
);
2176 memcpy(netdev
->name
, token
, IFNAMSIZ
- 1);
2178 /* netdev->ifindex */
2179 token
= strtok_r(NULL
, ":", &saveptr
);
2181 ERROR("Failed to parse lxc-user-nic output");
2185 ret
= lxc_safe_int(token
, &netdev
->ifindex
);
2188 SYSERROR("Failed to convert string \"%s\" to integer", token
);
2192 /* netdev->priv.veth_attr.veth1 */
2193 token
= strtok_r(NULL
, ":", &saveptr
);
2195 ERROR("Failed to parse lxc-user-nic output");
2199 retlen
= strlcpy(netdev
->priv
.veth_attr
.veth1
, token
, IFNAMSIZ
);
2200 if (retlen
>= IFNAMSIZ
) {
2201 ERROR("Host side veth device name returned by lxc-user-nic is "
2206 /* netdev->priv.veth_attr.ifindex */
2207 token
= strtok_r(NULL
, ":", &saveptr
);
2209 ERROR("Failed to parse lxc-user-nic output");
2213 ret
= lxc_safe_int(token
, &netdev
->priv
.veth_attr
.ifindex
);
2216 SYSERROR("Failed to convert string \"%s\" to integer", token
);
2220 if (netdev
->upscript
) {
2224 netdev
->priv
.veth_attr
.veth1
,
2228 ret
= run_script_argv(lxcname
,
2229 hooks_version
, "net",
2230 netdev
->upscript
, "up", argv
);
2238 static int lxc_delete_network_unpriv_exec(const char *lxcpath
, const char *lxcname
,
2239 struct lxc_netdev
*netdev
,
2240 const char *netns_path
)
2245 char buffer
[MAXPATHLEN
] = {0};
2247 if (netdev
->type
!= LXC_NET_VETH
) {
2248 ERROR("Network type %d not support for unprivileged use", netdev
->type
);
2254 SYSERROR("Failed to create pipe");
2260 SYSERROR("Failed to create new process");
2272 ret
= dup2(pipefd
[1], STDOUT_FILENO
);
2274 ret
= dup2(pipefd
[1], STDERR_FILENO
);
2277 SYSERROR("Failed to duplicate std{err,out} file descriptor");
2278 _exit(EXIT_FAILURE
);
2281 if (netdev
->priv
.veth_attr
.pair
[0] != '\0')
2282 hostveth
= netdev
->priv
.veth_attr
.pair
;
2284 hostveth
= netdev
->priv
.veth_attr
.veth1
;
2285 if (hostveth
[0] == '\0') {
2286 SYSERROR("Host side veth device name is missing");
2287 _exit(EXIT_FAILURE
);
2290 if (netdev
->link
[0] == '\0') {
2291 SYSERROR("Network link for network device \"%s\" is "
2292 "missing", netdev
->priv
.veth_attr
.veth1
);
2293 _exit(EXIT_FAILURE
);
2296 INFO("Execing lxc-user-nic delete %s %s %s veth %s %s", lxcpath
,
2297 lxcname
, netns_path
, netdev
->link
, hostveth
);
2298 execlp(LXC_USERNIC_PATH
, LXC_USERNIC_PATH
, "delete", lxcpath
,
2299 lxcname
, netns_path
, "veth", netdev
->link
, hostveth
,
2301 SYSERROR("Failed to exec lxc-user-nic.");
2302 _exit(EXIT_FAILURE
);
2307 bytes
= lxc_read_nointr(pipefd
[0], &buffer
, MAXPATHLEN
);
2309 SYSERROR("Failed to read from pipe file descriptor.");
