X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fflow.c;h=f769b14d6abc9244d2a6a75d3cd4e545a980f7ec;hb=573c1db99dbfd655d20e0fdf27ff4d460d6529c1;hp=ae60617dd27e08a167e6735420a603e580af51c9;hpb=a26ef51703333b9ab337ba13f397c28c062422fd;p=sliver-openvswitch.git diff --git a/datapath/flow.c b/datapath/flow.c index ae60617dd..f769b14d6 100644 --- a/datapath/flow.c +++ b/datapath/flow.c @@ -1,18 +1,20 @@ /* * Distributed under the terms of the GNU GPL version 2. - * Copyright (c) 2007, 2008, 2009 Nicira Networks. + * Copyright (c) 2007, 2008, 2009, 2010 Nicira Networks. * * Significant portions of this file may be copied from parts of the Linux * kernel, by Linus Torvalds and others. */ #include "flow.h" +#include "datapath.h" #include #include #include #include #include #include +#include #include #include #include @@ -24,11 +26,13 @@ #include #include #include +#include #include #include "compat.h" struct kmem_cache *flow_cache; +static unsigned int hash_seed; struct arp_eth_header { @@ -88,31 +92,22 @@ static inline int icmphdr_ok(struct sk_buff *skb) #define TCP_FLAGS_OFFSET 13 #define TCP_FLAG_MASK 0x3f -static inline struct ovs_tcphdr *ovs_tcp_hdr(const struct sk_buff *skb) -{ - return (struct ovs_tcphdr *)skb_transport_header(skb); -} - void flow_used(struct sw_flow *flow, struct sk_buff *skb) { - unsigned long flags; u8 tcp_flags = 0; - if (flow->key.dl_type == htons(ETH_P_IP) && iphdr_ok(skb)) { - struct iphdr *nh = ip_hdr(skb); - flow->ip_tos = nh->tos; - if (flow->key.nw_proto == IPPROTO_TCP && tcphdr_ok(skb)) { - u8 *tcp = (u8 *)tcp_hdr(skb); - tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK; - } + if (flow->key.dl_type == htons(ETH_P_IP) && + flow->key.nw_proto == IPPROTO_TCP) { + u8 *tcp = (u8 *)tcp_hdr(skb); + tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK; } - spin_lock_irqsave(&flow->lock, flags); - getnstimeofday(&flow->used); + spin_lock_bh(&flow->lock); + flow->used = jiffies; flow->packet_count++; flow->byte_count += skb->len; flow->tcp_flags |= tcp_flags; - spin_unlock_irqrestore(&flow->lock, flags); + spin_unlock_bh(&flow->lock); } struct sw_flow_actions *flow_actions_alloc(size_t n_actions) @@ -133,7 +128,7 @@ struct sw_flow_actions *flow_actions_alloc(size_t n_actions) /* Frees 'flow' immediately. */ -void flow_free(struct sw_flow *flow) +static void flow_free(struct sw_flow *flow) { if (unlikely(!flow)) return; @@ -141,6 +136,12 @@ void flow_free(struct sw_flow *flow) kmem_cache_free(flow_cache, flow); } +void flow_free_tbl(struct tbl_node *node) +{ + struct sw_flow *flow = flow_cast(node); + flow_free(flow); +} + /* RCU callback used by flow_deferred_free. */ static void rcu_free_flow_callback(struct rcu_head *rcu) { @@ -170,23 +171,52 @@ void flow_deferred_free_acts(struct sw_flow_actions *sf_acts) call_rcu(&sf_acts->rcu, rcu_free_acts_callback); } -#define SNAP_OUI_LEN 3 - -struct eth_snap_hdr +static void parse_vlan(struct sk_buff *skb, struct odp_flow_key *key) { - struct ethhdr eth; - u8 dsap; /* Always 