[ovs-dev] [PATCH] datapath: refactor netlink flow handling for mega flow support
Andy Zhou
azhou at nicira.com
Thu May 2 19:36:35 UTC 2013
This patch refactors the netlink flow handling code to make
the processing, error checking and prerequisite checking
more stream lined.
Signed-off-by: Andy Zhou <azhou at nicira.com>
---
datapath/datapath.c | 44 +++-
datapath/flow.c | 652 ++++++++++++++++++++++++++++-----------------------
datapath/flow.h | 16 +-
3 files changed, 411 insertions(+), 301 deletions(-)
diff --git a/datapath/datapath.c b/datapath/datapath.c
index 42af315..8cf75bf 100644
--- a/datapath/datapath.c
+++ b/datapath/datapath.c
@@ -619,10 +619,14 @@ static int validate_tp_port(const struct sw_flow_key *flow_key)
static int validate_and_copy_set_tun(const struct nlattr *attr,
struct sw_flow_actions **sfa)
{
- struct ovs_key_ipv4_tunnel tun_key;
+ struct sw_flow_match match;
+ struct sw_flow_key key;
int err, start;
- err = ipv4_tun_from_nlattr(nla_data(attr), &tun_key);
+ memset(&key, 0, sizeof key);
+ ovs_match_init(&match, &key);
+
+ err = ipv4_tun_from_nlattr(nla_data(attr), &match);
if (err)
return err;
@@ -630,7 +634,8 @@ static int validate_and_copy_set_tun(const struct nlattr *attr,
if (start < 0)
return start;
- err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &tun_key, sizeof(tun_key));
+ err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &match.key->tun_key,
+ sizeof(match.key->tun_key));
add_nested_action_end(*sfa, start);
return err;
@@ -1234,6 +1239,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
struct datapath *dp;
struct flow_table *table;
struct sw_flow_actions *acts = NULL;
+ struct sw_flow_match match;
int error;
int key_len;
@@ -1241,10 +1247,18 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
error = -EINVAL;
if (!a[OVS_FLOW_ATTR_KEY])
goto error;
- error = ovs_flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]);
+
+ memset(&key, 0, sizeof key);
+ ovs_match_init(&match, &key);
+ error = ovs_match_from_nlattrs(&match, a[OVS_FLOW_ATTR_KEY]);
if (error)
goto error;
+ if (ovs_match_check_prerequisite(&match) == false)
+ goto error;
+
+ key_len = ovs_match_get_len(&match);
+
/* Validate actions. */
if (a[OVS_FLOW_ATTR_ACTIONS]) {
acts = ovs_flow_actions_alloc(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
@@ -1358,15 +1372,24 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
struct sw_flow *flow;
struct datapath *dp;
struct flow_table *table;
+ struct sw_flow_match match;
int err;
int key_len;
+ memset(&key, 0, sizeof key);
+ ovs_match_init(&match, &key);
+
if (!a[OVS_FLOW_ATTR_KEY])
return -EINVAL;
- err = ovs_flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]);
+ err = ovs_match_from_nlattrs(&match, a[OVS_FLOW_ATTR_KEY]);
if (err)
return err;
+ if (ovs_match_check_prerequisite(&match) == false)
+ return -EINVAL;
+
+ key_len = ovs_match_get_len(&match);
+
ovs_lock();
dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
if (!dp) {
@@ -1404,6 +1427,7 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
struct sw_flow *flow;
struct datapath *dp;
struct flow_table *table;
+ struct sw_flow_match match;
int err;
int key_len;
@@ -1418,10 +1442,18 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
err = flush_flows(dp);
goto unlock;
}
- err = ovs_flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]);
+
+ memset(&key, 0, sizeof key);
