#include "flow.h"
#include "datapath.h"
-#include <asm/uaccess.h>
+#include <linux/uaccess.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/if_ether.h>
return 0;
}
-static inline bool arphdr_ok(struct sk_buff *skb)
+static bool arphdr_ok(struct sk_buff *skb)
{
return pskb_may_pull(skb, skb_network_offset(skb) +
sizeof(struct arp_eth_header));
}
-static inline int check_iphdr(struct sk_buff *skb)
+static int check_iphdr(struct sk_buff *skb)
{
unsigned int nh_ofs = skb_network_offset(skb);
unsigned int ip_len;
return 0;
}
-static inline bool tcphdr_ok(struct sk_buff *skb)
+static bool tcphdr_ok(struct sk_buff *skb)
{
int th_ofs = skb_transport_offset(skb);
int tcp_len;
return true;
}
-static inline bool udphdr_ok(struct sk_buff *skb)
+static bool udphdr_ok(struct sk_buff *skb)
{
return pskb_may_pull(skb, skb_transport_offset(skb) +
sizeof(struct udphdr));
}
-static inline bool icmphdr_ok(struct sk_buff *skb)
+static bool icmphdr_ok(struct sk_buff *skb)
{
return pskb_may_pull(skb, skb_transport_offset(skb) +
sizeof(struct icmphdr));
}
#define SW_FLOW_KEY_OFFSET(field) \
- offsetof(struct sw_flow_key, field) + \
- FIELD_SIZEOF(struct sw_flow_key, field)
+ (offsetof(struct sw_flow_key, field) + \
+ FIELD_SIZEOF(struct sw_flow_key, field))
+
+/**
+ * skip_exthdr - skip any IPv6 extension headers
+ * @skb: skbuff to parse
+ * @start: offset of first extension header
+ * @nexthdrp: Initially, points to the type of the extension header at @start.
+ * This function updates it to point to the extension header at the final
+ * offset.
+ * @frag: Points to the @frag member in a &struct sw_flow_key. This
+ * function sets an appropriate %OVS_FRAG_TYPE_* value.
+ *
+ * This is based on ipv6_skip_exthdr() but adds the updates to *@frag.
+ *
+ * When there is more than one fragment header, this version reports whether
+ * the final fragment header that it examines is a first fragment.
+ *
+ * Returns the final payload offset, or -1 on error.
+ */
+static int skip_exthdr(const struct sk_buff *skb, int start, u8 *nexthdrp,
+ u8 *frag)
+{
+ u8 nexthdr = *nexthdrp;
+
+ while (ipv6_ext_hdr(nexthdr)) {
+ struct ipv6_opt_hdr _hdr, *hp;
+ int hdrlen;
+
+ if (nexthdr == NEXTHDR_NONE)
+ return -1;
+ hp = skb_header_pointer(skb, start, sizeof(_hdr), &_hdr);
+ if (hp == NULL)
+ return -1;
+ if (nexthdr == NEXTHDR_FRAGMENT) {
+ __be16 _frag_off, *fp;
+ fp = skb_header_pointer(skb,
+ start+offsetof(struct frag_hdr,
+ frag_off),
+ sizeof(_frag_off),
+ &_frag_off);
+ if (fp == NULL)
+ return -1;
+
+ if (ntohs(*fp) & ~0x7) {
+ *frag = OVS_FRAG_TYPE_LATER;
+ break;
+ }
+ *frag = OVS_FRAG_TYPE_FIRST;
+ hdrlen = 8;
+ } else if (nexthdr == NEXTHDR_AUTH)
+ hdrlen = (hp->hdrlen+2)<<2;
+ else
+ hdrlen = ipv6_optlen(hp);
+
+ nexthdr = hp->nexthdr;
+ start += hdrlen;
+ }
+
+ *nexthdrp = nexthdr;
+ return start;
+}
static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key,
int *key_lenp)
uint8_t nexthdr;
int err;
- *key_lenp = SW_FLOW_KEY_OFFSET(ipv6.addr);
+ *key_lenp = SW_FLOW_KEY_OFFSET(ipv6.label);
err = check_header(skb, nh_ofs + sizeof(*nh));
if (unlikely(err))
key->ip.proto = NEXTHDR_NONE;
key->ip.tos = ipv6_get_dsfield(nh) & ~INET_ECN_MASK;
+ key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL);