2312 buffer
[bytes
- 1] = '\0';
2315 ret
= wait_for_pid(child
);
2317 if (ret
!= 0 || bytes
< 0) {
2318 ERROR("lxc-user-nic failed to delete requested network: %s",
2319 buffer
[0] != '\0' ? buffer
: "(null)");
2326 bool lxc_delete_network_unpriv(struct lxc_handler
*handler
)
2329 struct lxc_list
*iterator
;
2330 struct lxc_list
*network
= &handler
->conf
->network
;
2331 /* strlen("/proc/") = 6
2333 * INTTYPE_TO_STRLEN(pid_t)
2335 * strlen("/fd/") = 4
2337 * INTTYPE_TO_STRLEN(int)
2341 char netns_path
[6 + INTTYPE_TO_STRLEN(pid_t
) + 4 + INTTYPE_TO_STRLEN(int) + 1];
2345 if (handler
->nsfd
[LXC_NS_NET
] < 0) {
2346 DEBUG("Cannot not guarantee safe deletion of network devices. "
2347 "Manual cleanup maybe needed");
2351 ret
= snprintf(netns_path
, sizeof(netns_path
), "/proc/%d/fd/%d",
2352 lxc_raw_getpid(), handler
->nsfd
[LXC_NS_NET
]);
2353 if (ret
< 0 || ret
>= sizeof(netns_path
))
2356 lxc_list_for_each(iterator
, network
) {
2357 char *hostveth
= NULL
;
2358 struct lxc_netdev
*netdev
= iterator
->elem
;
2360 /* We can only delete devices whose ifindex we have. If we don't
2361 * have the index it means that we didn't create it.
2363 if (!netdev
->ifindex
)
2366 if (netdev
->type
== LXC_NET_PHYS
) {
2367 ret
= lxc_netdev_rename_by_index(netdev
->ifindex
,
2370 WARN("Failed to rename interface with index %d "
2371 "to its initial name \"%s\"",
2372 netdev
->ifindex
, netdev
->link
);
2374 TRACE("Renamed interface with index %d to its "
2375 "initial name \"%s\"",
2376 netdev
->ifindex
, netdev
->link
);
2377 goto clear_ifindices
;
2380 ret
= netdev_deconf
[netdev
->type
](handler
, netdev
);
2382 WARN("Failed to deconfigure network device");
2384 if (netdev
->type
!= LXC_NET_VETH
)
2385 goto clear_ifindices
;
2387 if (netdev
->link
[0] == '\0' || !is_ovs_bridge(netdev
->link
))
2388 goto clear_ifindices
;
2390 if (netdev
->priv
.veth_attr
.pair
[0] != '\0')
2391 hostveth
= netdev
->priv
.veth_attr
.pair
;
2393 hostveth
= netdev
->priv
.veth_attr
.veth1
;
2394 if (hostveth
[0] == '\0')
2395 goto clear_ifindices
;
2397 ret
= lxc_delete_network_unpriv_exec(handler
->lxcpath
,
2398 handler
->name
, netdev
,
2401 WARN("Failed to remove port \"%s\" from openvswitch "
2402 "bridge \"%s\"", hostveth
, netdev
->link
);
2403 goto clear_ifindices
;
2405 INFO("Removed interface \"%s\" from \"%s\"", hostveth
,
2409 /* We need to clear any ifindeces we recorded so liblxc won't
2410 * have cached stale data which would cause it to fail on reboot
2411 * we're we don't re-read the on-disk config file.