+ ovs_match_init(&match, &key);
+ err = ovs_match_from_nlattrs(&match, a[OVS_FLOW_ATTR_KEY]);
if (err)
goto unlock;
+ if (ovs_match_check_prerequisite(&match) == false)
+ goto unlock;
+
+ key_len = ovs_match_get_len(&match);
+
table = ovsl_dereference(dp->table);
flow = ovs_flow_tbl_lookup(table, &key, key_len);
if (!flow) {
diff --git a/datapath/flow.c b/datapath/flow.c
index 3ce926e..914051d 100644
--- a/datapath/flow.c
+++ b/datapath/flow.c
@@ -47,6 +47,110 @@
static struct kmem_cache *flow_cache;
+static void field_add_writes__(struct sw_flow_match *match,
+ size_t offset, size_t size)
+{
+ int i;
+ u8 *op = &(match->n_writes[offset]);
+
+ for (i = 0; i < size; i++, op++) {
+ ++*op;
+ }
+}
+
+#define SW_FLOW_KEY_PUT(match, field, value) \
+ do { \
+ field_add_writes__(match, offsetof(struct sw_flow_key, field), \
+ sizeof (match)->key->field ); \
+ (match)->key->field = value; \
+ } while (0)
+
+#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len) \
+ do { \
+ field_add_writes__(match, \
+ offsetof(struct sw_flow_key, field), len); \
+ memcpy(&(match)->key->field, value_p, len); \
+ } while (0)
+
+void ovs_match_init(struct sw_flow_match *match, struct sw_flow_key *key)
+{
+ memset(match, 0, sizeof *match);
+ match->key = key;
+}
+
+bool ovs_match_check_prerequisite(const struct sw_flow_match *match)
+{
+ int i;
+ u64 expected = 0;
+
+ /* Any duplicate writes to the matching key is invalid. */
+ for (i=0; i< sizeof *(match->key); i++)
+ if (match->n_writes[i] >= 2) {
+ return false;
+ }
+
+ if (match->key->eth.type == htons(ETH_P_ARP))
+ expected |= 1ULL << OVS_KEY_ATTR_ARP;
+
+ if (match->key->eth.type == htons(ETH_P_RARP))
+ expected |= 1ULL << OVS_KEY_ATTR_ARP;
+
+ if (match->key->eth.type == htons(ETH_P_IP)) {
+ expected |= 1ULL << OVS_KEY_ATTR_IPV4;
+
+ if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
+ if (match->key->ip.proto == IPPROTO_UDP)
+ expected |= 1ULL << OVS_KEY_ATTR_UDP;
+
+ if (match->key->ip.proto == IPPROTO_TCP)
+ expected |= 1ULL << OVS_KEY_ATTR_TCP;
+
+ if (match->key->ip.proto == IPPROTO_ICMP)
+ expected |= 1ULL << OVS_KEY_ATTR_ICMP;
+ }
+ }
+
+ if (match->key->eth.type == htons(ETH_P_IPV6)) {
+ expected |= 1ULL << OVS_KEY_ATTR_IPV6;
+
+ if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
+ if (match->key->ip.proto == IPPROTO_UDP)
+ expected |= 1ULL << OVS_KEY_ATTR_UDP;
+
+ if (match->key->ip.proto == IPPROTO_TCP)
+ expected |= 1ULL << OVS_KEY_ATTR_TCP;
+
+ if (match->key->ip.proto == IPPROTO_ICMPV6) {
+ expected |= 1ULL << OVS_KEY_ATTR_ICMPV6;
+
+ if (match->key->ipv6.tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION))
+ expected |= 1ULL << OVS_KEY_ATTR_ND;
+
+ if (match->key->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT))
+ expected |= 1ULL << OVS_KEY_ATTR_ND;
+ }
+ }
+ }
+
+ if ((match->key_attrs & expected) != expected)
+ return false;
+
+ return true;
+}
+
+int ovs_match_get_len(struct sw_flow_match *match)
+{
+ int len;
+
+ for (len = sizeof *(match->key); len >0; len--) {
+ if (match->n_writes[len-1] > 0) {
+ return len;
+ }
+ }
+
+ return 0;
+}
+
static int check_header(struct sk_buff *skb, int len)
{
if (unlikely(skb->len < len))
@@ -850,112 +954,6 @@ const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
[OVS_KEY_ATTR_TUNNEL] = -1,
};
-static int ipv4_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len,