ipv6_addr_copy(&key->ipv6.addr.src, &nh->saddr);
ipv6_addr_copy(&key->ipv6.addr.dst, &nh->daddr);
- payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr);
+ payload_ofs = skip_exthdr(skb, payload_ofs, &nexthdr, &key->ip.frag);
if (unlikely(payload_ofs < 0))
return -EINVAL;
tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK;
}
- spin_lock_bh(&flow->lock);
+ spin_lock(&flow->lock);
flow->used = jiffies;
flow->packet_count++;
flow->byte_count += skb->len;
flow->tcp_flags |= tcp_flags;
- spin_unlock_bh(&flow->lock);
+ spin_unlock(&flow->lock);
}
struct sw_flow_actions *flow_actions_alloc(const struct nlattr *actions)
static struct flex_array __rcu *alloc_buckets(unsigned int n_buckets)
{
- struct flex_array __rcu * buckets;
+ struct flex_array __rcu *buckets;
int i, err;
buckets = flex_array_alloc(sizeof(struct hlist_head *),
return buckets;
}
-static void free_buckets(struct flex_array * buckets)
+static void free_buckets(struct flex_array *buckets)
{
flex_array_free(buckets);
}
void flow_tbl_deferred_destroy(struct flow_table *table)
{
- if (!table)
- return;
+ if (!table)
+ return;
- call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb);
+ call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb);
}
struct sw_flow *flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *last)
icmp_len -= sizeof(*nd);
offset = 0;
while (icmp_len >= 8) {
- struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd->opt + offset);
+ struct nd_opt_hdr *nd_opt =
+ (struct nd_opt_hdr *)(nd->opt + offset);
int opt_len = nd_opt->nd_opt_len * 8;
if (unlikely(!opt_len || opt_len > icmp_len))
* @in_port: port number on which @skb was received.
* @key: output flow key
* @key_lenp: length of output flow key
- * @is_frag: set to 1 if @skb contains an IPv4 fragment, or to 0 if @skb does
- * not contain an IPv4 packet or if it is not a fragment.
*
* The caller must ensure that skb->len >= ETH_HLEN.
*
* For other key->dl_type values it is left untouched.
*/
int flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key,
- int *key_lenp, bool *is_frag)
+ int *key_lenp)
{
int error = 0;
int key_len = SW_FLOW_KEY_OFFSET(eth);
struct ethhdr *eth;
memset(key, 0, sizeof(*key));
- key->eth.tun_id = OVS_CB(skb)->tun_id;
- key->eth.in_port = in_port;
- *is_frag = false;
+
+ key->phy.priority = skb->priority;
+ key->phy.tun_id = OVS_CB(skb)->tun_id;
+ key->phy.in_port = in_port;
skb_reset_mac_header(skb);
/* Network layer. */
if (key->eth.type == htons(ETH_P_IP)) {
struct iphdr *nh;
+ __be16 offset;
key_len = SW_FLOW_KEY_OFFSET(ipv4.addr);
nh = ip_hdr(skb);
key->ipv4.addr.src = nh->saddr;
key->ipv4.addr.dst = nh->daddr;
- key->ip.tos = nh->tos & ~INET_ECN_MASK;
+
key->ip.proto = nh->protocol;
+ key->ip.tos = nh->tos & ~INET_ECN_MASK;
- /* Transport layer. */
- if ((nh->frag_off & htons(IP_MF | IP_OFFSET)) ||
- (skb_shinfo(skb)->gso_type & SKB_GSO_UDP))
- *is_frag = true;
+ offset = nh->frag_off & htons(IP_OFFSET);
+ if (offset) {
+ key->ip.frag = OVS_FRAG_TYPE_LATER;
+ goto out;
+ }
+ if (nh->frag_off & htons(IP_MF) ||
+ skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
+ key->ip.frag = OVS_FRAG_TYPE_FIRST;
+ /* Transport layer. */
if (key->ip.proto == IPPROTO_TCP) {
