net: validate veth and vxcan peer ifindexes
Subsystems (3)
Jakub Kicinski
maintainerPATCH 1/1 net: validate veth and vxcan peer ifindexes
veth and vxcan need to make sure the ifindexes of the peer are not negative, core does not validate this.
Using iproute2 with user-space-level checking removed:
Before:
# ./ip link add index 10 type veth peer index -1
# ip link show
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: enp1s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP mode DEFAULT group default qlen 1000
link/ether 52:54:00:74:b2:03 brd ff:ff:ff:ff:ff:ff
10: veth1@veth0: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 8a:90:ff:57:6d:5d brd ff:ff:ff:ff:ff:ff
-1: veth0@veth1: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether ae:ed:18:e6:fa:7f brd ff:ff:ff:ff:ff:ffNow:
$ ./ip link add index 10 type veth peer index -1 Error: ifindex can't be negative.
This problem surfaced in net-next because an explicit WARN() was added, the root cause is older.
---
CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]
CC: [email protected]
--- drivers/net/can/vxcan.c | 7 +------ drivers/net/veth.c | 5 +---- include/net/rtnetlink.h | 4 ++-- net/core/rtnetlink.c | 22 ++++++++++++++++++---- 4 files changed, 22 insertions(+), 16 deletions(-)
drivers/net/can/vxcan.c +1 -6
@@ -192,12 +192,7 @@ static int vxcan_newlink(struct net *net, struct net_device *dev, nla_peer = data[VXCAN_INFO_PEER ]; ifmp =nla_data (nla_peer); - err =rtnl_nla_parse_ifla (peer_tb, -nla_data (nla_peer) + -sizeof (struct ifinfomsg ), -nla_len (nla_peer) - -sizeof (struct ifinfomsg ), -NULL ); + err =rtnl_nla_parse_ifinfomsg (peer_tb, nla_peer, extack);if (err <0 )return err;
drivers/net/veth.c +1 -4
@@ -1861,10 +1861,7 @@ static int veth_newlink(struct net *src_net, struct net_device *dev, nla_peer = data[VETH_INFO_PEER ]; ifmp =nla_data (nla_peer); - err =rtnl_nla_parse_ifla (peer_tb, -nla_data (nla_peer) +sizeof (struct ifinfomsg ), -nla_len (nla_peer) -sizeof (struct ifinfomsg ), -NULL ); + err =rtnl_nla_parse_ifinfomsg (peer_tb, nla_peer, extack);if (err <0 )return err;
include/net/rtnetlink.h +2 -2
@@ -190,8 +190,8 @@ int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *int rtnl_configure_link (struct net_device *dev,const struct ifinfomsg *ifm,u32 portid,const struct nlmsghdr *nlh); -int rtnl_nla_parse_ifla (struct nlattr **tb,const struct nlattr *head,int len, -struct netlink_ext_ack *exterr); +int rtnl_nla_parse_ifinfomsg (struct nlattr **tb,const struct nlattr *nla_peer, +struct netlink_ext_ack *exterr);struct net *rtnl_get_net_ns_capable (struct sock *sk,int netnsid);#define MODULE_ALIAS_RTNL_LINK (kind) MODULE_ALIAS("rtnl-link-" kind)
net/core/rtnetlink.c +18 -4
@@ -2268,13 +2268,27 @@ static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)return err; } -int rtnl_nla_parse_ifla (struct nlattr **tb,const struct nlattr *head,int len, -struct netlink_ext_ack *exterr) +int rtnl_nla_parse_ifinfomsg (struct nlattr **tb,const struct nlattr *nla_peer, +struct netlink_ext_ack *exterr) { -return nla_parse_deprecated (tb,IFLA_MAX , head, len, ifla_policy, +const struct ifinfomsg *ifmp; +const struct nlattr *attrs; +size_t len; + + ifmp =nla_data (nla_peer); + attrs =nla_data (nla_peer) +sizeof (struct ifinfomsg ); + len =nla_len (nla_peer) -sizeof (struct ifinfomsg ); + +if (ifmp->ifi_index <0 ) { +NL_SET_ERR_MSG_ATTR (exterr, nla_peer, +"ifindex can't be negative" ); +return -EINVAL ; + } + +return nla_parse_deprecated (tb,IFLA_MAX , attrs, len, ifla_policy, exterr); } -EXPORT_SYMBOL (rtnl_nla_parse_ifla); +EXPORT_SYMBOL (rtnl_nla_parse_ifinfomsg);struct net *rtnl_link_get_net (struct net *src_net,struct nlattr *tb[]) {
Signature · 1 line
2.41.0
Quoted 24 lines · Jakub Kicinski
veth and vxcan need to make sure the ifindexes of the peer are not negative, core does not validate this.