2413 netdev
->ifindex
= 0;
2414 if (netdev
->type
== LXC_NET_PHYS
) {
2415 netdev
->priv
.phys_attr
.ifindex
= 0;
2416 } else if (netdev
->type
== LXC_NET_VETH
) {
2417 netdev
->priv
.veth_attr
.veth1
[0] = '\0';
2418 netdev
->priv
.veth_attr
.ifindex
= 0;
2425 int lxc_create_network_priv(struct lxc_handler
*handler
)
2427 struct lxc_list
*iterator
;
2428 struct lxc_list
*network
= &handler
->conf
->network
;
2430 if (!handler
->am_root
)
2433 lxc_list_for_each(iterator
, network
) {
2434 struct lxc_netdev
*netdev
= iterator
->elem
;
2436 if (netdev
->type
< 0 || netdev
->type
> LXC_NET_MAXCONFTYPE
) {
2437 ERROR("Invalid network configuration type %d", netdev
->type
);
2441 if (netdev_conf
[netdev
->type
](handler
, netdev
)) {
2442 ERROR("Failed to create network device");
2451 int lxc_network_move_created_netdev_priv(const char *lxcpath
, const char *lxcname
,
2452 struct lxc_list
*network
, pid_t pid
)
2455 char ifname
[IFNAMSIZ
];
2456 struct lxc_list
*iterator
;
2458 if (am_guest_unpriv())
2461 lxc_list_for_each(iterator
, network
) {
2462 struct lxc_netdev
*netdev
= iterator
->elem
;
2464 if (!netdev
->ifindex
)
2467 /* retrieve the name of the interface */
2468 if (!if_indextoname(netdev
->ifindex
, ifname
)) {
2469 ERROR("No interface corresponding to ifindex \"%d\"",
2474 ret
= lxc_netdev_move_by_name(ifname
, pid
, NULL
);
2477 SYSERROR("Failed to move network device \"%s\" to "
2478 "network namespace %d", ifname
, pid
);
2482 DEBUG("Moved network device \"%s\"/\"%s\" to network namespace "
2484 ifname
, netdev
->name
[0] != '\0' ? netdev
->name
: "(null)",
2491 int lxc_create_network_unpriv(const char *lxcpath
, const char *lxcname
,
2492 struct lxc_list
*network
, pid_t pid
, unsigned int hooks_version
)
2494 struct lxc_list
*iterator
;
2496 if (!am_guest_unpriv())
2499 lxc_list_for_each(iterator
, network
) {
2500 struct lxc_netdev
*netdev
= iterator
->elem
;
2502 if (netdev
->type
== LXC_NET_EMPTY
)
2505 if (netdev
->type
== LXC_NET_NONE
)
2508 if (netdev
->type
!= LXC_NET_VETH
) {
2509 ERROR("Networks of type %s are not supported by "
2510 "unprivileged containers",
2511 lxc_net_type_to_str(netdev
->type
));
2516 INFO("mtu ignored due to insufficient privilege");
2518 if (lxc_create_network_unpriv_exec(lxcpath
, lxcname
, netdev
, pid
, hooks_version
))
2525 bool lxc_delete_network_priv(struct lxc_handler
*handler
)
2528 struct lxc_list
*iterator
;
2529 struct lxc_list
*network
= &handler
->conf
->network
;
2531 lxc_list_for_each(iterator
, network
) {
2532 char *hostveth
= NULL
;
2533 struct lxc_netdev
*netdev
= iterator
->elem
;
2535 /* We can only delete devices whose ifindex we have. If we don't
2536 * have the index it means that we didn't create it.
2538 if (!netdev
->ifindex
)
2541 if (netdev
->type
== LXC_NET_PHYS
) {
2542 ret
= lxc_netdev_rename_by_index(netdev
->ifindex
, netdev
->link
);
2544 WARN("Failed to rename interface with index %d "
2545 "from \"%s\" to its initial name \"%s\"",
2546 netdev
->ifindex
, netdev
->name
, netdev
->link
);
2548 TRACE("Renamed interface with index %d from "
2549 "\"%s\" to its initial name \"%s\"",
2550 netdev
->ifindex
, netdev
->name
,
2552 goto clear_ifindices
;
2555 ret
= netdev_deconf
[netdev
->type
](handler
, netdev
);
2557 WARN("Failed to deconfigure network device");
2559 /* Recent kernels remove the virtual interfaces when the network
2560 * namespace is destroyed but in case we did not move the
2561 * interface to the network namespace, we have to destroy it.