key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
- if (!*is_frag && tcphdr_ok(skb)) {
+ if (tcphdr_ok(skb)) {
struct tcphdr *tcp = tcp_hdr(skb);
key->ipv4.tp.src = tcp->source;
key->ipv4.tp.dst = tcp->dest;
}
} else if (key->ip.proto == IPPROTO_UDP) {
key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
- if (!*is_frag && udphdr_ok(skb)) {
+ if (udphdr_ok(skb)) {
struct udphdr *udp = udp_hdr(skb);
key->ipv4.tp.src = udp->source;
key->ipv4.tp.dst = udp->dest;
}
} else if (key->ip.proto == IPPROTO_ICMP) {
key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
- if (!*is_frag && icmphdr_ok(skb)) {
+ if (icmphdr_ok(skb)) {
struct icmphdr *icmp = icmp_hdr(skb);
/* The ICMP type and code fields use the 16-bit
- * transport port fields, so we need to store them
- * in 16-bit network byte order. */
+ * transport port fields, so we need to store
+ * them in 16-bit network byte order. */
key->ipv4.tp.src = htons(icmp->type);
key->ipv4.tp.dst = htons(icmp->code);
}
goto out;
}
+ if (key->ip.frag == OVS_FRAG_TYPE_LATER)
+ goto out;
+ if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
+ key->ip.frag = OVS_FRAG_TYPE_FIRST;
+
/* Transport layer. */
if (key->ip.proto == NEXTHDR_TCP) {
key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
u32 flow_hash(const struct sw_flow_key *key, int key_len)
{
- return jhash2((u32*)key, DIV_ROUND_UP(key_len, sizeof(u32)), hash_seed);
+ return jhash2((u32 *)key, DIV_ROUND_UP(key_len, sizeof(u32)), hash_seed);
}
-struct sw_flow * flow_tbl_lookup(struct flow_table *table,
+struct sw_flow *flow_tbl_lookup(struct flow_table *table,
struct sw_flow_key *key, int key_len)
{
struct sw_flow *flow;
}
/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
-static const u32 key_lens[OVS_KEY_ATTR_MAX + 1] = {
+const u32 ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
+ [OVS_KEY_ATTR_PRIORITY] = 4,
[OVS_KEY_ATTR_TUN_ID] = 8,
[OVS_KEY_ATTR_IN_PORT] = 4,
[OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
* This state machine accepts the following forms, with [] for optional
* elements and | for alternatives:
*
- * [tun_id] [in_port] ethernet [8021q] [ethertype \
+ * [priority] [tun_id] [in_port] ethernet [8021q] [ethertype \
* [IPv4 [TCP|UDP|ICMP] | IPv6 [TCP|UDP|ICMPv6 [ND]] | ARP]]
+ *
+ * except that IPv4 or IPv6 terminates the sequence if its @ipv4_frag or
+ * @ipv6_frag member, respectively, equals %OVS_FRAG_TYPE_LATER.
*/
int flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
const struct nlattr *attr)
int key_len;
memset(swkey, 0, sizeof(*swkey));
- swkey->eth.in_port = USHRT_MAX;
+ swkey->phy.in_port = USHRT_MAX;
swkey->eth.type = htons(ETH_P_802_2);
key_len = SW_FLOW_KEY_OFFSET(eth);
const struct ovs_key_arp *arp_key;
const struct ovs_key_nd *nd_key;
- int type = nla_type(nla);
+ int type = nla_type(nla);
- if (type > OVS_KEY_ATTR_MAX || nla_len(nla) != key_lens[type])
- goto invalid;
+ if (type > OVS_KEY_ATTR_MAX ||
+ nla_len(nla) != ovs_key_lens[type])
+ goto invalid;
#define TRANSITION(PREV_TYPE, TYPE) (((PREV_TYPE) << 16) | (TYPE))
switch (TRANSITION(prev_type, type)) {
+ case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_PRIORITY):
+ swkey->phy.priority = nla_get_u32(nla);
+ break;
+
case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_TUN_ID):