- const struct nlattr *a[], u64 *attrs)
-{
- const struct ovs_key_icmp *icmp_key;
- const struct ovs_key_tcp *tcp_key;
- const struct ovs_key_udp *udp_key;
-
- switch (swkey->ip.proto) {
- case IPPROTO_TCP:
- if (!(*attrs & (1 << OVS_KEY_ATTR_TCP)))
- return -EINVAL;
- *attrs &= ~(1 << OVS_KEY_ATTR_TCP);
-
- *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
- tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
- swkey->ipv4.tp.src = tcp_key->tcp_src;
- swkey->ipv4.tp.dst = tcp_key->tcp_dst;
- break;
-
- case IPPROTO_UDP:
- if (!(*attrs & (1 << OVS_KEY_ATTR_UDP)))
- return -EINVAL;
- *attrs &= ~(1 << OVS_KEY_ATTR_UDP);
-
- *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
- udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
- swkey->ipv4.tp.src = udp_key->udp_src;
- swkey->ipv4.tp.dst = udp_key->udp_dst;
- break;
-
- case IPPROTO_ICMP:
- if (!(*attrs & (1 << OVS_KEY_ATTR_ICMP)))
- return -EINVAL;
- *attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
-
- *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
- icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
- swkey->ipv4.tp.src = htons(icmp_key->icmp_type);
- swkey->ipv4.tp.dst = htons(icmp_key->icmp_code);
- break;
- }
-
- return 0;
-}
-
-static int ipv6_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len,
- const struct nlattr *a[], u64 *attrs)
-{
- const struct ovs_key_icmpv6 *icmpv6_key;
- const struct ovs_key_tcp *tcp_key;
- const struct ovs_key_udp *udp_key;
-
- switch (swkey->ip.proto) {
- case IPPROTO_TCP:
- if (!(*attrs & (1 << OVS_KEY_ATTR_TCP)))
- return -EINVAL;
- *attrs &= ~(1 << OVS_KEY_ATTR_TCP);
-
- *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
- tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
- swkey->ipv6.tp.src = tcp_key->tcp_src;
- swkey->ipv6.tp.dst = tcp_key->tcp_dst;
- break;
-
- case IPPROTO_UDP:
- if (!(*attrs & (1 << OVS_KEY_ATTR_UDP)))
- return -EINVAL;
- *attrs &= ~(1 << OVS_KEY_ATTR_UDP);
-
- *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
- udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
- swkey->ipv6.tp.src = udp_key->udp_src;
- swkey->ipv6.tp.dst = udp_key->udp_dst;
- break;
-
- case IPPROTO_ICMPV6:
- if (!(*attrs & (1 << OVS_KEY_ATTR_ICMPV6)))
- return -EINVAL;
- *attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
-
- *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
- icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
- swkey->ipv6.tp.src = htons(icmpv6_key->icmpv6_type);
- swkey->ipv6.tp.dst = htons(icmpv6_key->icmpv6_code);
-
- if (swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
- swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
- const struct ovs_key_nd *nd_key;
-
- if (!(*attrs & (1 << OVS_KEY_ATTR_ND)))
- return -EINVAL;
- *attrs &= ~(1 << OVS_KEY_ATTR_ND);
-
- *key_len = SW_FLOW_KEY_OFFSET(ipv6.nd);
- nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
- memcpy(&swkey->ipv6.nd.target, nd_key->nd_target,
- sizeof(swkey->ipv6.nd.target));
- memcpy(swkey->ipv6.nd.sll, nd_key->nd_sll, ETH_ALEN);
- memcpy(swkey->ipv6.nd.tll, nd_key->nd_tll, ETH_ALEN);
- }
- break;
- }
-
- return 0;
-}
-
static int parse_flow_nlattrs(const struct nlattr *attr,
const struct nlattr *a[], u64 *attrsp)
{
@@ -986,13 +984,12 @@ static int parse_flow_nlattrs(const struct nlattr *attr,
}
int ipv4_tun_from_nlattr(const struct nlattr *attr,
- struct ovs_key_ipv4_tunnel *tun_key)
