Thanks a lot for the review Pravin. I will send out v2 soon.
On Tue, Apr 15, 2014 at 7:14 AM, Pravin Shelar <pshe...@nicira.com> wrote:
> On Sat, Apr 12, 2014 at 3:30 AM, Andy Zhou <az...@nicira.com> wrote:
>> Implements Linux kernel datapath recirc action. Recirc action allows
>> kernel data patch reprocess the packet as if it is received from its
>> original input port, with a recirc_id added to its flow key.
>>
>> The recirc action follows a sub routine call model. The packet change
>> by actions during recirculation is invisible across a recirculation;
>> any action following the recirc action will process the same packet
>> prior to the recirc action.
>>
>> Signed-off-by: Andy Zhou <az...@nicira.com>
>> ---
>> datapath/actions.c | 54
>> ++++++++++++++++++++++++++++++++++++++++++++++-
>> datapath/datapath.c | 37 +++++++++++++++++++-------------
>> datapath/datapath.h | 3 ++-
>> datapath/flow_netlink.c | 31 ++++++++++++++++++++++++++-
>> 4 files changed, 107 insertions(+), 18 deletions(-)
>>
>> diff --git a/datapath/actions.c b/datapath/actions.c
>> index cb239c8..c833c69 100644
>> --- a/datapath/actions.c
>> +++ b/datapath/actions.c
>> @@ -1,5 +1,5 @@
>> /*
>> - * Copyright (c) 2007-2013 Nicira, Inc.
>> + * Copyright (c) 2007-2014 Nicira, Inc.
>> *
>> * This program is free software; you can redistribute it and/or
>> * modify it under the terms of version 2 of the GNU General Public
>> @@ -521,6 +521,34 @@ static int execute_set_action(struct sk_buff *skb,
>> return err;
>> }
>>
>> +static int execute_recirc(struct datapath *dp, struct sk_buff *skb,
>> + const struct nlattr *recirc_attr,
>> + uint16_t in_port, uint32_t dp_hash)
>> +{
>> + struct sw_flow_key recirc_key;
>> + struct vport *p;
>> + int err;
>> +
>> + err = ovs_flow_extract(skb, in_port, &recirc_key);
>> + if (err)
>> + return err;
>> +
>> + recirc_key.dp_hash = dp_hash;
>> + recirc_key.recirc_id = nla_get_u32(recirc_attr);
>> + OVS_CB(skb)->pkt_key = &recirc_key;
>> + OVS_CB(skb)->flow = NULL;
>> +
> Do we need to reset flow here?
>
Yes, we can skip it.
>> + p = ovs_vport_rcu(dp, in_port);
>> + if (unlikely(!p)) {
>> + kfree_skb(skb);
>> + return -ENODEV;
>> + }
>> +
>> + ovs_dp_process_packet_with_key(p, skb);
>> +
> Can we avoid flow-extract and vport lookup. it does add per packet
> processing cost.
I agree this may be optimized more. But should not affect correctness. Also
notice this will not affect current use case where the recirc action
is the last action.
flow_extract is needed in case the packet is changed by previous
actions. One obvious
alternative is to maintain pkt_key whenever set action is applied to
the packet. But this
may affect the efficiency of non-recirc use case. At any rate, we need
to collect some data
to help decide. It is better to be addressed in a separate patch. What
do you think?
>
>> + return 0;
>> +}
>> +
>> /* Execute a list of actions against 'skb'. */
>> static int do_execute_actions(struct datapath *dp, struct sk_buff *skb,
>> const struct nlattr *attr, int len, bool keep_skb)
>> @@ -565,12 +593,36 @@ static int do_execute_actions(struct datapath *dp,
>> struct sk_buff *skb,
>> err = pop_vlan(skb);
>> break;
>>
>> + case OVS_ACTION_ATTR_RECIRC: {
>> + struct sk_buff *recirc_skb;
>> + const bool last_action = (a->nla_len == rem);
>> + uint16_t in_port = OVS_CB(skb)->pkt_key->phy.in_port;
>> + uint32_t dp_hash = OVS_CB(skb)->pkt_key->dp_hash;
>> +
>> + if (!last_action || keep_skb)
>> + recirc_skb = skb_clone(skb, GFP_ATOMIC);
>> + else
>> + recirc_skb = skb;
>> +
>> + err = execute_recirc(dp, recirc_skb, a, in_port,
>> + dp_hash);
>> +
> Why are hash and in-port are explicitly passed to execute_recirc()?
