#include <linux/wait.h>
#include <asm/system.h>
#include <asm/div64.h>
-#include <asm/bug.h>
#include <linux/highmem.h>
#include <linux/netfilter_bridge.h>
#include <linux/netfilter_ipv4.h>
#include <linux/inetdevice.h>
#include <linux/list.h>
+#include <linux/openvswitch.h>
#include <linux/rculist.h>
#include <linux/dmi.h>
#include <net/inet_ecn.h>
#include <net/genetlink.h>
-#include "openvswitch/datapath-protocol.h"
#include "checksum.h"
#include "datapath.h"
#include "actions.h"
#include "vport-internal_dev.h"
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18) || \
- LINUX_VERSION_CODE >= KERNEL_VERSION(3,1,0)
-#error Kernels before 2.6.18 or after 3.0 are not supported by this version of Open vSwitch.
+ LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0)
+#error Kernels before 2.6.18 or after 3.1 are not supported by this version of Open vSwitch.
#endif
int (*dp_ioctl_hook)(struct net_device *dev, struct ifreq *rq, int cmd);
* etc.) are protected by RTNL.
*
* Writes to other state (flow table modifications, set miscellaneous datapath
- * parameters such as drop frags, etc.) are protected by genl_mutex. The RTNL
- * lock nests inside genl_mutex.
+ * parameters, etc.) are protected by genl_mutex. The RTNL lock nests inside
+ * genl_mutex.
*
* Reads are protected by RCU.
*
static LIST_HEAD(dps);
static struct vport *new_vport(const struct vport_parms *);
-static int queue_userspace_packets(struct datapath *, struct sk_buff *,
- const struct dp_upcall_info *);
+static int queue_gso_packets(int dp_ifindex, struct sk_buff *,
+ const struct dp_upcall_info *);
+static int queue_userspace_packet(int dp_ifindex, struct sk_buff *,
+ const struct dp_upcall_info *);
/* Must be called with rcu_read_lock, genl_mutex, or RTNL lock. */
struct datapath *get_dp(int dp_ifindex)
return vport_get_name(rcu_dereference_rtnl(dp->ports[OVSP_LOCAL]));
}
-static inline size_t br_nlmsg_size(void)
+static int get_dpifindex(struct datapath *dp)
+{
+ struct vport *local;
+ int ifindex;
+
+ rcu_read_lock();
+
+ local = get_vport_protected(dp, OVSP_LOCAL);
+ if (local)
+ ifindex = vport_get_ifindex(local);
+ else
+ ifindex = 0;
+
+ rcu_read_unlock();
+
+ return ifindex;
+}
+
+static size_t br_nlmsg_size(void)
{
return NLMSG_ALIGN(sizeof(struct ifinfomsg))
+ nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
hdr->ifi_change = 0;
NLA_PUT_STRING(skb, IFLA_IFNAME, vport_get_name(port));
- NLA_PUT_U32(skb, IFLA_MASTER,
- vport_get_ifindex(get_vport_protected(dp, OVSP_LOCAL)));
+ NLA_PUT_U32(skb, IFLA_MASTER, get_dpifindex(dp));
NLA_PUT_U32(skb, IFLA_MTU, vport_get_mtu(port));
#ifdef IFLA_OPERSTATE
NLA_PUT_U8(skb, IFLA_OPERSTATE,
struct datapath *dp = p->dp;
struct sw_flow *flow;
struct dp_stats_percpu *stats;
- int stats_counter_off;
+ u64 *stats_counter;
int error;
+ stats = per_cpu_ptr(dp->stats_percpu, smp_processor_id());
OVS_CB(skb)->vport = p;
if (!OVS_CB(skb)->flow) {
struct sw_flow_key key;
int key_len;
- bool is_frag;
/* Extract flow from 'skb' into 'key'. */
- error = flow_extract(skb, p->port_no, &key, &key_len, &is_frag);
+ error = flow_extract(skb, p->port_no, &key, &key_len);
if (unlikely(error)) {
kfree_skb(skb);
return;
}
- if (is_frag && dp->drop_frags) {
- consume_skb(skb);
- stats_counter_off = offsetof(struct dp_stats_percpu, n_frags);
- goto out;
- }
-
/* Look up flow. */
- flow = flow_tbl_lookup(rcu_dereference(dp->table), &key, key_len);
+ flow = flow_tbl_lookup(rcu_dereference(dp->table),
+ &key, key_len);
if (unlikely(!flow)) {
struct dp_upcall_info upcall;
upcall.cmd = OVS_PACKET_CMD_MISS;
upcall.key = &key;
- upcall.userdata = 0;
- upcall.sample_pool = 0;
- upcall.actions = NULL;
- upcall.actions_len = 0;
+ upcall.userdata = NULL;