2563 ret
= lxc_netdev_delete_by_index(netdev
->ifindex
);
2564 if (-ret
== ENODEV
) {
2565 INFO("Interface \"%s\" with index %d already "
2566 "deleted or existing in different network "
2568 netdev
->name
[0] != '\0' ? netdev
->name
: "(null)",
2570 } else if (ret
< 0) {
2572 SYSWARN("Failed to remove interface \"%s\" with index %d",
2573 netdev
->name
[0] != '\0' ? netdev
->name
: "(null)",
2575 goto clear_ifindices
;
2577 INFO("Removed interface \"%s\" with index %d",
2578 netdev
->name
[0] != '\0' ? netdev
->name
: "(null)",
2581 if (netdev
->type
!= LXC_NET_VETH
)
2582 goto clear_ifindices
;
2584 /* Explicitly delete host veth device to prevent lingering
2585 * devices. We had issues in LXD around this.
2587 if (netdev
->priv
.veth_attr
.pair
[0] != '\0')
2588 hostveth
= netdev
->priv
.veth_attr
.pair
;
2590 hostveth
= netdev
->priv
.veth_attr
.veth1
;
2591 if (hostveth
[0] == '\0')
2592 goto clear_ifindices
;
2594 ret
= lxc_netdev_delete_by_name(hostveth
);
2597 SYSWARN("Failed to remove interface \"%s\" from \"%s\"",
2598 hostveth
, netdev
->link
);
2599 goto clear_ifindices
;
2601 INFO("Removed interface \"%s\" from \"%s\"", hostveth
, netdev
->link
);
2603 if (netdev
->link
[0] == '\0' || !is_ovs_bridge(netdev
->link
)) {
2604 netdev
->priv
.veth_attr
.veth1
[0] = '\0';
2605 netdev
->ifindex
= 0;
2606 netdev
->priv
.veth_attr
.ifindex
= 0;
2607 goto clear_ifindices
;
2610 /* Delete the openvswitch port. */
2611 ret
= lxc_ovs_delete_port(netdev
->link
, hostveth
);
2613 WARN("Failed to remove port \"%s\" from openvswitch "
2614 "bridge \"%s\"", hostveth
, netdev
->link
);
2616 INFO("Removed port \"%s\" from openvswitch bridge \"%s\"",
2617 hostveth
, netdev
->link
);
2620 /* We need to clear any ifindeces we recorded so liblxc won't
2621 * have cached stale data which would cause it to fail on reboot
2622 * we're we don't re-read the on-disk config file.
2624 netdev
->ifindex
= 0;
2625 if (netdev
->type
== LXC_NET_PHYS
) {
2626 netdev
->priv
.phys_attr
.ifindex
= 0;
2627 } else if (netdev
->type
== LXC_NET_VETH
) {
2628 netdev
->priv
.veth_attr
.veth1
[0] = '\0';
2629 netdev
->priv
.veth_attr
.ifindex
= 0;
2636 int lxc_requests_empty_network(struct lxc_handler
*handler
)
2638 struct lxc_list
*network
= &handler
->conf
->network
;
2639 struct lxc_list
*iterator
;
2640 bool found_none
= false, found_nic
= false;
2642 if (lxc_list_empty(network
))
2645 lxc_list_for_each(iterator
, network
) {
2646 struct lxc_netdev
*netdev
= iterator
->elem
;
2648 if (netdev
->type
== LXC_NET_NONE
)
2653 if (found_none
&& !found_nic
)
2658 /* try to move physical nics to the init netns */
2659 int lxc_restore_phys_nics_to_netns(struct lxc_handler
*handler
)
2663 char ifname
[IFNAMSIZ
];
2664 struct lxc_list
*iterator
;
2665 int netnsfd
= handler
->nsfd
[LXC_NS_NET
];
2666 struct lxc_conf
*conf
= handler
->conf
;
2668 /* We need CAP_NET_ADMIN in the parent namespace in order to setns() to
2669 * the parent network namespace. We won't have this capability if we are
2672 if (!handler
->am_root
)
2675 TRACE("Moving physical network devices back to parent network namespace");
2677 oldfd
= lxc_preserve_ns(lxc_raw_getpid(), "net");
2679 SYSERROR("Failed to preserve network namespace");
2683 ret
= setns(netnsfd
, CLONE_NEWNET
);
