Re: [PATCH] netfilter: check if the socket netns is correct.

Re: [PATCH] netfilter: check if the socket netns is correct.

[Guenter Roeck]

Hi Flavio,

On Wed, Jun 27, 2018 at 10:34:25AM -0300, Flavio Leitner wrote:
> Netfilter assumes that if the socket is present in the skb, then
> it can be used because that reference is cleaned up while the skb
> is crossing netns.
> 
> We want to change that to preserve the socket reference in a future
> patch, so this is a preparation updating netfilter to check if the
> socket netns matches before use it.
> 
> Signed-off-by: Flavio Leitner <fbl@redhat.com>
> Acked-by: Florian Westphal <fw@strlen.de>
> Signed-off-by: David S. Miller <davem@davemloft.net>
> ---
...
> --- a/net/netfilter/xt_socket.c
> +++ b/net/netfilter/xt_socket.c
> @@ -56,8 +56,12 @@ socket_match(const struct sk_buff *skb, struct xt_action_param *par,
>  	struct sk_buff *pskb = (struct sk_buff *)skb;
>  	struct sock *sk = skb->sk;
>  
> +	if (!net_eq(xt_net(par), sock_net(sk)))
> +		sk = NULL;
> +

I am having trouble with this code. With CONFIG_NET_NS enabled, it crashes
for me in read_pnet() because sk is NULL.

>  	if (!sk)
>  		sk = nf_sk_lookup_slow_v4(xt_net(par), skb, xt_in(par));

The old code seems to suggest that sk == NULL was possible.

I see the problem with the Chrome OS kernel rebased to v4.19-rc5, so I
can not guarantee that this really an upstream problem. The change seems
odd, though. Are you sure that it is not (or, rather, no longer) necessary
to check if sk == NULL before dereferencing it in sock_net() ?

> +
>  	if (sk) {
>  		bool wildcard;
>  		bool transparent = true;
> @@ -113,8 +117,12 @@ socket_mt6_v1_v2_v3(const struct sk_buff *skb, struct xt_action_param *par)
>  	struct sk_buff *pskb = (struct sk_buff *)skb;
>  	struct sock *sk = skb->sk;
>  
> +	if (!net_eq(xt_net(par), sock_net(sk)))
> +		sk = NULL;
> +
Same here.

>  	if (!sk)
>  		sk = nf_sk_lookup_slow_v6(xt_net(par), skb, xt_in(par));
> +
>  	if (sk) {
>  		bool wildcard;
>  		bool transparent = true;

Thanks,
Guenter

Re: [PATCH] netfilter: check if the socket netns is correct.

[Flavio Leitner]

On Thu, Sep 27, 2018 at 01:46:29PM -0700, Guenter Roeck wrote:
> Hi Flavio,
> 
> On Wed, Jun 27, 2018 at 10:34:25AM -0300, Flavio Leitner wrote:
> > Netfilter assumes that if the socket is present in the skb, then
> > it can be used because that reference is cleaned up while the skb
> > is crossing netns.
> > 
> > We want to change that to preserve the socket reference in a future
> > patch, so this is a preparation updating netfilter to check if the
> > socket netns matches before use it.
> > 
> > Signed-off-by: Flavio Leitner <fbl@redhat.com>
> > Acked-by: Florian Westphal <fw@strlen.de>
> > Signed-off-by: David S. Miller <davem@davemloft.net>
> > ---
> ...
> > --- a/net/netfilter/xt_socket.c
> > +++ b/net/netfilter/xt_socket.c
> > @@ -56,8 +56,12 @@ socket_match(const struct sk_buff *skb, struct xt_action_param *par,
> >  	struct sk_buff *pskb = (struct sk_buff *)skb;
> >  	struct sock *sk = skb->sk;
> >  
> > +	if (!net_eq(xt_net(par), sock_net(sk)))
> > +		sk = NULL;
> > +
> 
> I am having trouble with this code. With CONFIG_NET_NS enabled, it crashes
> for me in read_pnet() because sk is NULL.
> 
> >  	if (!sk)
> >  		sk = nf_sk_lookup_slow_v4(xt_net(par), skb, xt_in(par));
> 
> The old code seems to suggest that sk == NULL was possible.
> 
> I see the problem with the Chrome OS kernel rebased to v4.19-rc5, so I
> can not guarantee that this really an upstream problem. The change seems
> odd, though. Are you sure that it is not (or, rather, no longer) necessary
> to check if sk == NULL before dereferencing it in sock_net() ?

Oops, it is necessary but if it's not and the netns doesn't match, we need
do the lookup. So, could you check if this fixes the problem for you?

>From a5f927e7f1368d753f87cb978d630d786d5adb62 Mon Sep 17 00:00:00 2001
From: Flavio Leitner <fbl@redhat.com>
Date: Thu, 27 Sep 2018 19:36:28 -0300
Subject: [PATCH] xt_socket: check sk before checking for netns.

Only check for the network namespace if the socket is available.

