Re: [PATCH net-next V2 2/2] tun: allow to attach ebpf socket filter

[Jason Wang <jasowang@redhat.com>]

On 2018年01月06日 00:21, Willem de Bruijn wrote:
> On Fri, Jan 5, 2018 at 4:54 AM, Jason Wang <jasowang@redhat.com> wrote:
>> This patch allows userspace to attach eBPF filter to tun. This will
>> allow to implement VM dataplane filtering in a more efficient way
>> compared to cBPF filter by allowing either qemu or libvirt to
>> attach eBPF filter to tun.
>>
>> Signed-off-by: Jason Wang <jasowang@redhat.com>
>> ---
>>   drivers/net/tun.c           | 39 +++++++++++++++++++++++++++++++++++----
>>   include/uapi/linux/if_tun.h |  1 +
>>   2 files changed, 36 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
>> index 0853829..9fc8b70 100644
>> --- a/drivers/net/tun.c
>> +++ b/drivers/net/tun.c
>> @@ -238,6 +238,12 @@ struct tun_struct {
>>          struct tun_pcpu_stats __percpu *pcpu_stats;
>>          struct bpf_prog __rcu *xdp_prog;
>>          struct tun_prog __rcu *steering_prog;
>> +       struct tun_prog __rcu *filter_prog;
>> +};
>> +
>> +struct veth {
>> +       __be16 h_vlan_proto;
>> +       __be16 h_vlan_TCI;
>>   };
>>
>>   static int tun_napi_receive(struct napi_struct *napi, int budget)
>> @@ -984,12 +990,25 @@ static void tun_automq_xmit(struct tun_struct *tun, struct sk_buff *skb)
>>   #endif
>>   }
>>
>> +static unsigned int run_ebpf_filter(struct tun_struct *tun,
>> +                                   struct sk_buff *skb,
>> +                                   int len)
>> +{
>> +       struct tun_prog *prog = rcu_dereference(tun->filter_prog);
>> +
>> +       if (prog)
>> +               len = bpf_prog_run_clear_cb(prog->prog, skb);
>> +
>> +       return len;
>> +}
>> +
>>   /* Net device start xmit */
>>   static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
>>   {
>>          struct tun_struct *tun = netdev_priv(dev);
>>          int txq = skb->queue_mapping;
>>          struct tun_file *tfile;
>> +       int len = skb->len;
>>
>>          rcu_read_lock();
>>          tfile = rcu_dereference(tun->tfiles[txq]);
>> @@ -1015,6 +1034,16 @@ static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
>>              sk_filter(tfile->socket.sk, skb))
>>                  goto drop;
>>
>> +       len = run_ebpf_filter(tun, skb, len);
>> +
>> +       /* Trim extra bytes since we may inster vlan proto & TCI
> inster -> insert

Will fix.

>
>> +        * in tun_put_user().
>> +        */
>> +       if (skb_vlan_tag_present(skb))
>> +               len -= skb_vlan_tag_present(skb) ? sizeof(struct veth) : 0;
> no need for testing skb_vlan_tag_present twice.

Right.

> more importantly, why trim these bytes unconditionally?
>
> only if the filter trims a packet to a length shorter than the the minimum
> could this cause problems. sk_filter_trim_cap with a lower bound avoids
> that: skb_vlan_tag_present(skb) ? sizeof(struct vlan_ethhdr) : 0;

The problem is, if the filter want to trim to packet to 50 bytes, we may 
get 54 bytes if vlan tag is existed. This seems wrong.

Thanks