2685 SYSERROR("Failed to enter network namespace");
2690 lxc_list_for_each(iterator
, &conf
->network
) {
2691 struct lxc_netdev
*netdev
= iterator
->elem
;
2693 if (netdev
->type
!= LXC_NET_PHYS
)
2696 /* Retrieve the name of the interface in the container's network
2699 if (!if_indextoname(netdev
->ifindex
, ifname
)) {
2700 WARN("No interface corresponding to ifindex %d",
2705 ret
= lxc_netdev_move_by_name(ifname
, 1, netdev
->link
);
2707 WARN("Error moving network device \"%s\" back to "
2708 "network namespace", ifname
);
2710 TRACE("Moved network device \"%s\" back to network "
2711 "namespace", ifname
);
2714 ret
= setns(oldfd
, CLONE_NEWNET
);
2717 SYSERROR("Failed to enter network namespace");
2724 static int setup_hw_addr(char *hwaddr
, const char *ifname
)
2726 struct sockaddr sockaddr
;
2730 ret
= lxc_convert_mac(hwaddr
, &sockaddr
);
2733 SYSERROR("Mac address \"%s\" conversion failed", hwaddr
);
2737 memcpy(ifr
.ifr_name
, ifname
, IFNAMSIZ
);
2738 ifr
.ifr_name
[IFNAMSIZ
-1] = '\0';
2739 memcpy((char *) &ifr
.ifr_hwaddr
, (char *) &sockaddr
, sizeof(sockaddr
));
2741 fd
= socket(AF_INET
, SOCK_DGRAM
, 0);
2745 ret
= ioctl(fd
, SIOCSIFHWADDR
, &ifr
);
2747 SYSERROR("Failed to perform ioctl");
2751 DEBUG("Mac address \"%s\" on \"%s\" has been setup", hwaddr
,
2757 static int setup_ipv4_addr(struct lxc_list
*ip
, int ifindex
)
2759 struct lxc_list
*iterator
;
2762 lxc_list_for_each(iterator
, ip
) {
2763 struct lxc_inetdev
*inetdev
= iterator
->elem
;
2765 err
= lxc_ipv4_addr_add(ifindex
, &inetdev
->addr
,
2766 &inetdev
->bcast
, inetdev
->prefix
);
2769 SYSERROR("Failed to setup ipv4 address for network device "
2770 "with eifindex %d", ifindex
);
2778 static int setup_ipv6_addr(struct lxc_list
*ip
, int ifindex
)
2780 struct lxc_list
*iterator
;
2783 lxc_list_for_each(iterator
, ip
) {
2784 struct lxc_inet6dev
*inet6dev
= iterator
->elem
;
2786 err
= lxc_ipv6_addr_add(ifindex
, &inet6dev
->addr
,
2787 &inet6dev
->mcast
, &inet6dev
->acast
,
2791 SYSERROR("Failed to setup ipv6 address for network device "
2792 "with eifindex %d", ifindex
);
2800 static int lxc_setup_netdev_in_child_namespaces(struct lxc_netdev
*netdev
)
2802 char ifname
[IFNAMSIZ
];
2804 const char *net_type_name
;
2805 char *current_ifname
= ifname
;
2807 /* empty network namespace */
2808 if (!netdev
->ifindex
) {
2809 if (netdev
->flags
& IFF_UP
) {
2810 err
= lxc_netdev_up("lo");
2813 SYSERROR("Failed to set the loopback network device up");
2818 if (netdev
->type
== LXC_NET_EMPTY
)
2821 if (netdev
->type
== LXC_NET_NONE
)
2824 if (netdev
->type
!= LXC_NET_VETH
) {
2825 net_type_name
= lxc_net_type_to_str(netdev
->type
);
2826 ERROR("%s networks are not supported for containers "
2827 "not setup up by privileged users", net_type_name
);
2831 netdev
->ifindex
= if_nametoindex(netdev
->name
);
2834 /* get the new ifindex in case of physical netdev */
2835 if (netdev
->type
== LXC_NET_PHYS
) {
2836 netdev
->ifindex
= if_nametoindex(netdev
->link
);
2837 if (!netdev
->ifindex
) {
2838 ERROR("Failed to get ifindex for network device \"%s\"",
2844 /* retrieve the name of the interface */
2845 if (!if_indextoname(netdev
->ifindex
, current_ifname
)) {
2846 ERROR("Failed get name for network device with ifindex %d",
2851 /* Default: let the system to choose one interface name.