+ upcall.pid = p->upcall_pid;
dp_upcall(dp, skb, &upcall);
- stats_counter_off = offsetof(struct dp_stats_percpu, n_missed);
+ consume_skb(skb);
+ stats_counter = &stats->n_missed;
goto out;
}
OVS_CB(skb)->flow = flow;
}
- stats_counter_off = offsetof(struct dp_stats_percpu, n_hit);
+ stats_counter = &stats->n_hit;
flow_used(OVS_CB(skb)->flow, skb);
execute_actions(dp, skb);
out:
/* Update datapath statistics. */
- local_bh_disable();
- stats = per_cpu_ptr(dp->stats_percpu, smp_processor_id());
write_seqcount_begin(&stats->seqlock);
- (*(u64 *)((u8 *)stats + stats_counter_off))++;
+ (*stats_counter)++;
write_seqcount_end(&stats->seqlock);
-
- local_bh_enable();
}
static void copy_and_csum_skb(struct sk_buff *skb, void *to)
.id = GENL_ID_GENERATE,
.hdrsize = sizeof(struct ovs_header),
.name = OVS_PACKET_FAMILY,
- .version = 1,
+ .version = OVS_PACKET_VERSION,
.maxattr = OVS_PACKET_ATTR_MAX
};
-/* Generic Netlink multicast groups for upcalls.
- *
- * We really want three unique multicast groups per datapath, but we can't even
- * get one, because genl_register_mc_group() takes genl_lock, which is also
- * held during Generic Netlink message processing, so trying to acquire
- * multicast groups during OVS_DP_NEW processing deadlocks. Instead, we
- * preallocate a few groups and use them round-robin for datapaths. Collision
- * isn't fatal--multicast listeners should check that the family is the one
- * that they want and discard others--but it wastes time and memory to receive
- * unwanted messages.
- */
-#define PACKET_N_MC_GROUPS 16
-static struct genl_multicast_group packet_mc_groups[PACKET_N_MC_GROUPS];
-
-static u32 packet_mc_group(struct datapath *dp, u8 cmd)
-{
- u32 idx;
- BUILD_BUG_ON_NOT_POWER_OF_2(PACKET_N_MC_GROUPS);
-
- idx = jhash_2words(dp->dp_ifindex, cmd, 0) & (PACKET_N_MC_GROUPS - 1);
- return packet_mc_groups[idx].id;
-}
-
-static int packet_register_mc_groups(void)
-{
- int i;
-
- for (i = 0; i < PACKET_N_MC_GROUPS; i++) {
- struct genl_multicast_group *group = &packet_mc_groups[i];
- int error;
-
- sprintf(group->name, "packet%d", i);
- error = genl_register_mc_group(&dp_packet_genl_family, group);
- if (error)
- return error;
- }
- return 0;
-}
-
-int dp_upcall(struct datapath *dp, struct sk_buff *skb, const struct dp_upcall_info *upcall_info)
+int dp_upcall(struct datapath *dp, struct sk_buff *skb,
+ const struct dp_upcall_info *upcall_info)
{
struct dp_stats_percpu *stats;
+ int dp_ifindex;
int err;
- WARN_ON_ONCE(skb_shared(skb));
-
- forward_ip_summed(skb, true);
+ if (upcall_info->pid == 0) {
+ err = -ENOTCONN;
+ goto err;
+ }
- /* Break apart GSO packets into their component pieces. Otherwise
- * userspace may try to stuff a 64kB packet into a 1500-byte MTU. */
- if (skb_is_gso(skb)) {
- struct sk_buff *nskb = skb_gso_segment(skb, NETIF_F_SG | NETIF_F_HW_CSUM);
-
- if (IS_ERR(nskb)) {
- kfree_skb(skb);
- err = PTR_ERR(nskb);
- goto err;
- }
- consume_skb(skb);
- skb = nskb;
+ dp_ifindex = get_dpifindex(dp);
+ if (!dp_ifindex) {
+ err = -ENODEV;
+ goto err;
}
- err = queue_userspace_packets(dp, skb, upcall_info);
+ forward_ip_summed(skb, true);
+
+ if (!skb_is_gso(skb))
+ err = queue_userspace_packet(dp_ifindex, skb, upcall_info);
+ else
+ err = queue_gso_packets(dp_ifindex, skb, upcall_info);
if (err)
goto err;
return 0;
err:
- local_bh_disable();
stats = per_cpu_ptr(dp->stats_percpu, smp_processor_id());
write_seqcount_begin(&stats->seqlock);
stats->n_lost++;
write_seqcount_end(&stats->seqlock);
- local_bh_enable();
-
return err;
}
-/* Send each packet in the 'skb' list to userspace for 'dp' as directed by
- * 'upcall_info'. There will be only one packet unless we broke up a GSO
- * packet.