Using iproute2 with user-space-level checking removed:
Before:
# ./ip link add index 10 type veth peer index -1 # ip link show 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000 link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 2: enp1s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP mode DEFAULT group default qlen 1000 link/ether 52:54:00:74:b2:03 brd ff:ff:ff:ff:ff:ff 10: veth1@veth0: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000 link/ether 8a:90:ff:57:6d:5d brd ff:ff:ff:ff:ff:ff -1: veth0@veth1: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000 link/ether ae:ed:18:e6:fa:7f brd ff:ff:ff:ff:ff:ffNow:
$ ./ip link add index 10 type veth peer index -1 Error: ifindex can't be negative.
Quoted 7 lines
This problem surfaced in net-next because an explicit
WARN()was added, the root cause is older.Fixes e6f8f1a739b6 ("veth: Allow to create peer link with given ifindex")Fixes a8f820a380a2 ("can: add Virtual CAN Tunnel driver (vxcan)")Reported-by [email protected]Signed-off-by Jakub Kicinski <[email protected]>
SGTM, I was not sure how to fix this myself ;)
This patch was applied to netdev/net.git (main) by David S. Miller <[email protected]>:
Hello:
This patch was applied to netdev/net.git (main)
by David S. Miller <[email protected]>:
Quoted 8 lines · On Fri, 18 Aug 2023 18:26:02 -0700 you wrote:
veth and vxcan need to make sure the ifindexes of the peer are not negative, core does not validate this.
Using iproute2 with user-space-level checking removed:
Before:
[...]
Here is the summary with links:
- [net] net: validate veth and vxcan peer ifindexes
https://git.kernel.org/netdev/net/c/f534f6581ec0
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
Quoted 23 lines · Jakub Kicinski
veth and vxcan need to make sure the ifindexes of the peer are not negative, core does not validate this.
Using iproute2 with user-space-level checking removed:
Before:
# ./ip link add index 10 type veth peer index -1 # ip link show 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000 link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 2: enp1s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP mode DEFAULT group default qlen 1000 link/ether 52:54:00:74:b2:03 brd ff:ff:ff:ff:ff:ff 10: veth1@veth0: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000 link/ether 8a:90:ff:57:6d:5d brd ff:ff:ff:ff:ff:ff -1: veth0@veth1: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000 link/ether ae:ed:18:e6:fa:7f brd ff:ff:ff:ff:ff:ffNow:
$ ./ip link add index 10 type veth peer index -1 Error: ifindex can't be negative.
Quoted 7 lines
This problem surfaced in net-next because an explicit
WARN()was added, the root cause is older.Fixes e6f8f1a739b6 ("veth: Allow to create peer link with given ifindex")Fixes a8f820a380a2 ("can: add Virtual CAN Tunnel driver (vxcan)")Reported-by [email protected]Signed-off-by Jakub Kicinski <[email protected]>
There is another report here [1] with a reproducer [2]. Even with this patch, the reproducer can still trigger the warning on net-next. Don't we also need to reject a negative ifindex in the ancillary header? At least with the following diff the warning does not trigger anymore:
net/core/rtnetlink.c +3 -0
@@ -3560,6 +3560,9 @@ static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,if (ifm->ifi_index >0 ) { link_specified = true; dev =__dev_get_by_index (net, ifm->ifi_index); + }else if (ifm->ifi_index <0 ) { +NL_SET_ERR_MSG (extack,"ifindex can't be negative" ); +return -EINVAL ; }else if (tb[IFLA_IFNAME ] || tb[IFLA_ALT_IFNAME ]) { link_specified = true; dev =rtnl_dev_get (net, tb);
[1] https://syzkaller.appspot.com/text?tag=CrashReport&x=178edad3a80000
[2] https://syzkaller.appspot.com/text?tag=ReproC&x=166ed6bba80000
Jakub Kicinski
maintainerQuoted 4 lines · On Sun, 20 Aug 2023 19:08:13 +0300 Ido Schimmel wrote:
There is another report here [1] with a reproducer [2]. Even with this patch, the reproducer can still trigger the warning on net-next. Don't we also need to reject a negative ifindex in the ancillary header? At least with the following diff the warning does not trigger anymore:
Yeah, definitely, please go ahead and submit.
Is "ancillary header" used more commonly as a term? in gnel we usually call this thing "user header" or "fixed header".
Quoted 7 lines · Jakub Kicinski
Quoted 4 lines · On Sun, 20 Aug 2023 19:08:13 +0300 Ido Schimmel wrote:
There is another report here [1] with a reproducer [2]. Even with this patch, the reproducer can still trigger the warning on net-next. Don't we also need to reject a negative ifindex in the ancillary header? At least with the following diff the warning does not trigger anymore:
Yeah, definitely, please go ahead and submit.
Sure, will submit tomorrow morning.
Quoted 2 lines
Is "ancillary header" used more commonly as a term? in gnel we usually call this thing "user header" or "fixed header".
I honestly don't know. IIRC I saw David using the term a few years ago and decided to adopt it.