2852 * When the IFLA_IFNAME attribute is passed something like "<prefix>%d"
2853 * netlink will replace the format specifier with an appropriate index.
2855 if (netdev
->name
[0] == '\0') {
2856 if (netdev
->type
== LXC_NET_PHYS
)
2857 (void)strlcpy(netdev
->name
, netdev
->link
, IFNAMSIZ
);
2859 (void)strlcpy(netdev
->name
, "eth%d", IFNAMSIZ
);
2862 /* rename the interface name */
2863 if (strcmp(ifname
, netdev
->name
) != 0) {
2864 err
= lxc_netdev_rename_by_name(ifname
, netdev
->name
);
2867 SYSERROR("Failed to rename network device \"%s\" to \"%s\"",
2868 ifname
, netdev
->name
);
2873 /* Re-read the name of the interface because its name has changed
2874 * and would be automatically allocated by the system
2876 if (!if_indextoname(netdev
->ifindex
, current_ifname
)) {
2877 ERROR("Failed get name for network device with ifindex %d",
2882 /* Now update the recorded name of the network device to reflect the
2883 * name of the network device in the child's network namespace. We will
2884 * later on send this information back to the parent.
2886 (void)strlcpy(netdev
->name
, current_ifname
, IFNAMSIZ
);
2888 /* set a mac address */
2889 if (netdev
->hwaddr
) {
2890 if (setup_hw_addr(netdev
->hwaddr
, current_ifname
)) {
2891 ERROR("Failed to setup hw address for network device \"%s\"",
2897 /* setup ipv4 addresses on the interface */
2898 if (setup_ipv4_addr(&netdev
->ipv4
, netdev
->ifindex
)) {
2899 ERROR("Failed to setup ip addresses for network device \"%s\"",
2904 /* setup ipv6 addresses on the interface */
2905 if (setup_ipv6_addr(&netdev
->ipv6
, netdev
->ifindex
)) {
2906 ERROR("Failed to setup ipv6 addresses for network device \"%s\"",
2911 /* set the network device up */
2912 if (netdev
->flags
& IFF_UP
) {
2915 err
= lxc_netdev_up(current_ifname
);
2918 SYSERROR("Failed to set network device \"%s\" up",
2923 /* the network is up, make the loopback up too */
2924 err
= lxc_netdev_up("lo");
2927 SYSERROR("Failed to set the loopback network device up");
2932 /* We can only set up the default routes after bringing
2933 * up the interface, sine bringing up the interface adds
2934 * the link-local routes and we can't add a default
2935 * route if the gateway is not reachable. */
2937 /* setup ipv4 gateway on the interface */
2938 if (netdev
->ipv4_gateway
) {
2939 if (!(netdev
->flags
& IFF_UP
)) {
2940 ERROR("Cannot add ipv4 gateway for network device "
2941 "\"%s\" when not bringing up the interface", ifname
);
2945 if (lxc_list_empty(&netdev
->ipv4
)) {
2946 ERROR("Cannot add ipv4 gateway for network device "
2947 "\"%s\" when not assigning an address", ifname
);
2951 err
= lxc_ipv4_gateway_add(netdev
->ifindex
, netdev
->ipv4_gateway
);
2953 err
= lxc_ipv4_dest_add(netdev
->ifindex
, netdev
->ipv4_gateway
);
2956 SYSERROR("Failed to add ipv4 dest for network device \"%s\"",
2960 err
= lxc_ipv4_gateway_add(netdev
->ifindex
, netdev