+ struct sw_flow_match *match)
{
struct nlattr *a;
int rem;
bool ttl = false;
-
- memset(tun_key, 0, sizeof(*tun_key));
+ u16 tun_flags = 0;
nla_for_each_nested(a, attr, rem) {
int type = nla_type(a);
@@ -1012,37 +1009,39 @@ int ipv4_tun_from_nlattr(const struct nlattr *attr,
switch (type) {
case OVS_TUNNEL_KEY_ATTR_ID:
- tun_key->tun_id = nla_get_be64(a);
- tun_key->tun_flags |= OVS_TNL_F_KEY;
+ SW_FLOW_KEY_PUT(match, tun_key.tun_id, nla_get_be64(a));
+ tun_flags |= OVS_TNL_F_KEY;
break;
case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
- tun_key->ipv4_src = nla_get_be32(a);
+ SW_FLOW_KEY_PUT(match, tun_key.ipv4_src, nla_get_be32(a));
break;
case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
- tun_key->ipv4_dst = nla_get_be32(a);
+ SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst, nla_get_be32(a));
break;
case OVS_TUNNEL_KEY_ATTR_TOS:
- tun_key->ipv4_tos = nla_get_u8(a);
+ SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos, nla_get_u8(a));
break;
case OVS_TUNNEL_KEY_ATTR_TTL:
- tun_key->ipv4_ttl = nla_get_u8(a);
+ SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl, nla_get_u8(a));
ttl = true;
break;
case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
- tun_key->tun_flags |= OVS_TNL_F_DONT_FRAGMENT;
+ tun_flags |= OVS_TNL_F_DONT_FRAGMENT;
break;
case OVS_TUNNEL_KEY_ATTR_CSUM:
- tun_key->tun_flags |= OVS_TNL_F_CSUM;
+ tun_flags |= OVS_TNL_F_CSUM;
break;
default:
return -EINVAL;
}
}
+ SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags);
+
if (rem > 0)
return -EINVAL;
- if (!tun_key->ipv4_dst)
+ if (!match->key->tun_key.ipv4_dst)
return -EINVAL;
if (!ttl)
@@ -1086,180 +1085,267 @@ int ipv4_tun_to_nlattr(struct sk_buff *skb,
/**
* ovs_flow_from_nlattrs - parses Netlink attributes into a flow key.
- * @swkey: receives the extracted flow key.
- * @key_lenp: number of bytes used in @swkey.
+ * @match: receives the extracted flow match information.
* @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
* sequence.
*/
-int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
- const struct nlattr *attr)
+int ovs_match_from_nlattrs(struct sw_flow_match *match,
+ const struct nlattr *attr)
{
const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
- const struct ovs_key_ethernet *eth_key;
- int key_len;
+ const struct nlattr *encap;
u64 attrs;
int err;
-
- memset(swkey, 0, sizeof(struct sw_flow_key));
- key_len = SW_FLOW_KEY_OFFSET(eth);
+ int k;
err = parse_flow_nlattrs(attr, a, &attrs);
if (err)
return err;
- /* Metadata attributes. */
- if (attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
- swkey->phy.priority = nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]);
- attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
- }
- if (attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
- u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
- if (in_port >= DP_MAX_PORTS)
- return -EINVAL;
- swkey->phy.in_port = in_port;
- attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
- } else {
- swkey->phy.in_port = DP_MAX_PORTS;
+ if (attrs & 1ULL << OVS_KEY_ATTR_ENCAP) {
+ encap = a[OVS_KEY_ATTR_ENCAP];
+ attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
+ if (nla_len(encap)) {
+ __be16 tci = 0;
+ __be16 eth_type = 0; /* ETH_P_8021Q */
+
+ if (a[OVS_KEY_ATTR_ETHERTYPE])
+ eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
