Skip to contentSkip to conversation

net: validate veth and vxcan peer ifindexes

PATCH Applied Jakub Kicinski 4 participants 6 messages Aug 19, 2023 – Aug 22, 2023

Subsystems (3)
lore
Share & tools
lore

Jakub Kicinski

author maintainer
#0 lore raw

PATCH 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:ff

Now:

  $ ./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.

Fixes e6f8f1a739b6 ("veth: Allow to create peer link with given ifindex")
Fixes a8f820a380a2 ("can: add Virtual CAN Tunnel driver (vxcan)")
Reported-by [email protected]

---
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
index 4068d962203d..98c669ad5141 100644
@@ -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
index 509e901da41d..ef8eacb596f7 100644
@@ -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
index d9076a7a430c..6506221c5fe3 100644
@@ -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
index aef25aa5cf1d..bcebdeb59163 100644
@@ -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
Reviewed-by on v1 1/1
#1 lore raw
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:ff

Now:

  $ ./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]

SGTM, I was not sure how to fix this myself ;)

#2 lore raw
In reply to Jakub Kicinski
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
#3 lore raw
In reply to Jakub Kicinski
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:ff

Now:

  $ ./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]

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
index 7aba4d63b069..4a2ec33bfb51 100644
@@ -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

author maintainer
#4 lore raw
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.

Is "ancillary header" used more commonly as a term? in gnel we usually call this thing "user header" or "fixed header".

#5 lore raw
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.