->ipv4_gateway
);
2963 SYSERROR("Failed to setup ipv4 gateway for network device \"%s\"",
2966 if (netdev
->ipv4_gateway_auto
) {
2967 char buf
[INET_ADDRSTRLEN
];
2968 inet_ntop(AF_INET
, netdev
->ipv4_gateway
, buf
, sizeof(buf
));
2969 ERROR("Fried to set autodetected ipv4 gateway \"%s\"", buf
);
2976 /* setup ipv6 gateway on the interface */
2977 if (netdev
->ipv6_gateway
) {
2978 if (!(netdev
->flags
& IFF_UP
)) {
2979 ERROR("Cannot add ipv6 gateway for network device "
2980 "\"%s\" when not bringing up the interface", ifname
);
2984 if (lxc_list_empty(&netdev
->ipv6
) && !IN6_IS_ADDR_LINKLOCAL(netdev
->ipv6_gateway
)) {
2985 ERROR("Cannot add ipv6 gateway for network device "
2986 "\"%s\" when not assigning an address", ifname
);
2990 err
= lxc_ipv6_gateway_add(netdev
->ifindex
, netdev
->ipv6_gateway
);
2992 err
= lxc_ipv6_dest_add(netdev
->ifindex
, netdev
->ipv6_gateway
);
2995 SYSERROR("Failed to add ipv6 dest for network device \"%s\"",
2999 err
= lxc_ipv6_gateway_add(netdev
->ifindex
, netdev
->ipv6_gateway
);
3002 SYSERROR("Failed to setup ipv6 gateway for network device \"%s\"",
3005 if (netdev
->ipv6_gateway_auto
) {
3006 char buf
[INET6_ADDRSTRLEN
];
3007 inet_ntop(AF_INET6
, netdev
->ipv6_gateway
, buf
, sizeof(buf
));
3008 ERROR("Tried to set autodetected ipv6 "
3009 "gateway for network device "
3017 DEBUG("Network device \"%s\" has been setup", current_ifname
);
3022 int lxc_setup_network_in_child_namespaces(const struct lxc_conf
*conf
,
3023 struct lxc_list
*network
)
3025 struct lxc_list
*iterator
;
3026 struct lxc_netdev
*netdev
;
3028 lxc_list_for_each(iterator
, network
) {
3029 netdev
= iterator
->elem
;
3031 if (lxc_setup_netdev_in_child_namespaces(netdev
)) {
3032 ERROR("failed to setup netdev");
3037 if (!lxc_list_empty(network
))
3038 INFO("network has been setup");
3043 int lxc_network_send_veth_names_to_child(struct lxc_handler
*handler
)
3045 struct lxc_list
*iterator
;
3046 struct lxc_list
*network
= &handler
->conf
->network
;
3047 int data_sock
= handler
->data_sock
[0];
3049 if (handler
->am_root
)
3052 lxc_list_for_each(iterator
, network
) {
3054 struct lxc_netdev
*netdev
= iterator
->elem
;
3056 if (netdev
->type
!= LXC_NET_VETH
)
3059 ret
= send(data_sock
, netdev
->name
, IFNAMSIZ
, MSG_NOSIGNAL
);
3062 TRACE("Sent network device name \"%s\" to child", netdev
->name
);
3068 int lxc_network_recv_veth_names_from_parent(struct lxc_handler
*handler
)
3070 struct lxc_list
*iterator
;
3071 struct lxc_list
*network
= &handler
->conf
->network
;
3072 int data_sock
= handler
->data_sock
[1];
3074 if (handler
->am_root
)
3077 lxc_list_for_each(iterator
, network
) {
3079 struct lxc_netdev
*netdev
= iterator
->elem
;
3081 if (netdev
->type
!= LXC_NET_VETH
)
3084 ret
= recv(data_sock
, netdev
->name
, IFNAMSIZ
, 0);
3087 TRACE("Received network device name \"%s\" from parent", netdev
->name
);