>
>> + /* Return directly if recirc is the last action
>> + * or err. */
>> + if (last_action || err)
>> + return err;
>> +
>> + break;
>> + }
>> +
>> case OVS_ACTION_ATTR_SET:
>> err = execute_set_action(skb, nla_data(a));
>> break;
>>
>> case OVS_ACTION_ATTR_SAMPLE:
>> err = sample(dp, skb, a);
>> + if (unlikely(err)) /* skb already freed. */
>> + return err;
>> break;
> good catch.
> This is bug fix and need to be back-ported. Can you send separate patch?
Yes,
>
>> }
>>
>> diff --git a/datapath/datapath.c b/datapath/datapath.c
>> index 25edd7d..dab4a6b 100644
>> --- a/datapath/datapath.c
>> +++ b/datapath/datapath.c
>> @@ -240,33 +240,24 @@ void ovs_dp_detach_port(struct vport *p)
>> ovs_vport_del(p);
>> }
>>
>> -/* Must be called with rcu_read_lock. */
>> -void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
>> +void ovs_dp_process_packet_with_key(const struct vport *p, struct sk_buff
>> *skb)
>> {
>> struct datapath *dp = p->dp;
>> struct sw_flow *flow;
>> struct dp_stats_percpu *stats;
>> - struct sw_flow_key key;
>> + struct sw_flow_key *key = OVS_CB(skb)->pkt_key;
>> u64 *stats_counter;
>> u32 n_mask_hit;
>> - int error;
>>
>> stats = this_cpu_ptr(dp->stats_percpu);
>>
>> - /* Extract flow from 'skb' into 'key'. */
>> - error = ovs_flow_extract(skb, p->port_no, &key);
>> - if (unlikely(error)) {
>> - kfree_skb(skb);
>> - return;
>> - }
>> -
>> /* Look up flow. */
>> - flow = ovs_flow_tbl_lookup_stats(&dp->table, &key, &n_mask_hit);
>> + flow = ovs_flow_tbl_lookup_stats(&dp->table, key, &n_mask_hit);
>> if (unlikely(!flow)) {
>> struct dp_upcall_info upcall;
>>
>> upcall.cmd = OVS_PACKET_CMD_MISS;
>> - upcall.key = &key;
>> + upcall.key = key;
>> upcall.userdata = NULL;
>> upcall.portid = p->upcall_portid;
>> ovs_dp_upcall(dp, skb, &upcall);
>> @@ -276,9 +267,8 @@ void ovs_dp_process_received_packet(struct vport *p,
>> struct sk_buff *skb)
>> }
>>
>> OVS_CB(skb)->flow = flow;
>> - OVS_CB(skb)->pkt_key = &key;
>>
>> - ovs_flow_stats_update(OVS_CB(skb)->flow, key.tp.flags, skb);
>> + ovs_flow_stats_update(OVS_CB(skb)->flow, key->tp.flags, skb);
>> ovs_execute_actions(dp, skb);
>> stats_counter = &stats->n_hit;
>>
>> @@ -290,6 +280,23 @@ out:
>> u64_stats_update_end(&stats->sync);
>> }
>>
>> +/* Must be called with rcu_read_lock. */
>> +void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
>> +{
>> + int error;
>> + struct sw_flow_key key;
>> +
>> + /* Extract flow from 'skb' into 'key'. */
>> + error = ovs_flow_extract(skb, p->port_no, &key);
>> + if (unlikely(error)) {
>> + kfree_skb(skb);
>> + return;
>> + }
>> +
>> + OVS_CB(skb)->pkt_key = &key;
>> + ovs_dp_process_packet_with_key(p, skb);
>> +}
> Can you pass key ovs_dp_process_packet_with_key(), so that we can only
> one function can set OVS_CB flow and key.
O.K.
>
>> +
>> int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
>> const struct dp_upcall_info *upcall_info)
>> {
>> diff --git a/datapath/datapath.h b/datapath/datapath.h
>> index 40e0f90..d3b06ee 100644
>> --- a/datapath/datapath.h
>> +++ b/datapath/datapath.h
>> @@ -1,5 +1,5 @@
>> /*
>> - * Copyright (c) 2007-2012 Nicira, Inc.
>> + * Copyright (c) 2007-2014 Nicira, Inc.
