Re: netfilter: Use netlink_ns_capable to verify the permisions of netlink messages

On Wed, Jul 07, 2021 at 04:05:33PM +0800, iLifetruth wrote:

Hi, we have found that the same fix pattern of CVE-2014-0181 may not
forward ported to some netlink-related places in the latest linux
kernel(v5.13)

=============
Here is the description of CVE-2014-0181:

The Netlink implementation in the Linux kernel through 3.14.1 does not
provide a mechanism for authorizing socket operations based on the opener
of a socket, which allows local users to bypass intended access
restrictions and modify network configurations by using a Netlink socket
for the (1) stdout or (2) stderr of a setuid program.

==========
And here is the solution to CVE-2014-0181:

To keep this from happening, replace bare capable and ns_capable calls with
netlink_capable, netlink_net_calls and netlink_ns_capable calls. Which act
the same as the previous calls *except they verify that the opener of the
socket had the desired permissions as well.*

==========
The upstream patch commit of this vulnerability described in CVE-2014-0181
is:
    90f62cf30a78721641e08737bda787552428061e (committed about 7 years ago)

=========
Capable() checks were added to these netlink-related places listed below
in netfilter by another upstream commit:
4b380c42f7d00a395feede754f0bc2292eebe6e5(committed about 4 years ago)

In kernel v5.13:
    File_1: linux/net/netfilter/nfnetlink_cthelper.c
                       in line 424, line 623 and line 691
    File_2: linux/net/netfilter/nfnetlink_osf.c
                       in line 305 and line 351

These subsystems depend on nfnetlink.

nfnetlink_rcv() is called before passing the message to the
corresponding backend, e.g. nfnetlink_osf.

static void nfnetlink_rcv(struct sk_buff *skb)
{
        struct nlmsghdr *nlh = nlmsg_hdr(skb);

        if (skb->len < NLMSG_HDRLEN ||
            nlh->nlmsg_len < NLMSG_HDRLEN ||
            skb->len < nlh->nlmsg_len)
                return;

        if (!netlink_net_capable(skb, CAP_NET_ADMIN)) {
                netlink_ack(skb, nlh, -EPERM, NULL);
                return;
        }
        [...]

which is calling netlink_net_capable().

But these checkers are still using bare capable instead of netlink_capable
calls. So this is likely to trigger the vulnerability described in the
CVE-2014-0181 without checking the desired permissions of the socket
opener. Now, shall we forward port the fix pattern from the patch of
CVE-2014-0181?

We would like to contact you to confirm this problem.

I think these capable() calls in nfnetlink_cthelper and nfnetlink_osf
are dead code that can be removed. As I explained these subsystems
stay behind nfnetlink.

I see.
There is no need to check the capability again in the
nfnetlink_cthelper and nfnetlink_osf now.

Regards and thanks for your analyze,
- iLifetruth




On Wed, Jul 7, 2021 at 5:18 PM Pablo Neira Ayuso [off-list ref] wrote:

On Wed, Jul 07, 2021 at 04:05:33PM +0800, iLifetruth wrote:

quoted

Hi, we have found that the same fix pattern of CVE-2014-0181 may not
forward ported to some netlink-related places in the latest linux
kernel(v5.13)

=============
Here is the description of CVE-2014-0181:

The Netlink implementation in the Linux kernel through 3.14.1 does not
provide a mechanism for authorizing socket operations based on the opener
of a socket, which allows local users to bypass intended access
restrictions and modify network configurations by using a Netlink socket
for the (1) stdout or (2) stderr of a setuid program.

==========
And here is the solution to CVE-2014-0181:

To keep this from happening, replace bare capable and ns_capable calls with
netlink_capable, netlink_net_calls and netlink_ns_capable calls. Which act
the same as the previous calls *except they verify that the opener of the
socket had the desired permissions as well.*

==========
The upstream patch commit of this vulnerability described in CVE-2014-0181
is:
    90f62cf30a78721641e08737bda787552428061e (committed about 7 years ago)

=========
Capable() checks were added to these netlink-related places listed below
in netfilter by another upstream commit:
4b380c42f7d00a395feede754f0bc2292eebe6e5(committed about 4 years ago)

In kernel v5.13:
    File_1: linux/net/netfilter/nfnetlink_cthelper.c
                       in line 424, line 623 and line 691
    File_2: linux/net/netfilter/nfnetlink_osf.c
                       in line 305 and line 351

These subsystems depend on nfnetlink.

nfnetlink_rcv() is called before passing the message to the
corresponding backend, e.g. nfnetlink_osf.

static void nfnetlink_rcv(struct sk_buff *skb)
{
        struct nlmsghdr *nlh = nlmsg_hdr(skb);

        if (skb->len < NLMSG_HDRLEN ||
            nlh->nlmsg_len < NLMSG_HDRLEN ||
            skb->len < nlh->nlmsg_len)
                return;

        if (!netlink_net_capable(skb, CAP_NET_ADMIN)) {
                netlink_ack(skb, nlh, -EPERM, NULL);
                return;
        }
        [...]

which is calling netlink_net_capable().

quoted

But these checkers are still using bare capable instead of netlink_capable
calls. So this is likely to trigger the vulnerability described in the
CVE-2014-0181 without checking the desired permissions of the socket
opener. Now, shall we forward port the fix pattern from the patch of
CVE-2014-0181?

We would like to contact you to confirm this problem.

I think these capable() calls in nfnetlink_cthelper and nfnetlink_osf
are dead code that can be removed. As I explained these subsystems
stay behind nfnetlink.