3093 int lxc_network_send_name_and_ifindex_to_parent(struct lxc_handler
*handler
)
3095 struct lxc_list
*iterator
, *network
;
3096 int data_sock
= handler
->data_sock
[0];
3098 if (!handler
->am_root
)
3101 network
= &handler
->conf
->network
;
3102 lxc_list_for_each(iterator
, network
) {
3104 struct lxc_netdev
*netdev
= iterator
->elem
;
3106 /* Send network device name in the child's namespace to parent. */
3107 ret
= send(data_sock
, netdev
->name
, IFNAMSIZ
, MSG_NOSIGNAL
);
3111 /* Send network device ifindex in the child's namespace to
3114 ret
= send(data_sock
, &netdev
->ifindex
, sizeof(netdev
->ifindex
), MSG_NOSIGNAL
);
3119 TRACE("Sent network device names and ifindeces to parent");
3123 int lxc_network_recv_name_and_ifindex_from_child(struct lxc_handler
*handler
)
3125 struct lxc_list
*iterator
, *network
;
3126 int data_sock
= handler
->data_sock
[1];
3128 if (!handler
->am_root
)
3131 network
= &handler
->conf
->network
;
3132 lxc_list_for_each(iterator
, network
) {
3134 struct lxc_netdev
*netdev
= iterator
->elem
;
3136 /* Receive network device name in the child's namespace to
3139 ret
= recv(data_sock
, netdev
->name
, IFNAMSIZ
, 0);
3143 /* Receive network device ifindex in the child's namespace to
3146 ret
= recv(data_sock
, &netdev
->ifindex
, sizeof(netdev
->ifindex
), 0);
3154 void lxc_delete_network(struct lxc_handler
*handler
)
3158 if (handler
->am_root
)
3159 bret
= lxc_delete_network_priv(handler
);
3161 bret
= lxc_delete_network_unpriv(handler
);
3163 DEBUG("Failed to delete network devices");
3165 DEBUG("Deleted network devices");
3168 int addattr(struct nlmsghdr
*n
, size_t maxlen
, int type
, const void *data
, size_t alen
)
3170 int len
= RTA_LENGTH(alen
);
3173 if (NLMSG_ALIGN(n
->nlmsg_len
) + RTA_ALIGN(len
) > maxlen
)
3176 rta
= NLMSG_TAIL(n
);
3177 rta
->rta_type
= type
;
3180 memcpy(RTA_DATA(rta
), data
, alen
);
3181 n
->nlmsg_len
= NLMSG_ALIGN(n
->nlmsg_len
) + RTA_ALIGN(len
);
3186 /* Attributes of RTM_NEWNSID/RTM_GETNSID messages */
3189 #define __LXC_NETNSA_NSID_NOT_ASSIGNED -1
3196 int lxc_netns_set_nsid(int fd
)
3199 char buf
[NLMSG_ALIGN(sizeof(struct nlmsghdr
)) +
3200 NLMSG_ALIGN(sizeof(struct rtgenmsg
)) +
3202 struct nl_handler nlh
;
3203 struct nlmsghdr
*hdr
;
3204 struct rtgenmsg
*msg
;
3206 __u32 netns_fd
= fd
;
3208 ret
= netlink_open(&nlh
, NETLINK_ROUTE
);
3212 memset(buf
, 0, sizeof(buf
));
3213 hdr
= (struct nlmsghdr
*)buf
;
3214 msg
= (struct rtgenmsg
*)NLMSG_DATA(hdr
);
3216 hdr
->nlmsg_len
= NLMSG_LENGTH(sizeof(*msg
));
3217 hdr
->nlmsg_type
= RTM_NEWNSID
;
3218 hdr
->nlmsg_flags
= NLM_F_REQUEST
| NLM_F_ACK
;
3220 hdr
->nlmsg_seq
= RTM_NEWNSID
;
3221 msg
->rtgen_family
= AF_UNSPEC
;
3223 addattr(hdr
, 1024, __LXC_NETNSA_FD
, &netns_fd
, sizeof(netns_fd
));
3224 addattr(hdr
, 1024, __LXC_NETNSA_NSID
, &ns_id
, sizeof(ns_id
));
3226 ret
= __netlink_transaction(&nlh
, hdr
, hdr
);
3227 netlink_close(&nlh
);