+ if (a[OVS_KEY_ATTR_VLAN])
+ tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
+
+ if ( (eth_type == htons(ETH_P_8021Q))
+ && (tci & htons(VLAN_TAG_PRESENT)) )
+ err = parse_flow_nlattrs(encap, a, &attrs);
+ else
+ err = -EINVAL;
+
+ if (err)
+ return err;
+ }
}
- if (attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
- uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
+
+ /* Save all attributes specified for prerequisite check */
+ match->key_attrs = attrs;
+
+ /* Set default values if not specified. */
+ if ((attrs & (1 << OVS_KEY_ATTR_IN_PORT)) == 0)
+ SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS);
+
+ if ((attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) == 0)
+ SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2));
+
+ while (attrs) {
+ k = fls64(attrs) -1;
+ attrs &= ~(1ULL << k);
+
+ switch(k) {
+ case OVS_KEY_ATTR_PRIORITY:
+ SW_FLOW_KEY_PUT(match, phy.priority,
+ nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]));
+ break;
+
+ case OVS_KEY_ATTR_IN_PORT:
+ {
+ u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
+
+ if (in_port >= DP_MAX_PORTS)
+ return -EINVAL;
+ SW_FLOW_KEY_PUT(match, phy.in_port, in_port);
+ }
+ break;
+
+ case OVS_KEY_ATTR_SKB_MARK:
+ {
+ uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
+
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20) && !defined(CONFIG_NETFILTER)
- if (mark != 0)
- return -EINVAL;
+ if (mark != 0)
+ return -EINVAL;
#endif
- swkey->phy.skb_mark = mark;
- attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
- }
+ SW_FLOW_KEY_PUT(match, phy.skb_mark, mark);
+ }
+ break;
- if (attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
- err = ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], &swkey->tun_key);
- if (err)
- return err;
+ case OVS_KEY_ATTR_TUNNEL:
+ if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match))
+ return -EINVAL;
- attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
- }
+ break;
- /* Data attributes. */
- if (!(attrs & (1 << OVS_KEY_ATTR_ETHERNET)))
- return -EINVAL;
- attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
+ case OVS_KEY_ATTR_ETHERNET:
+ {
+ const struct ovs_key_ethernet *eth_key;
- eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
- memcpy(swkey->eth.src, eth_key->eth_src, ETH_ALEN);
- memcpy(swkey->eth.dst, eth_key->eth_dst, ETH_ALEN);
+ eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
+ SW_FLOW_KEY_MEMCPY(match, eth.src,
+ eth_key->eth_src, ETH_ALEN);
+ SW_FLOW_KEY_MEMCPY(match, eth.dst,
+ eth_key->eth_dst, ETH_ALEN);
+ }
+ break;
- if (attrs & (1u << OVS_KEY_ATTR_ETHERTYPE) &&
- nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q)) {
- const struct nlattr *encap;
- __be16 tci;
+ case OVS_KEY_ATTR_VLAN:
+ {
+ __be16 tci;
- if (attrs != ((1 << OVS_KEY_ATTR_VLAN) |
- (1 << OVS_KEY_ATTR_ETHERTYPE) |
- (1 << OVS_KEY_ATTR_ENCAP)))
- return -EINVAL;
+ tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
+ SW_FLOW_KEY_PUT(match, eth.tci, tci);
+ }
+ break;
- encap = a[OVS_KEY_ATTR_ENCAP];
- tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
- if (tci & htons(VLAN_TAG_PRESENT)) {
- swkey->eth.tci = tci;
+ case OVS_KEY_ATTR_ETHERTYPE:
+ {
+ __be16 eth_type;
- err = parse_flow_nlattrs(encap, a, &attrs);
- if (err)
- return err;
- } else if (!tci) {
- /* Corner case for truncated 802.1Q header. */
- if (nla_len(encap))