>> *
>> * This program is free software; you can redistribute it and/or
>> * modify it under the terms of version 2 of the GNU General Public
>> @@ -188,6 +188,7 @@ extern struct genl_family dp_vport_genl_family;
>> extern struct genl_multicast_group ovs_dp_vport_multicast_group;
>>
>> void ovs_dp_process_received_packet(struct vport *, struct sk_buff *);
>> +void ovs_dp_process_packet_with_key(const struct vport *, struct sk_buff *);
>> void ovs_dp_detach_port(struct vport *);
>> int ovs_dp_upcall(struct datapath *, struct sk_buff *,
>> const struct dp_upcall_info *);
>> diff --git a/datapath/flow_netlink.c b/datapath/flow_netlink.c
>> index e85326b..f5249ee 100644
>> --- a/datapath/flow_netlink.c
>> +++ b/datapath/flow_netlink.c
>> @@ -129,7 +129,8 @@ static bool match_validate(const struct sw_flow_match
>> *match,
>> /* Always allowed mask fields. */
>> mask_allowed |= ((1ULL << OVS_KEY_ATTR_TUNNEL)
>> | (1ULL << OVS_KEY_ATTR_IN_PORT)
>> - | (1ULL << OVS_KEY_ATTR_ETHERTYPE));
>> + | (1ULL << OVS_KEY_ATTR_ETHERTYPE)
>> + | (1ULL << OVS_KEY_ATTR_RECIRC_ID));
>>
>> /* Check key attributes. */
>> if (match->key->dp_hash) {
>> @@ -258,6 +259,7 @@ static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
>> [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
>> [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
>> [OVS_KEY_ATTR_DP_HASH] = sizeof(u32),
>> + [OVS_KEY_ATTR_RECIRC_ID] = sizeof(u32),
>> [OVS_KEY_ATTR_TUNNEL] = -1,
>> };
>>
>> @@ -474,6 +476,23 @@ static int metadata_from_nlattrs(struct sw_flow_match
>> *match, u64 *attrs,
>> *attrs &= ~(1ULL << OVS_KEY_ATTR_DP_HASH);
>> }
>>
>> + if (*attrs & (1ULL << OVS_KEY_ATTR_RECIRC_ID)) {
>> + u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
>> +
>> + if (is_mask && (recirc_id > 0 && recirc_id < UINT_MAX)) {
>> + OVS_NLERR("Reicrc_id mask is neither wildcard nor
>> exact match\n");
>> + return -EINVAL;
>> + }
>> +
>> + SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
>> + *attrs &= ~(1ULL << OVS_KEY_ATTR_RECIRC_ID);
>> + }
>> +
>> + if (is_mask) {
>> + /* Always exact match recirc_id. */
>> + SW_FLOW_KEY_PUT(match, recirc_id, UINT_MAX, is_mask);
>> + }
> I am not sure why is this done unconditionally? This does increases
> range for flow match.
>> +
>> if (*attrs & (1ULL << OVS_KEY_ATTR_PRIORITY)) {
>> SW_FLOW_KEY_PUT(match, phy.priority,
>> nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
>> @@ -879,6 +898,7 @@ int ovs_nla_get_flow_metadata(struct sw_flow *flow,
>> flow->key.phy.priority = 0;
>> flow->key.phy.skb_mark = 0;
>> flow->key.dp_hash = 0;
>> + flow->key.recirc_id = 0;
>
> Comment need a update.
>> memset(tun_key, 0, sizeof(flow->key.tun_key));
>>
>> err = parse_flow_nlattrs(attr, a, &attrs);
>> @@ -902,10 +922,15 @@ int ovs_nla_put_flow(const struct sw_flow_key *swkey,
>> struct nlattr *nla, *encap;
>> bool is_mask = (swkey != output);
>>
>> + /* dp_hash and recric_id are only valid if they are nonzero. */
>> if (swkey->dp_hash)
>> if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->dp_hash))
>> goto nla_put_failure;
>>
>> + if (swkey->recirc_id)
>> + if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID,
>> output->recirc_id))
>> + goto nla_put_failure;
>> +
> Can you use && rather than nested if statement?
>
>> if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
>> goto nla_put_failure;
>>
>> @@ -1442,6 +1467,7 @@ int ovs_nla_copy_actions(const struct nlattr *attr,
>> /* Expected argument lengths, (u32)-1 for variable length. */
>> static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
>> [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
>> + [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
>> [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
>> [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct
>> ovs_action_push_vlan),
>> [OVS_ACTION_ATTR_POP_VLAN] = 0,
>> @@ -1488,6 +1514,9 @@ int ovs_nla_copy_actions(const struct nlattr *attr,
>> return -EINVAL;
>> break;
>>
>> + case OVS_ACTION_ATTR_RECIRC:
>> + break;
>> +
>> case OVS_ACTION_ATTR_SET:
>> err = validate_set(a, key, sfa, &skip_copy);
>> if (err)
>> --
>> 1.7.9.5
>>
>> _______________________________________________
>> dev mailing list
>> dev@openvswitch.org
>> http://openvswitch.org/mailman/listinfo/dev
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
