Nicolas Dichtel <nicolas.dichtel@xxxxxxxxx> writes:
> Le 06/10/2015 20:53, Eric W. Biederman a écrit :
>> It is confusing and silly hiding a paramater so modify all of
>> the callers to pass in the appropriate socket or skb->sk if
>> no socket is known.
>>
>> Signed-off-by: "Eric W. Biederman" <ebiederm@xxxxxxxxxxxx>
>> ---
> [snip]
>> @@ -456,7 +456,7 @@ packet_routed:
>> skb->priority = sk->sk_priority;
>> skb->mark = sk->sk_mark;
>>
>> - res = ip_local_out(skb);
>> + res = ip_local_out(sk, skb);
> As stated in the comment at the top of this function (ip_queue_xmit()), skb->sk
> can be different from sk. See also commit b0270e91014d ("ipv4: add a sock
> pointer to ip_queue_xmit()").
> Not sure if this change is right.
Good catch. This change should not have been buried in this patch. It
needs to be it's own separate bug fix.
As I read the code we actually do want to pass sk not skb->sk into
ip_local_out. For all of the reasons that sk is potentially different
from skb->sk already.
The way I understand this is we have pushed an sk parameter through the
output path so that sk_mc_loop(sk) can be called with the tunnel's
socket not whatever is on skb->sk. This allows for looking to see if
local multicast loopback is configured on the tunnels socket not on the
originating socket of the packet.
I am going to respin my series with that change made into a separate bug
fix, that can potentially be backported.
Eric
--
To unsubscribe from this list: send the line "unsubscribe lvs-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