- return -EINVAL;
+ eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
- swkey->eth.type = htons(ETH_P_8021Q);
- *key_lenp = key_len;
- return 0;
- } else {
- return -EINVAL;
- }
- }
+ if (ntohs(eth_type) < ETH_P_802_3_MIN)
+ return -EINVAL;
- if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
- swkey->eth.type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
- if (ntohs(swkey->eth.type) < ETH_P_802_3_MIN)
- return -EINVAL;
- attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
- } else {
- swkey->eth.type = htons(ETH_P_802_2);
- }
+ SW_FLOW_KEY_PUT(match, eth.type, eth_type);
+ }
+ break;
- if (swkey->eth.type == htons(ETH_P_IP)) {
- const struct ovs_key_ipv4 *ipv4_key;
+ case OVS_KEY_ATTR_IPV4:
+ {
+ const struct ovs_key_ipv4 *ipv4_key;
- if (!(attrs & (1 << OVS_KEY_ATTR_IPV4)))
- return -EINVAL;
- attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
+ ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
+ if (ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX)
+ return -EINVAL;
+ SW_FLOW_KEY_PUT(match, ip.proto,
+ ipv4_key->ipv4_proto);
+ SW_FLOW_KEY_PUT(match, ip.tos,
+ ipv4_key->ipv4_tos);
+ SW_FLOW_KEY_PUT(match, ip.ttl,
+ ipv4_key->ipv4_ttl);
+ SW_FLOW_KEY_PUT(match, ip.frag,
+ ipv4_key->ipv4_frag);
+ SW_FLOW_KEY_PUT(match, ipv4.addr.src,
+ ipv4_key->ipv4_src);
+ SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
+ ipv4_key->ipv4_dst);
+ }
+ break;
- key_len = SW_FLOW_KEY_OFFSET(ipv4.addr);
- ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
- if (ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX)
- return -EINVAL;
- swkey->ip.proto = ipv4_key->ipv4_proto;
- swkey->ip.tos = ipv4_key->ipv4_tos;
- swkey->ip.ttl = ipv4_key->ipv4_ttl;
- swkey->ip.frag = ipv4_key->ipv4_frag;
- swkey->ipv4.addr.src = ipv4_key->ipv4_src;
- swkey->ipv4.addr.dst = ipv4_key->ipv4_dst;
-
- if (swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
- err = ipv4_flow_from_nlattrs(swkey, &key_len, a, &attrs);
- if (err)
- return err;
- }
- } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
- const struct ovs_key_ipv6 *ipv6_key;
+ case OVS_KEY_ATTR_IPV6:
+ {
+ const struct ovs_key_ipv6 *ipv6_key;
- if (!(attrs & (1 << OVS_KEY_ATTR_IPV6)))
- return -EINVAL;
- attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
+ ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
+ if (ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX)
+ return -EINVAL;
+ SW_FLOW_KEY_PUT(match, ipv6.label,
+ ipv6_key->ipv6_label);
+ SW_FLOW_KEY_PUT(match, ip.proto,
+ ipv6_key->ipv6_proto);
+ SW_FLOW_KEY_PUT(match, ip.tos,
+ ipv6_key->ipv6_tclass);
+ SW_FLOW_KEY_PUT(match, ip.ttl,
+ ipv6_key->ipv6_hlimit);
+ SW_FLOW_KEY_PUT(match, ip.frag,
+ ipv6_key->ipv6_frag);
+ SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
+ ipv6_key->ipv6_src,
+ sizeof(match->key->ipv6.addr.src));
+ SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
+ ipv6_key->ipv6_dst,
+ sizeof(match->key->ipv6.addr.dst));
- key_len = SW_FLOW_KEY_OFFSET(ipv6.label);
- ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
- if (ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX)
- return -EINVAL;
- swkey->ipv6.label = ipv6_key->ipv6_label;
- swkey->ip.proto = ipv6_key->ipv6_proto;
- swkey->ip.tos = ipv6_key->ipv6_tclass;
- swkey->ip.ttl = ipv6_key->ipv6_hlimit;
- swkey->ip.frag = ipv6_key->ipv6_frag;
- memcpy(&swkey->ipv6.addr.src, ipv6_key->ipv6_src,