- swkey->eth.tun_id = nla_get_be64(nla);
+ case TRANSITION(OVS_KEY_ATTR_PRIORITY, OVS_KEY_ATTR_TUN_ID):
+ swkey->phy.tun_id = nla_get_be64(nla);
break;
case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_IN_PORT):
+ case TRANSITION(OVS_KEY_ATTR_PRIORITY, OVS_KEY_ATTR_IN_PORT):
case TRANSITION(OVS_KEY_ATTR_TUN_ID, OVS_KEY_ATTR_IN_PORT):
if (nla_get_u32(nla) >= DP_MAX_PORTS)
goto invalid;
- swkey->eth.in_port = nla_get_u32(nla);
+ swkey->phy.in_port = nla_get_u32(nla);
break;
case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_ETHERNET):
+ case TRANSITION(OVS_KEY_ATTR_PRIORITY, OVS_KEY_ATTR_ETHERNET):
case TRANSITION(OVS_KEY_ATTR_TUN_ID, OVS_KEY_ATTR_ETHERNET):
case TRANSITION(OVS_KEY_ATTR_IN_PORT, OVS_KEY_ATTR_ETHERNET):
eth_key = nla_data(nla);
if (swkey->eth.type != htons(ETH_P_IP))
goto invalid;
ipv4_key = nla_data(nla);
+ if (ipv4_key->ipv4_tos & INET_ECN_MASK)
+ goto invalid;
+ if (ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX)
+ goto invalid;
swkey->ip.proto = ipv4_key->ipv4_proto;
swkey->ip.tos = ipv4_key->ipv4_tos;
+ 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.tos & INET_ECN_MASK)
- goto invalid;
break;
case TRANSITION(OVS_KEY_ATTR_ETHERTYPE, OVS_KEY_ATTR_IPV6):
- key_len = SW_FLOW_KEY_OFFSET(ipv6.addr);
+ key_len = SW_FLOW_KEY_OFFSET(ipv6.label);
if (swkey->eth.type != htons(ETH_P_IPV6))
goto invalid;
ipv6_key = nla_data(nla);
+ if (ipv6_key->ipv6_tos & INET_ECN_MASK)
+ goto invalid;
+ if (ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX)
+ goto invalid;
+ swkey->ipv6.label = ipv6_key->ipv6_label;
swkey->ip.proto = ipv6_key->ipv6_proto;
swkey->ip.tos = ipv6_key->ipv6_tos;
+ 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.tos & INET_ECN_MASK)
- goto invalid;
break;
case TRANSITION(OVS_KEY_ATTR_IPV4, OVS_KEY_ATTR_TCP):
case OVS_KEY_ATTR_UNSPEC:
goto invalid;
+ case OVS_KEY_ATTR_PRIORITY:
case OVS_KEY_ATTR_TUN_ID:
case OVS_KEY_ATTR_IN_PORT:
goto invalid;
case OVS_KEY_ATTR_ETHERTYPE:
if (swkey->eth.type == htons(ETH_P_IP) ||
+ swkey->eth.type == htons(ETH_P_IPV6) ||
swkey->eth.type == htons(ETH_P_ARP))
goto invalid;
goto ok;
case OVS_KEY_ATTR_IPV4:
+ if (swkey->ip.frag == OVS_FRAG_TYPE_LATER)
+ goto ok;
if (swkey->ip.proto == IPPROTO_TCP ||
swkey->ip.proto == IPPROTO_UDP ||
swkey->ip.proto == IPPROTO_ICMP)
goto ok;
case OVS_KEY_ATTR_IPV6:
+ if (swkey->ip.frag == OVS_FRAG_TYPE_LATER)
+ goto ok;
if (swkey->ip.proto == IPPROTO_TCP ||
swkey->ip.proto == IPPROTO_UDP ||
swkey->ip.proto == IPPROTO_ICMPV6)
case OVS_KEY_ATTR_ICMPV6:
if (swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
- swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT))
+ swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT) ||
+ swkey->ip.frag == OVS_FRAG_TYPE_LATER)
goto invalid;
goto ok;
case OVS_KEY_ATTR_TCP:
case OVS_KEY_ATTR_UDP:
case OVS_KEY_ATTR_ICMP:
- case OVS_KEY_ATTR_ARP:
case OVS_KEY_ATTR_ND:
+ if (swkey->ip.frag == OVS_FRAG_TYPE_LATER)
+ goto invalid;
+ goto ok;
+
+ case OVS_KEY_ATTR_ARP:
goto ok;
default:
error = -EINVAL;
ok:
- WARN_ON_ONCE(!key_len && !error);
*key_lenp = key_len;
return error;
}
* get the metadata, that is, the parts of the flow key that cannot be
* extracted from the packet itself.
