On 27/08/2022 16:17, Ido Schimmel wrote:
> On Sat, Aug 27, 2022 at 02:30:25PM +0300, Nikolay Aleksandrov wrote:
>> On 26/08/2022 14:45, Hans Schultz wrote:
>> Please add the blackhole flag in a separate patch.
>
> +1
>
> [...]
>
>>> @@ -185,6 +196,9 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
>>> if (test_bit(BR_FDB_LOCAL, &dst->flags))
>>> return br_pass_frame_up(skb);
>>>
>>> + if (test_bit(BR_FDB_BLACKHOLE, &dst->flags))
>>> + goto drop;
>>> +
>> Not happy about adding a new test in arguably the most used fast-path, but I don't see
>> a better way to do blackhole right now. Could you please make it an unlikely() ?
>>
>> I guess the blackhole flag will be allowed for user-space to set at some point, why
>> not do it from the start?
>>
>> Actually adding a BR_FDB_LOCAL and BR_FDB_BLACKHOLE would be a conflict above -
>> the packet will be received. So you should move the blackhole check above the
>> BR_FDB_LOCAL one if user-space is allowed to set it to any entry.
>
> Agree about unlikely() and making it writeable from user space from the
> start. This flag is different from the "locked" flag that should only be
> ever set by the kernel.
>
> Regarding BR_FDB_LOCAL, I think BR_FDB_BLACKHOLE should only be allowed
> with BR_FDB_LOCAL as these entries are similar in the following ways:
>
> 1. It doesn't make sense to associate a blackhole entry with a specific
> port. The packet will never be forwarded to this port, but dropped by
> the bridge. This means user space will add them on the bridge itself:
>
Right, good point.
> # bridge fdb add 00:11:22:33:44:55 dev br0 self local blackhole
>
> 2. If you agree that these entries should not be associated with a
> specific port, then it also does not make sense to subject them to
> ageing and roaming, just like existing local/permanent entries.
>
> The above allows us to push the new check under the BR_FDB_LOCAL check:
>
hmm.. so only the driver will be allowed to add non-BR_FDB_LOCAL blackhole
entries with locked flag set as well, that sounds ok as they will be extern_learn
and enforced by it. It is a little discrepancy as we cannot add similar entries in SW
but it really doesn't make any sense to have blackhole fdbs pointing to a port.
SW won't be able to have a locked entry w/ blackhole set, but I like that it is hidden
in the fdb local case when fwding and that's good enough for me.
> diff --git a/net/bridge/br_input.c b/net/bridge/br_input.c
> index 68b3e850bcb9..4357445529a5 100644
> --- a/net/bridge/br_input.c
> +++ b/net/bridge/br_input.c
> @@ -182,8 +182,11 @@ int br_handle_frame_finish(struct net *net, struct sock *sk, struct sk_buff *skb
> if (dst) {
> unsigned long now = jiffies;
>
> - if (test_bit(BR_FDB_LOCAL, &dst->flags))
> + if (test_bit(BR_FDB_LOCAL, &dst->flags)) {
> + if (unlikely(test_bit(BR_FDB_BLACKHOLE, &dst->flags)))
> + goto drop;
> return br_pass_frame_up(skb);
> + }
>
> if (now != dst->used)
> dst->used = now;