Fixes: f564650106a6 ("netfilter: check if the socket netns is correct.")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Flavio Leitner <fbl@redhat.com>
---
 net/netfilter/xt_socket.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/netfilter/xt_socket.c b/net/netfilter/xt_socket.c
index 0472f3472842..ada144e5645b 100644
--- a/net/netfilter/xt_socket.c
+++ b/net/netfilter/xt_socket.c
@@ -56,7 +56,7 @@ socket_match(const struct sk_buff *skb, struct xt_action_param *par,
 	struct sk_buff *pskb = (struct sk_buff *)skb;
 	struct sock *sk = skb->sk;
 
-	if (!net_eq(xt_net(par), sock_net(sk)))
+	if (sk && !net_eq(xt_net(par), sock_net(sk)))
 		sk = NULL;
 
 	if (!sk)
@@ -117,7 +117,7 @@ socket_mt6_v1_v2_v3(const struct sk_buff *skb, struct xt_action_param *par)
 	struct sk_buff *pskb = (struct sk_buff *)skb;
 	struct sock *sk = skb->sk;
 
-	if (!net_eq(xt_net(par), sock_net(sk)))
+	if (sk && !net_eq(xt_net(par), sock_net(sk)))
 		sk = NULL;
 
 	if (!sk)
-- 
2.14.4

Re: [PATCH] netfilter: check if the socket netns is correct.

[Guenter Roeck]

On Thu, Sep 27, 2018 at 07:58:24PM -0300, Flavio Leitner wrote:
> On Thu, Sep 27, 2018 at 01:46:29PM -0700, Guenter Roeck wrote:
> > Hi Flavio,
> > 
> > On Wed, Jun 27, 2018 at 10:34:25AM -0300, Flavio Leitner wrote:
> > > Netfilter assumes that if the socket is present in the skb, then
> > > it can be used because that reference is cleaned up while the skb
> > > is crossing netns.
> > > 
> > > We want to change that to preserve the socket reference in a future
> > > patch, so this is a preparation updating netfilter to check if the
> > > socket netns matches before use it.
> > > 
> > > Signed-off-by: Flavio Leitner <fbl@redhat.com>
> > > Acked-by: Florian Westphal <fw@strlen.de>
> > > Signed-off-by: David S. Miller <davem@davemloft.net>
> > > ---
> > ...
> > > --- a/net/netfilter/xt_socket.c
> > > +++ b/net/netfilter/xt_socket.c
> > > @@ -56,8 +56,12 @@ socket_match(const struct sk_buff *skb, struct xt_action_param *par,
> > >  	struct sk_buff *pskb = (struct sk_buff *)skb;
> > >  	struct sock *sk = skb->sk;
> > >  
> > > +	if (!net_eq(xt_net(par), sock_net(sk)))
> > > +		sk = NULL;
> > > +
> > 
> > I am having trouble with this code. With CONFIG_NET_NS enabled, it crashes
> > for me in read_pnet() because sk is NULL.
> > 
> > >  	if (!sk)
> > >  		sk = nf_sk_lookup_slow_v4(xt_net(par), skb, xt_in(par));
> > 
> > The old code seems to suggest that sk == NULL was possible.
> > 
> > I see the problem with the Chrome OS kernel rebased to v4.19-rc5, so I
> > can not guarantee that this really an upstream problem. The change seems
> > odd, though. Are you sure that it is not (or, rather, no longer) necessary
> > to check if sk == NULL before dereferencing it in sock_net() ?
> 
> Oops, it is necessary but if it's not and the netns doesn't match, we need
> do the lookup. So, could you check if this fixes the problem for you?
> 
> From a5f927e7f1368d753f87cb978d630d786d5adb62 Mon Sep 17 00:00:00 2001
> From: Flavio Leitner <fbl@redhat.com>
> Date: Thu, 27 Sep 2018 19:36:28 -0300
> Subject: [PATCH] xt_socket: check sk before checking for netns.
> 
> Only check for the network namespace if the socket is available.
> 
> Fixes: f564650106a6 ("netfilter: check if the socket netns is correct.")
> Reported-by: Guenter Roeck <linux@roeck-us.net>
> Signed-off-by: Flavio Leitner <fbl@redhat.com>

This fixes the problem for me.

Tested-by: Guenter Roeck <linux@roeck-us.net>

Thanks,
Guenter

> ---
>  net/netfilter/xt_socket.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/net/netfilter/xt_socket.c b/net/netfilter/xt_socket.c
> index 0472f3472842..ada144e5645b 100644
> --- a/net/netfilter/xt_socket.c
> +++ b/net/netfilter/xt_socket.c
> @@ -56,7 +56,7 @@ socket_match(const struct sk_buff *skb, struct xt_action_param *par,
>  	struct sk_buff *pskb = (struct sk_buff *)skb;
>  	struct sock *sk = skb->sk;
>  
> -	if (!net_eq(xt_net(par), sock_net(sk)))
> +	if (sk && !net_eq(xt_net(par), sock_net(sk)))
>  		sk = NULL;
>  
>  	if (!sk)
> @@ -117,7 +117,7 @@ socket_mt6_v1_v2_v3(const struct sk_buff *skb, struct xt_action_param *par)
>  	struct sk_buff *pskb = (struct sk_buff *)skb;
>  	struct sock *sk = skb->sk;
>  
> -	if (!net_eq(xt_net(par), sock_net(sk)))
> +	if (sk && !net_eq(xt_net(par), sock_net(sk)))
>  		sk = NULL;
>  
>  	if (!sk)
> -- 
> 2.14.4
> 

Re: [PATCH] netfilter: check if the socket netns is correct.

[Pablo Neira Ayuso]

On Thu, Sep 27, 2018 at 07:58:24PM -0300, Flavio Leitner wrote:
[...]
> From a5f927e7f1368d753f87cb978d630d786d5adb62 Mon Sep 17 00:00:00 2001
> From: Flavio Leitner <fbl@redhat.com>
> Date: Thu, 27 Sep 2018 19:36:28 -0300
> Subject: [PATCH] xt_socket: check sk before checking for netns.
> 
> Only check for the network namespace if the socket is available.

Applied.

Please, Cc netfilter-devel@vger.kernel.org next time for your
netfilter patches.

Thanks.