net/openvswitch: check Ethernet header length in key_extract()

[PATCH net v3] net/openvswitch: check Ethernet header length in key_extract()

Posted by Cen Zhang (Microsoft) 21 hours ago

When a packet arrives on an ARPHRD_NONE device (e.g. TUN),
ovs_flow_key_extract() trusts the user-provided skb->protocol field: if
it is ETH_P_TEB, the packet is classified as MAC_PROTO_ETHERNET and
key_extract() is called without ensuring the skb has ETH_HLEN (14) bytes
of linear data. key_extract() unconditionally pulls 2 * ETH_ALEN bytes
for MAC addresses and parse_ethertype() pulls 2 more, either of which
triggers a kernel BUG in __skb_pull() when the linear area is too small.

  kernel BUG at include/linux/skbuff.h:2848!
  RIP: 0010:key_extract+0xa7e/0xd90 net/openvswitch/flow.c:933
  ovs_flow_key_extract+0x419/0xa70
  ovs_vport_receive+0x222/0x390
  netdev_frame_hook+0x3e0/0x630
  tun_get_user+0x2d0c/0x38e0

Fixed by adding checks in key_extract() before pulling the Ethernet
header.

Fixes: 217ac77a3c25 ("openvswitch: allow L3 netdev ports")
Reported-by: AutonomousCodeSecurity@microsoft.com
Signed-off-by: Cen Zhang (Microsoft) <blbllhy@gmail.com>
---
v3: Use check_header() per review, fix format issue.
v2: Moved the check into key_extract() per Ilya Maximets.
Link: https://lore.kernel.org/all/20260721143602.64677-1-blbllhy@gmail.com

 net/openvswitch/flow.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/net/openvswitch/flow.c b/net/openvswitch/flow.c
index 66366982f604..fc4002353e17 100644
--- a/net/openvswitch/flow.c
+++ b/net/openvswitch/flow.c
@@ -911,6 +911,7 @@ static int key_extract_l3l4(struct sk_buff *skb, struct sw_flow_key *key)
 static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
 {
 	struct ethhdr *eth;
+	int err;
 
 	/* Flags are always used as part of stats */
 	key->tp.flags = 0;
@@ -926,6 +927,10 @@ static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
 		skb_reset_network_header(skb);
 		key->eth.type = skb->protocol;
 	} else {
+		err = check_header(skb, ETH_HLEN);
+		if (unlikely(err))
+			return err;
+
 		eth = eth_hdr(skb);
 		ether_addr_copy(key->eth.src, eth->h_source);
 		ether_addr_copy(key->eth.dst, eth->h_dest);
-- 
2.53.0

Re: [PATCH net v3] net/openvswitch: check Ethernet header length in key_extract()

Posted by Eelco Chaudron 11 hours ago


On 24 Jul 2026, at 1:37, Cen Zhang (Microsoft) wrote:

> When a packet arrives on an ARPHRD_NONE device (e.g. TUN),
> ovs_flow_key_extract() trusts the user-provided skb->protocol field: if
> it is ETH_P_TEB, the packet is classified as MAC_PROTO_ETHERNET and
> key_extract() is called without ensuring the skb has ETH_HLEN (14) bytes
> of linear data. key_extract() unconditionally pulls 2 * ETH_ALEN bytes
> for MAC addresses and parse_ethertype() pulls 2 more, either of which
> triggers a kernel BUG in __skb_pull() when the linear area is too small.
>
>   kernel BUG at include/linux/skbuff.h:2848!
>   RIP: 0010:key_extract+0xa7e/0xd90 net/openvswitch/flow.c:933
>   ovs_flow_key_extract+0x419/0xa70
>   ovs_vport_receive+0x222/0x390
>   netdev_frame_hook+0x3e0/0x630
>   tun_get_user+0x2d0c/0x38e0
>
> Fixed by adding checks in key_extract() before pulling the Ethernet
> header.
>
> Fixes: 217ac77a3c25 ("openvswitch: allow L3 netdev ports")
> Reported-by: AutonomousCodeSecurity@microsoft.com
> Signed-off-by: Cen Zhang (Microsoft) <blbllhy@gmail.com>

Hi Cen,

Thanks for the new revision.

The doc comment for key_extract() also needs updating.  It currently
states:

  * The caller must ensure that skb->len >= ETH_HLEN.

The comment should be removed or replaced with a note that the
function returns -EINVAL when the frame is too short to contain an
Ethernet header.

One more comment below.