0xAA */ - u8 ssap; /* Always 0xAA */ - u8 ctrl; - u8 oui[SNAP_OUI_LEN]; - u16 ethertype; -} __attribute__ ((packed)); - -static int is_snap(const struct eth_snap_hdr *esh) + struct qtag_prefix { + __be16 eth_type; /* ETH_P_8021Q */ + __be16 tci; + }; + struct qtag_prefix *qp; + + if (skb->len < sizeof(struct qtag_prefix) + sizeof(__be16)) + return; + + qp = (struct qtag_prefix *) skb->data; + key->dl_vlan = qp->tci & htons(VLAN_VID_MASK); + key->dl_vlan_pcp = (ntohs(qp->tci) & VLAN_PCP_MASK) >> VLAN_PCP_SHIFT; + __skb_pull(skb, sizeof(struct qtag_prefix)); +} + +static __be16 parse_ethertype(struct sk_buff *skb) { - return (esh->dsap == LLC_SAP_SNAP - && esh->ssap == LLC_SAP_SNAP - && !memcmp(esh->oui, "\0\0\0", 3)); + struct llc_snap_hdr { + u8 dsap; /* Always 0xAA */ + u8 ssap; /* Always 0xAA */ + u8 ctrl; + u8 oui[3]; + u16 ethertype; + }; + struct llc_snap_hdr *llc; + __be16 proto; + + proto = *(__be16 *) skb->data; + __skb_pull(skb, sizeof(__be16)); + + if (ntohs(proto) >= ODP_DL_TYPE_ETH2_CUTOFF) + return proto; + + if (unlikely(skb->len < sizeof(struct llc_snap_hdr))) + return htons(ODP_DL_TYPE_NOT_ETH_TYPE); + + llc = (struct llc_snap_hdr *) skb->data; + if (llc->dsap != LLC_SAP_SNAP || + llc->ssap != LLC_SAP_SNAP || + (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0) + return htons(ODP_DL_TYPE_NOT_ETH_TYPE); + + __skb_pull(skb, sizeof(struct llc_snap_hdr)); + return llc->ethertype; } /* Parses the Ethernet frame in 'skb', which was received on 'in_port', @@ -195,54 +225,40 @@ static int is_snap(const struct eth_snap_hdr *esh) int flow_extract(struct sk_buff *skb, u16 in_port, struct odp_flow_key *key) { struct ethhdr *eth; - struct eth_snap_hdr *esh; int retval = 0; - int nh_ofs; memset(key, 0, sizeof *key); - key->dl_vlan = htons(ODP_VLAN_NONE); + key->tun_id = OVS_CB(skb)->tun_id; key->in_port = in_port; + key->dl_vlan = htons(ODP_VLAN_NONE); if (skb->len < sizeof *eth) return 0; - if (!pskb_may_pull(skb, skb->len >= 64 ? 64 : skb->len)) { + if (!pskb_may_pull(skb, skb->len >= 64 ? 64 : skb->len)) return 0; - } skb_reset_mac_header(skb); - eth = eth_hdr(skb); - esh = (struct eth_snap_hdr *) eth; - nh_ofs = sizeof *eth; - if (likely(ntohs(eth->h_proto) >= ODP_DL_TYPE_ETH2_CUTOFF)) - key->dl_type = eth->h_proto; - else if (skb->len >= sizeof *esh && is_snap(esh)) { - key->dl_type = esh->ethertype; - nh_ofs = sizeof *esh; - } else { - key->dl_type = htons(ODP_DL_TYPE_NOT_ETH_TYPE); - if (skb->len >= nh_ofs + sizeof(struct llc_pdu_un)) { - nh_ofs += sizeof(struct llc_pdu_un); - } - } - /* Check for a VLAN tag */ - if (key->dl_type == htons(ETH_P_8021Q) && - skb->len >= nh_ofs + sizeof(struct vlan_hdr)) { - struct vlan_hdr *vh = (struct vlan_hdr*)(skb->data + nh_ofs); - key->dl_type = vh->h_vlan_encapsulated_proto; - key->dl_vlan = vh->h_vlan_TCI & htons(VLAN_VID_MASK); - nh_ofs += sizeof(struct vlan_hdr); - } + /* Link layer. */ + eth = eth_hdr(skb); memcpy(key->dl_src, eth->h_source, ETH_ALEN); memcpy(key->dl_dst, eth->h_dest, ETH_ALEN); - skb_set_network_header(skb, nh_ofs); + + /* dl_type, dl_vlan, dl_vlan_pcp. */ + __skb_pull(skb, 2 * ETH_ALEN); + if (eth->h_proto == htons(ETH_P_8021Q)) + parse_vlan(skb, key); + key->dl_type = parse_ethertype(skb); + skb_reset_network_header(skb); + __skb_push(skb, skb->data - (unsigned char *)eth); /* Network layer. */ if (key->dl_type == htons(ETH_P_IP) && iphdr_ok(skb)) { struct iphdr *nh = ip_hdr(skb); - int th_ofs = nh_ofs + nh->ihl * 4; + int th_ofs = skb_network_offset(skb) + nh->ihl * 4; key->nw_src = nh->saddr; key->nw_dst = nh->daddr; + key->nw_tos = nh->tos & ~INET_ECN_MASK; key->nw_proto = nh->protocol; skb_set_transport_header(skb, th_ofs); @@ -293,28 +309,41 @@ int flow_extract(struct sk_buff *skb, u16 in_port, struct odp_flow_key *key) arp = (struct arp_eth_header *)skb_network_header(skb); - if (arp->ar_hrd == htons(1) - && arp->ar_pro == htons(ETH_P_IP) - && arp->ar_hln == ETH_ALEN - && arp->ar_pln == 4) { - - /* We only match on the lower 8 bits of the opcode. */ - if (ntohs(arp->ar_op) <= 0xff) { - key->nw_proto = ntohs(arp->ar_op); - } - - if (key->nw_proto == ARPOP_REQUEST - || key->nw_proto == ARPOP_REPLY) { - memcpy(&key->nw_src, arp->ar_sip, sizeof(key->nw_src)); - memcpy(&key->nw_dst, arp->ar_tip, sizeof(key->nw_dst)); - } - } + if (arp->ar_hrd == htons(ARPHRD_ETHER) + && arp->ar_pro == htons(ETH_P_IP) + && arp->ar_hln == ETH_ALEN + && arp->ar_pln == 4) { + + /* We only match on the lower 8 bits of the opcode. */ + if (ntohs(arp->ar_op) <= 0xff) { + key->nw_proto = ntohs(arp->ar_op); + } + + if (key->nw_proto == ARPOP_REQUEST + || key->nw_proto == ARPOP_REPLY) { + memcpy(&key->nw_src, arp->ar_sip, sizeof(key->nw_src)); + memcpy(&key->nw_dst, arp->ar_tip, sizeof(key->nw_dst)); + } + } } else { skb_reset_transport_header(skb); } return retval; } +u32 flow_hash(const struct odp_flow_key *key) +{ + return jhash2((u32*)key, sizeof *key / sizeof(u32), hash_seed); +} + +int flow_cmp(const struct tbl_node *node, void *key2_) +{ + const struct odp_flow_key *key1 = &flow_cast(node)->key; + const struct odp_flow_key *key2 = key2_; + + return !memcmp(key1, key2, sizeof(struct odp_flow_key)); +} + /* Initializes the flow module. * Returns zero if successful or a negative error code. */ int flow_init(void) @@ -324,6 +353,8 @@ int flow_init(void) if (flow_cache == NULL) return -ENOMEM; + get_random_bytes(&hash_seed, sizeof hash_seed); + return 0; } @@ -332,16 +363,3 @@ void flow_exit(void) { kmem_cache_destroy(flow_cache); } - -void print_flow(const struct odp_flow_key *key) -{ -#define MAC_FMT "%02x:%02x:%02x:%02x:%02x:%02x" -#define MAC_ARG(x) ((u8*)(x))[0],((u8*)(x))[1],((u8*)(x))[2],((u8*)(x))[3],((u8*)(x))[4],((u8*)(x))[5] - printk("port%04x:vlan%d mac"MAC_FMT"->"MAC_FMT" " - "type%04x proto%d ip%x->%x port%d->%d\n", - key->in_port, ntohs(key->dl_vlan), - MAC_ARG(key->dl_src), MAC_ARG(key->dl_dst), - ntohs(key->dl_type), key->nw_proto, - key->nw_src, key->nw_dst, - ntohs(key->tp_src), ntohs(key->tp_dst)); -}