- */
-static int queue_userspace_packets(struct datapath *dp, struct sk_buff *skb,
- const struct dp_upcall_info *upcall_info)
+static int queue_gso_packets(int dp_ifindex, struct sk_buff *skb,
+ const struct dp_upcall_info *upcall_info)
{
- u32 group = packet_mc_group(dp, upcall_info->cmd);
- struct sk_buff *nskb;
+ struct dp_upcall_info later_info;
+ struct sw_flow_key later_key;
+ struct sk_buff *segs, *nskb;
int err;
+ segs = skb_gso_segment(skb, NETIF_F_SG | NETIF_F_HW_CSUM);
+ if (IS_ERR(skb))
+ return PTR_ERR(skb);
+
+ /* Queue all of the segments. */
+ skb = segs;
do {
- struct ovs_header *upcall;
- struct sk_buff *user_skb; /* to be queued to userspace */
- struct nlattr *nla;
- unsigned int len;
+ err = queue_userspace_packet(dp_ifindex, skb, upcall_info);
+ if (err)
+ break;
- nskb = skb->next;
- skb->next = NULL;
+ if (skb == segs && skb_shinfo(skb)->gso_type & SKB_GSO_UDP) {
+ /* The initial flow key extracted by flow_extract() in
+ * this case is for a first fragment, so we need to
+ * properly mark later fragments.
+ */
+ later_key = *upcall_info->key;
+ later_key.ip.tos_frag &= ~OVS_FRAG_TYPE_MASK;
+ later_key.ip.tos_frag |= OVS_FRAG_TYPE_LATER;
+
+ later_info = *upcall_info;
+ later_info.key = &later_key;
+ upcall_info = &later_info;
+ }
+ } while ((skb = skb->next));
- err = vlan_deaccel_tag(skb);
- if (unlikely(err))
- goto err_kfree_skbs;
+ /* Free all of the segments. */
+ skb = segs;
+ do {
+ nskb = skb->next;
+ if (err)
+ kfree_skb(skb);
+ else
+ consume_skb(skb);
+ } while ((skb = nskb));
+ return err;
+}
- if (nla_attr_size(skb->len) > USHRT_MAX) {
- err = -EFBIG;
- goto err_kfree_skbs;
- }
+static int queue_userspace_packet(int dp_ifindex, struct sk_buff *skb,
+ const struct dp_upcall_info *upcall_info)
+{
+ struct ovs_header *upcall;
+ struct sk_buff *user_skb; /* to be queued to userspace */
+ struct nlattr *nla;
+ unsigned int len;
+ int err;
- len = sizeof(struct ovs_header);
- len += nla_total_size(skb->len);
- len += nla_total_size(FLOW_BUFSIZE);
- if (upcall_info->userdata)
- len += nla_total_size(8);
- if (upcall_info->sample_pool)
- len += nla_total_size(4);
- if (upcall_info->actions_len)
- len += nla_total_size(upcall_info->actions_len);
-
- user_skb = genlmsg_new(len, GFP_ATOMIC);
- if (!user_skb) {
- netlink_set_err(INIT_NET_GENL_SOCK, 0, group, -ENOBUFS);
- err = -ENOMEM;
- goto err_kfree_skbs;
- }
+ err = vlan_deaccel_tag(skb);
+ if (unlikely(err))
+ return err;
- upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family, 0, upcall_info->cmd);
- upcall->dp_ifindex = dp->dp_ifindex;
+ if (nla_attr_size(skb->len) > USHRT_MAX)
+ return -EFBIG;
- nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY);
- flow_to_nlattrs(upcall_info->key, user_skb);
- nla_nest_end(user_skb, nla);
+ len = sizeof(struct ovs_header);
+ len += nla_total_size(skb->len);
+ len += nla_total_size(FLOW_BUFSIZE);
+ if (upcall_info->cmd == OVS_PACKET_CMD_ACTION)
+ len += nla_total_size(8);
- if (upcall_info->userdata)
- nla_put_u64(user_skb, OVS_PACKET_ATTR_USERDATA, upcall_info->userdata);
- if (upcall_info->sample_pool)
- nla_put_u32(user_skb, OVS_PACKET_ATTR_SAMPLE_POOL, upcall_info->sample_pool);
- if (upcall_info->actions_len) {
- const struct nlattr *actions = upcall_info->actions;
- u32 actions_len = upcall_info->actions_len;