*/
-int flow_metadata_from_nlattrs(u16 *in_port, __be64 *tun_id,
+int flow_metadata_from_nlattrs(u32 *priority, u16 *in_port, __be64 *tun_id,
const struct nlattr *attr)
{
const struct nlattr *nla;
*in_port = USHRT_MAX;
*tun_id = 0;
+ *priority = 0;
prev_type = OVS_KEY_ATTR_UNSPEC;
nla_for_each_nested(nla, attr, rem) {
- int type = nla_type(nla);
+ int type = nla_type(nla);
- if (type > OVS_KEY_ATTR_MAX || nla_len(nla) != key_lens[type])
- return -EINVAL;
+ if (type > OVS_KEY_ATTR_MAX || nla_len(nla) != ovs_key_lens[type])
+ return -EINVAL;
switch (TRANSITION(prev_type, type)) {
+ case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_PRIORITY):
+ *priority = nla_get_u32(nla);
+ break;
+
case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_TUN_ID):
+ case TRANSITION(OVS_KEY_ATTR_PRIORITY, OVS_KEY_ATTR_TUN_ID):
*tun_id = nla_get_be64(nla);
break;
case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_IN_PORT):
+ case TRANSITION(OVS_KEY_ATTR_PRIORITY, OVS_KEY_ATTR_IN_PORT):
case TRANSITION(OVS_KEY_ATTR_TUN_ID, OVS_KEY_ATTR_IN_PORT):
if (nla_get_u32(nla) >= DP_MAX_PORTS)
return -EINVAL;
struct ovs_key_ethernet *eth_key;
struct nlattr *nla;
- /* This is an imperfect sanity-check that FLOW_BUFSIZE doesn't need
- * to be updated, but will at least raise awareness when new
- * datapath key types are added. */
- BUILD_BUG_ON(__OVS_KEY_ATTR_MAX != 14);
+ if (swkey->phy.priority)
+ NLA_PUT_U32(skb, OVS_KEY_ATTR_PRIORITY, swkey->phy.priority);
- if (swkey->eth.tun_id != cpu_to_be64(0))
- NLA_PUT_BE64(skb, OVS_KEY_ATTR_TUN_ID, swkey->eth.tun_id);
+ if (swkey->phy.tun_id != cpu_to_be64(0))
+ NLA_PUT_BE64(skb, OVS_KEY_ATTR_TUN_ID, swkey->phy.tun_id);
- if (swkey->eth.in_port != USHRT_MAX)
- NLA_PUT_U32(skb, OVS_KEY_ATTR_IN_PORT, swkey->eth.in_port);
+ if (swkey->phy.in_port != USHRT_MAX)
+ NLA_PUT_U32(skb, OVS_KEY_ATTR_IN_PORT, swkey->phy.in_port);
nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
if (!nla)
ipv4_key->ipv4_src = swkey->ipv4.addr.src;
ipv4_key->ipv4_dst = swkey->ipv4.addr.dst;
ipv4_key->ipv4_proto = swkey->ip.proto;
- ipv4_key->ipv4_tos = swkey->ip.tos;
+ ipv4_key->ipv4_tos = swkey->ip.tos & ~INET_ECN_MASK;
+ ipv4_key->ipv4_frag = swkey->ip.frag;
} else if (swkey->eth.type == htons(ETH_P_IPV6)) {
struct ovs_key_ipv6 *ipv6_key;
sizeof(ipv6_key->ipv6_src));
memcpy(ipv6_key->ipv6_dst, &swkey->ipv6.addr.dst,
sizeof(ipv6_key->ipv6_dst));
+ ipv6_key->ipv6_label = swkey->ipv6.label;
ipv6_key->ipv6_proto = swkey->ip.proto;
- ipv6_key->ipv6_tos = swkey->ip.tos;
+ ipv6_key->ipv6_tos = swkey->ip.tos & ~INET_ECN_MASK;
+ ipv6_key->ipv6_frag = swkey->ip.frag;
} else if (swkey->eth.type == htons(ETH_P_ARP)) {
struct ovs_key_arp *arp_key;
memcpy(arp_key->arp_tha, swkey->ipv4.arp.tha, ETH_ALEN);
}
- if (swkey->eth.type == htons(ETH_P_IP) ||
- swkey->eth.type == htons(ETH_P_IPV6)) {
+ if ((swkey->eth.type == htons(ETH_P_IP) ||
+ swkey->eth.type == htons(ETH_P_IPV6)) &&
+ swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
if (swkey->ip.proto == IPPROTO_TCP) {
struct ovs_key_tcp *tcp_key;
projects / openvswitch / blobdiff