- sizeof(swkey->ipv6.addr.src));
- memcpy(&swkey->ipv6.addr.dst, ipv6_key->ipv6_dst,
- sizeof(swkey->ipv6.addr.dst));
-
- if (swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
- err = ipv6_flow_from_nlattrs(swkey, &key_len, a, &attrs);
- if (err)
- return err;
- }
- } else if (swkey->eth.type == htons(ETH_P_ARP) ||
- swkey->eth.type == htons(ETH_P_RARP)) {
- const struct ovs_key_arp *arp_key;
+ }
+ break;
- if (!(attrs & (1 << OVS_KEY_ATTR_ARP)))
- return -EINVAL;
- attrs &= ~(1 << OVS_KEY_ATTR_ARP);
+ case OVS_KEY_ATTR_ARP:
+ {
+ const struct ovs_key_arp *arp_key;
- key_len = SW_FLOW_KEY_OFFSET(ipv4.arp);
- arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
- swkey->ipv4.addr.src = arp_key->arp_sip;
- swkey->ipv4.addr.dst = arp_key->arp_tip;
- if (arp_key->arp_op & htons(0xff00))
+ arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
+ SW_FLOW_KEY_PUT(match, ipv4.addr.src,
+ arp_key->arp_sip);
+ SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
+ arp_key->arp_tip);
+ if (arp_key->arp_op & htons(0xff00))
+ return -EINVAL;
+ SW_FLOW_KEY_PUT(match, ip.proto,
+ ntohs(arp_key->arp_op));
+ SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
+ arp_key->arp_sha,
+ ETH_ALEN);
+ SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
+ arp_key->arp_tha,
+ ETH_ALEN);
+ }
+ break;
+
+ case OVS_KEY_ATTR_TCP:
+ {
+ const struct ovs_key_tcp *tcp_key;
+
+ tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
+ SW_FLOW_KEY_PUT(match, ipv4.tp.src,
+ tcp_key->tcp_src);
+ SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
+ tcp_key->tcp_dst);
+ }
+ break;
+
+ case OVS_KEY_ATTR_UDP:
+ {
+ const struct ovs_key_udp *udp_key;
+
+ udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
+ SW_FLOW_KEY_PUT(match, ipv4.tp.src,
+ udp_key->udp_src);
+ SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
+ udp_key->udp_dst);
+ }
+ break;
+
+ case OVS_KEY_ATTR_ICMP:
+ {
+ const struct ovs_key_icmp *icmp_key;
+
+ icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
+ SW_FLOW_KEY_PUT(match, ipv4.tp.src,
+ htons(icmp_key->icmp_type));
+ SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
+ htons(icmp_key->icmp_code));
+ }
+ break;
+
+ case OVS_KEY_ATTR_ICMPV6:
+ {
+ const struct ovs_key_icmpv6 *icmpv6_key;
+
+ icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
+ SW_FLOW_KEY_PUT(match, ipv6.tp.src,
+ htons(icmpv6_key->icmpv6_type));
+ SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
+ htons(icmpv6_key->icmpv6_code));
+ }
+ break;
+
+ case OVS_KEY_ATTR_ND:
+ {
+ const struct ovs_key_nd *nd_key;
+
+ nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
+ SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
+ nd_key->nd_target,
+ sizeof(match->key->ipv6.nd.target));
+ SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
+ nd_key->nd_sll, ETH_ALEN);
+ SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
+ nd_key->nd_tll, ETH_ALEN);
+ }
+ break;
+
+ default:
return -EINVAL;
- swkey->ip.proto = ntohs(arp_key->arp_op);
- memcpy(swkey->ipv4.arp.sha, arp_key->arp_sha, ETH_ALEN);
- memcpy(swkey->ipv4.arp.tha, arp_key->arp_tha, ETH_ALEN);
+ }
}
- if (attrs)
- return -EINVAL;
- *key_lenp = key_len;
-
return 0;
}
@@ -1279,51 +1365,31 @@ int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
int ovs_flow_metadata_from_nlattrs(struct sw_flow *flow, int key_len, const struct nlattr *attr)
