On Wed, May 27, 2015 at 10:57 PM, Pravin Shelar <pshe...@nicira.com> wrote:
> On Wed, May 27, 2015 at 9:16 PM, Jesse Gross <je...@nicira.com> wrote:
>> On Wed, May 27, 2015 at 7:46 PM, Pravin Shelar <pshe...@nicira.com> wrote:
>>> On Wed, May 27, 2015 at 2:10 PM, Jesse Gross <je...@nicira.com> wrote:
>>>> On Fri, May 22, 2015 at 10:53 AM, Pravin Shelar <pshe...@nicira.com> wrote:
>>>>> On Wed, May 20, 2015 at 12:32 PM, Neil McKee <neil.mc...@inmon.com> wrote:
>>>>>> diff --git a/net/openvswitch/actions.c b/net/openvswitch/actions.c
>>>>>> index b491c1c..ee5760d 100644
>>>>>> --- a/net/openvswitch/actions.c
>>>>>> +++ b/net/openvswitch/actions.c
>>>>>> @@ -608,7 +608,8 @@ static void do_output(struct datapath *dp, struct
>>>>>> sk_buff *skb, int out_port)
>>>>>> }
>>>>>>
>>>>>> static int output_userspace(struct datapath *dp, struct sk_buff *skb,
>>>>>> - struct sw_flow_key *key, const struct nlattr
>>>>>> *attr)
>>>>>> + struct sw_flow_key *key, const struct nlattr
>>>>>> *attr,
>>>>>> + const struct nlattr *actions, int
>>>>>> actions_len)
>>>>>> {
>>>>>> struct ovs_tunnel_info info;
>>>>>> struct dp_upcall_info upcall;
>>>>>> @@ -619,6 +620,8 @@ static int output_userspace(struct datapath *dp,
>>>>>> struct sk_buff *skb,
>>>>>> upcall.userdata = NULL;
>>>>>> upcall.portid = 0;
>>>>>> upcall.egress_tun_info = NULL;
>>>>>> + upcall.actions = actions;
>>>>>> + upcall.actions_len = actions_len;
>>>>>>
>>>>> Rather than unconditionally passing actions to the upcall, there
>>>>> should be attribute in ovs_userspace_attr to request the actions list.
>>>>
>>>> Why? It seems simpler to just always pass the actions and I'm not sure
>>>> that this is really performance critical (which is the only reason
>>>> that comes to mind to not always pass this).
>>>
>>> This is only required for sFlow sampling so I do not think we should
>>> send it on every upcall.
>>
>> But what is the downside?
>
> This increases memory allocation in atomic context but if you think
> this makes code complicated then I am fine without the attribute.
OK, I see.
My guess is that there are only likely to be a significant set of
actions for sampling use cases anyways so if this is a real problem
then a flag is probably not going to make much of a difference.
One possibility is to retry with a smaller size if allocation fails
and not include the actions in that case. Userspace is already going
to have to handle the case where actions are omitted for existing
kernels.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
