netdev - Re: "wrong" ifindex on received VLAN tagged packet?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAOfa6T1QU+ukwTvujxN8xdSGjnEz=dXzMmGp1TQQGxaZU7ESaQ@mail.gmail.com>
Date:   Tue, 6 Mar 2018 17:27:43 -0800
From:   Lawrence Kreeger <lkreeger@...il.com>
To:     David Ahern <dsahern@...il.com>
Cc:     netdev@...r.kernel.org
Subject: Re: "wrong" ifindex on received VLAN tagged packet?

Using ETH_P_ALL instead of ETH_P_802_2, is causing mstpd to get 3
copies of the same BPDU.  One from eth0, one from eth0.100, and
another from vlan100 (the bridge).
mstpd will drop the one from vlan100, but since there is also an
instance of spanning tree running on the native VLAN, there is now no
way to differentiate BPDUs coming in
tagged vs untagged because they all show up with eth0.  So, there
isn't some kernel knob to get the BPDUs to only come from eth0.100?

On Tue, Mar 6, 2018 at 4:43 PM, David Ahern <dsahern@...il.com> wrote:
> On 3/6/18 3:02 PM, Lawrence Kreeger wrote:
>> Hello,
>>
>> I'm trying to run mstpd on a per VLAN basis using one traditional
>> linux bridge per VLAN.  I'm running it on kernel version 4.12.4.  It
>> works fine for untagged frames, but I'm having a problem with VLAN
>> tagged BPDUs arriving on the socket with the ifindex of the bridge
>> itself, and not the VLAN tagged interface.  For example, I have a
>> tagged interface eth0.100 connected to the bridge "vlan100".  When
>> packets arrive, they have the ifindex of vlan100, which mstpd doesn't
>> recognize as a valid spanning tree interface, so it drops them.  Is
>> there something needed to be set in the kernel to get the ifindex of
>> eth0.100 instead?  This is how mstpd opens the raw socket:
>>
>>
>> /* Berkeley Packet filter code to filter out spanning tree packets.
>>    from tcpdump -s 1152 -dd stp
>>  */
>> static struct sock_filter stp_filter[] = {
>>     { 0x28, 0, 0, 0x0000000c },
>>     { 0x25, 3, 0, 0x000005dc },
>>     { 0x30, 0, 0, 0x0000000e },
>>     { 0x15, 0, 1, 0x00000042 },
>>     { 0x6, 0, 0, 0x00000480 },
>>     { 0x6, 0, 0, 0x00000000 },
>> };
>>
>> /*
>>  * Open up a raw packet socket to catch all 802.2 packets.
>>  * and install a packet filter to only see STP (SAP 42)
>>  *
>>  * Since any bridged devices are already in promiscious mode
>>  * no need to add multicast address.
>>  */
>> int packet_sock_init(void)
>> {
>>     int s;
>>     struct sock_fprog prog =
>>     {
>>         .len = sizeof(stp_filter) / sizeof(stp_filter[0]),
>>         .filter = stp_filter,
>>     };
>>
>>     s = socket(PF_PACKET, SOCK_RAW, htons(ETH_P_802_2));
>
> try ETH_P_ALL
>
>>     if(s < 0)
>>     {
>>         ERROR("socket failed: %m");
>>         return -1;
>>     }
>>
>>     if(setsockopt(s, SOL_SOCKET, SO_ATTACH_FILTER, &prog, sizeof(prog)) < 0)
>>         ERROR("setsockopt packet filter failed: %m");
>>     else if(fcntl(s, F_SETFL, O_NONBLOCK) < 0)
>>         ERROR("fcntl set nonblock failed: %m");
>>     else
>>     {
>>         packet_event.fd = s;
>>         packet_event.handler = packet_rcv;
>
> And then packet_rcv using recvfrom:
>                 struct sockaddr_ll sll;
>                 char buf[4096];
>                 socklen_t alen;
>                 int len;
>
>                 alen = sizeof(sll);
>                 len = recvfrom(sd, buf, sizeof(buf), 0,
>                                 (struct sockaddr *)&sll, &alen);
>
> And sll.sll_ifindex will show vlan device indices.
>
>
>>
>>         if(0 == add_epoll(&packet_event))
>>             return 0;
>>     }
>>
>>     close(s);
>>     return -1;
>> }
>>
>> Thanks, Larry
>>
>