{
struct ovs_key_ipv4_tunnel *tun_key = &flow->key.tun_key;
- const struct nlattr *nla;
- int rem;
+ const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
+ u64 attrs;
+ int err;
+ struct sw_flow_match match;
flow->key.phy.in_port = DP_MAX_PORTS;
flow->key.phy.priority = 0;
flow->key.phy.skb_mark = 0;
memset(tun_key, 0, sizeof(flow->key.tun_key));
- nla_for_each_nested(nla, attr, rem) {
- int type = nla_type(nla);
-
- if (type <= OVS_KEY_ATTR_MAX && ovs_key_lens[type] > 0) {
- int err;
-
- if (nla_len(nla) != ovs_key_lens[type])
- return -EINVAL;
-
- switch (type) {
- case OVS_KEY_ATTR_PRIORITY:
- flow->key.phy.priority = nla_get_u32(nla);
- break;
+ err = parse_flow_nlattrs(attr, a, &attrs);
+ if (err)
+ return -EINVAL;
- case OVS_KEY_ATTR_TUNNEL:
- err = ipv4_tun_from_nlattr(nla, tun_key);
- if (err)
- return err;
- break;
+ attrs &= (1 << OVS_KEY_ATTR_PRIORITY);
+ attrs &= (1 << OVS_KEY_ATTR_TUNNEL);
+ attrs &= (1 << OVS_KEY_ATTR_IN_PORT);
+ attrs &= (1 << OVS_KEY_ATTR_SKB_MARK);
- case OVS_KEY_ATTR_IN_PORT:
- if (nla_get_u32(nla) >= DP_MAX_PORTS)
- return -EINVAL;
- flow->key.phy.in_port = nla_get_u32(nla);
- break;
+ if (attrs)
+ return -EINVAL;
- case OVS_KEY_ATTR_SKB_MARK:
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20) && !defined(CONFIG_NETFILTER)
- if (nla_get_u32(nla) != 0)
- return -EINVAL;
-#endif
- flow->key.phy.skb_mark = nla_get_u32(nla);
- break;
- }
- }
- }
- if (rem)
+ ovs_match_init(&match, &flow->key);
+ err = ovs_match_from_nlattrs(&match, attr);
+ if (err)
return -EINVAL;
flow->hash = ovs_flow_hash(&flow->key,
diff --git a/datapath/flow.h b/datapath/flow.h
index dba66cf..09b1482 100644
--- a/datapath/flow.h
+++ b/datapath/flow.h
@@ -129,6 +129,16 @@ struct sw_flow {
u8 tcp_flags; /* Union of seen TCP flags. */
};
+struct sw_flow_match {
+ struct sw_flow_key *key;
+ u8 n_writes[sizeof(struct sw_flow_key)]; /* Per byte write count. */
+ u64 key_attrs; /* OVS_KEY_ATTR_XXXs used to populate the keys */
+};
+
+void ovs_match_init(struct sw_flow_match *match, struct sw_flow_key *key);
+bool ovs_match_check_prerequisite(const struct sw_flow_match *match);
+int ovs_match_get_len(struct sw_flow_match *match);
+
struct arp_eth_header {
__be16 ar_hrd; /* format of hardware address */
__be16 ar_pro; /* format of protocol address */
@@ -159,7 +169,7 @@ void ovs_flow_used(struct sw_flow *, struct sk_buff *);
u64 ovs_flow_used_time(unsigned long flow_jiffies);
int ovs_flow_to_nlattrs(const struct sw_flow_key *, struct sk_buff *);
-int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
+int ovs_match_from_nlattrs(struct sw_flow_match *match,
const struct nlattr *);
int ovs_flow_metadata_from_nlattrs(struct sw_flow *flow, int key_len,
const struct nlattr *attr);
@@ -176,6 +186,8 @@ struct flow_table {
bool keep_flows;
};
+
+
static inline int ovs_flow_tbl_count(struct flow_table *table)
{
return table->count;
@@ -200,7 +212,7 @@ void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow);
struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *idx);
extern const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1];
int ipv4_tun_from_nlattr(const struct nlattr *attr,
- struct ovs_key_ipv4_tunnel *tun_key);
+ struct sw_flow_match *match);
int ipv4_tun_to_nlattr(struct sk_buff *skb,
const struct ovs_key_ipv4_tunnel *tun_key);
--
1.7.9.5
More information about the dev
mailing list