Cheers,

Eelco

> ---
> v3: Use check_header() per review, fix format issue.
> v2: Moved the check into key_extract() per Ilya Maximets.
> Link: https://lore.kernel.org/all/20260721143602.64677-1-blbllhy@gmail.com
>
>  net/openvswitch/flow.c | 5 +++++
>  1 file changed, 5 insertions(+)
>
> diff --git a/net/openvswitch/flow.c b/net/openvswitch/flow.c
> index 66366982f604..fc4002353e17 100644
> --- a/net/openvswitch/flow.c
> +++ b/net/openvswitch/flow.c
> @@ -911,6 +911,7 @@ static int key_extract_l3l4(struct sk_buff *skb, struct sw_flow_key *key)
>  static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
>  {
>  	struct ethhdr *eth;
> +	int err;

Since err is only used in the else block, it should be scoped there:

	} else {
		int err = check_header(skb, ETH_HLEN);

		if (unlikely(err))
			return err;

>
>  	/* Flags are always used as part of stats */
>  	key->tp.flags = 0;
> @@ -926,6 +927,10 @@ static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
>  		skb_reset_network_header(skb);
>  		key->eth.type = skb->protocol;
>  	} else {
> +		err = check_header(skb, ETH_HLEN);
> +		if (unlikely(err))
> +			return err;
> +
>  		eth = eth_hdr(skb);
>  		ether_addr_copy(key->eth.src, eth->h_source);
>  		ether_addr_copy(key->eth.dst, eth->h_dest);
> -- 
> 2.53.0

Re: [PATCH net v3] net/openvswitch: check Ethernet header length in key_extract()

Posted by Ilya Maximets 8 hours ago

On 7/24/26 12:12 PM, Eelco Chaudron wrote:
> 
> 
> On 24 Jul 2026, at 1:37, Cen Zhang (Microsoft) wrote:
> 
>> When a packet arrives on an ARPHRD_NONE device (e.g. TUN),
>> ovs_flow_key_extract() trusts the user-provided skb->protocol field: if
>> it is ETH_P_TEB, the packet is classified as MAC_PROTO_ETHERNET and
>> key_extract() is called without ensuring the skb has ETH_HLEN (14) bytes
>> of linear data. key_extract() unconditionally pulls 2 * ETH_ALEN bytes
>> for MAC addresses and parse_ethertype() pulls 2 more, either of which
>> triggers a kernel BUG in __skb_pull() when the linear area is too small.
>>
>>   kernel BUG at include/linux/skbuff.h:2848!
>>   RIP: 0010:key_extract+0xa7e/0xd90 net/openvswitch/flow.c:933
>>   ovs_flow_key_extract+0x419/0xa70
>>   ovs_vport_receive+0x222/0x390
>>   netdev_frame_hook+0x3e0/0x630
>>   tun_get_user+0x2d0c/0x38e0
>>
>> Fixed by adding checks in key_extract() before pulling the Ethernet
>> header.
>>
>> Fixes: 217ac77a3c25 ("openvswitch: allow L3 netdev ports")
>> Reported-by: AutonomousCodeSecurity@microsoft.com
>> Signed-off-by: Cen Zhang (Microsoft) <blbllhy@gmail.com>
> 
> Hi Cen,
> 
> Thanks for the new revision.
> 
> The doc comment for key_extract() also needs updating.  It currently
> states:
> 
>   * The caller must ensure that skb->len >= ETH_HLEN.
> 
> The comment should be removed or replaced with a note that the
> function returns -EINVAL when the frame is too short to contain an
> Ethernet header.
> 
> One more comment below.
> 
> Cheers,
> 
> Eelco
> 
>> ---
>> v3: Use check_header() per review, fix format issue.
>> v2: Moved the check into key_extract() per Ilya Maximets.
>> Link: https://lore.kernel.org/all/20260721143602.64677-1-blbllhy@gmail.com
>>
>>  net/openvswitch/flow.c | 5 +++++
>>  1 file changed, 5 insertions(+)
>>
>> diff --git a/net/openvswitch/flow.c b/net/openvswitch/flow.c
>> index 66366982f604..fc4002353e17 100644
>> --- a/net/openvswitch/flow.c
>> +++ b/net/openvswitch/flow.c
>> @@ -911,6 +911,7 @@ static int key_extract_l3l4(struct sk_buff *skb, struct sw_flow_key *key)
>>  static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
>>  {
>>  	struct ethhdr *eth;
>> +	int err;
> 
> Since err is only used in the else block, it should be scoped there:
> 
> 	} else {
> 		int err = check_header(skb, ETH_HLEN);
> 
> 		if (unlikely(err))
> 			return err;

The 'eth' is also only used in the 'else' block, but defined at the top.
It doesn't matter where, but I think, we should keep them together.

> 
>>
>>  	/* Flags are always used as part of stats */
>>  	key->tp.flags = 0;
>> @@ -926,6 +927,10 @@ static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
>>  		skb_reset_network_header(skb);
>>  		key->eth.type = skb->protocol;
>>  	} else {
>> +		err = check_header(skb, ETH_HLEN);
>> +		if (unlikely(err))
>> +			return err;
>> +
>>  		eth = eth_hdr(skb);
>>  		ether_addr_copy(key->eth.src, eth->h_source);
>>  		ether_addr_copy(key->eth.dst, eth->h_dest);
>> -- 
>> 2.53.0
>