On Wed, Aug 6, 2014 at 3:04 PM, Andy Zhou <[email protected]> wrote:
> Looks good in general. Some minor comments inline.
>
> Acked-by: Andy Zhou <[email protected]>
>
> On Tue, Aug 5, 2014 at 4:46 PM, Pravin B Shelar <[email protected]> wrote:
>> Recirc action needs to extract flow key from packet, it uses tun_info
>> from OVS_CB for setting tunnel meta data in flow key. But tun_info
>> can be overwritten by tunnel send action. This would result in wrong
>> flow key for the recirculation.
>> Following patch copies flow-key meta data from OVS_CB packet key
>> itself thus avoids this bug.
>>
>> Signed-off-by: Pravin B Shelar <[email protected]>
>> ---
>> datapath/actions.c | 6 ++----
>> datapath/flow.c | 10 ++++++++++
>> datapath/flow.h | 9 +++++++++
>> datapath/flow_netlink.c | 7 +------
>> 4 files changed, 22 insertions(+), 10 deletions(-)
>>
>> diff --git a/datapath/actions.c b/datapath/actions.c
>> index 5a0bfa9..6de65d3 100644
>> --- a/datapath/actions.c
>> +++ b/datapath/actions.c
>> @@ -649,17 +649,15 @@ static int execute_recirc(struct datapath *dp, struct
>> sk_buff *skb,
>> const struct nlattr *a)
>> {
>> struct sw_flow_key recirc_key;
>> - uint32_t hash = OVS_CB(skb)->pkt_key->ovs_flow_hash;
>> int err;
>>
>> - err = ovs_flow_key_extract(skb, &recirc_key);
>> + err = ovs_flow_key_extract_recirc(nla_get_u32(a),
>> OVS_CB(skb)->pkt_key,
>> + skb, &recirc_key);
>> if (err) {
>> kfree_skb(skb);
>> return err;
>> }
>>
>> - recirc_key.ovs_flow_hash = hash;
>> - recirc_key.recirc_id = nla_get_u32(a);
>>
>> ovs_dp_process_packet_with_key(skb, &recirc_key, true);
>>
>> diff --git a/datapath/flow.c b/datapath/flow.c
>> index 7fdce6a..1feca85 100644
>> --- a/datapath/flow.c
>> +++ b/datapath/flow.c
>> @@ -718,3 +718,13 @@ int ovs_flow_key_extract_userspace(const struct nlattr
>> *attr,
>>
>> return key_extract(skb, key);
>> }
>> +
>> +int ovs_flow_key_extract_recirc(u32 recirc_id,
>
> Could we make the recirc_id argument as const?
>
what is advantage of making recirc_is const.
>> + const struct sw_flow_key *key,
>> + struct sk_buff *skb,
>> + struct sw_flow_key *new_key)
>> +{
>> + memcpy(new_key, key, OVS_SW_FLOW_KEY_METADATA_SIZE);
> It is a bit odd to make recirc_id 0 then overwrite it just one line
> below. May be we can memcpy sizeof (recirc_id) less bytes?
I am also using same macro to clear flow key meta data, so lets keep
it as it is.
I will push first two patches shortly.
>> + new_key->recirc_id = recirc_id;
>> + return key_extract(skb, new_key);
>> +}
>> diff --git a/datapath/flow.h b/datapath/flow.h
>> index 69664cd..ee63b8b 100644
>> --- a/datapath/flow.h
>> +++ b/datapath/flow.h
>> @@ -87,6 +87,11 @@ static inline void ovs_flow_tun_info_init(struct
>> ovs_tunnel_info *tun_info,
>> tun_info->options_len = opts_len;
>> }
>>
>> +#define OVS_SW_FLOW_KEY_METADATA_SIZE \
>> + (offsetof(struct sw_flow_key, recirc_id) + \
>> + FIELD_SIZEOF(struct sw_flow_key, recirc_id))
>> +
>> +
>> struct sw_flow_key {
>> u8 tun_opts[255];
>> u8 tun_opts_len;
>> @@ -222,5 +227,9 @@ int ovs_flow_key_extract(struct sk_buff *skb, struct
>> sw_flow_key *key);
>> int ovs_flow_key_extract_userspace(const struct nlattr *attr,
>> struct sk_buff *skb,
>> struct sw_flow_key *key);
>> +int ovs_flow_key_extract_recirc(u32 recirc_id,
>
> same as above comment.
>> + const struct sw_flow_key *key,
>> + struct sk_buff *skb,
>> + struct sw_flow_key *new_key);
>>
>> #endif /* flow.h */
>> diff --git a/datapath/flow_netlink.c b/datapath/flow_netlink.c
>> index ada21ae..75172de 100644
>> --- a/datapath/flow_netlink.c
>> +++ b/datapath/flow_netlink.c
>> @@ -1016,13 +1016,8 @@ int ovs_nla_get_flow_metadata(const struct nlattr
>> *attr,
>> memset(&match, 0, sizeof(match));
>> match.key = key;
>>
>> - key->tun_opts_len = 0;
>> - memset(&key->tun_key, 0, sizeof(key->tun_key));
>> - key->phy.priority = 0;
>> - key->phy.skb_mark = 0;
>> + memset(key, 0, OVS_SW_FLOW_KEY_METADATA_SIZE);
>> key->phy.in_port = DP_MAX_PORTS;
>> - key->ovs_flow_hash = 0;
>> - key->recirc_id = 0;
>>
>> return metadata_from_nlattrs(&match, &attrs, a, false);
>> }
>> --
>> 1.7.1
>>
>> _______________________________________________
>> dev mailing list
>> [email protected]
>> http://openvswitch.org/mailman/listinfo/dev
_______________________________________________
dev mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/dev