+ user_skb = genlmsg_new(len, GFP_ATOMIC);
+ if (!user_skb)
+ return -ENOMEM;
- nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_ACTIONS);
- memcpy(__skb_put(user_skb, actions_len), actions, actions_len);
- nla_nest_end(user_skb, nla);
- }
+ upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family,
+ 0, upcall_info->cmd);
+ upcall->dp_ifindex = dp_ifindex;
- nla = __nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, skb->len);
- if (skb->ip_summed == CHECKSUM_PARTIAL)
- copy_and_csum_skb(skb, nla_data(nla));
- else
- skb_copy_bits(skb, 0, nla_data(nla), skb->len);
+ nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY);
+ flow_to_nlattrs(upcall_info->key, user_skb);
+ nla_nest_end(user_skb, nla);
- err = genlmsg_multicast(user_skb, 0, group, GFP_ATOMIC);
- if (err)
- goto err_kfree_skbs;
+ if (upcall_info->userdata)
+ nla_put_u64(user_skb, OVS_PACKET_ATTR_USERDATA,
+ nla_get_u64(upcall_info->userdata));
- consume_skb(skb);
- skb = nskb;
- } while (skb);
- return 0;
+ nla = __nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, skb->len);
+ if (skb->ip_summed == CHECKSUM_PARTIAL)
+ copy_and_csum_skb(skb, nla_data(nla));
+ else
+ skb_copy_bits(skb, 0, nla_data(nla), skb->len);
-err_kfree_skbs:
- kfree_skb(skb);
- while ((skb = nskb) != NULL) {
- nskb = skb->next;
- kfree_skb(skb);
- }
- return err;
+ return genlmsg_unicast(&init_net, user_skb, upcall_info->pid);
}
/* Called with genl_mutex. */
return 0;
}
-static int validate_actions(const struct nlattr *attr)
+static int validate_actions(const struct nlattr *attr,
+ const struct sw_flow_key *key, int depth);
+
+static int validate_sample(const struct nlattr *attr,
+ const struct sw_flow_key *key, int depth)
{
+ const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
+ const struct nlattr *probability, *actions;
const struct nlattr *a;
int rem;
+ memset(attrs, 0, sizeof(attrs));
+ nla_for_each_nested(a, attr, rem) {
+ int type = nla_type(a);
+ if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
+ return -EINVAL;
+ attrs[type] = a;
+ }
+ if (rem)
+ return -EINVAL;
+
+ probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
+ if (!probability || nla_len(probability) != sizeof(u32))
+ return -EINVAL;
+
+ actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
+ if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
+ return -EINVAL;
+ return validate_actions(actions, key, depth + 1);
+}
+
+static int validate_action_key(const struct nlattr *a,
+ const struct sw_flow_key *flow_key)
+{
+ int act_type = nla_type(a);
+ const struct nlattr *ovs_key = nla_data(a);
+ int key_type = nla_type(ovs_key);
+
+ /* There can be only one key in a action */
+ if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
+ return -EINVAL;
+
+ if (key_type > OVS_KEY_ATTR_MAX ||
+ nla_len(ovs_key) != ovs_key_lens[key_type])
+ return -EINVAL;
+
+#define ACTION(act, key) (((act) << 8) | (key))
+
+ switch (ACTION(act_type, key_type)) {
+ const struct ovs_key_ipv4 *ipv4_key;
+ const struct ovs_key_8021q *q_key;
+
+ case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_PRIORITY):
+ case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_TUN_ID):
+ case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_ETHERNET):
+ break;
+
+ case ACTION(OVS_ACTION_ATTR_PUSH, OVS_KEY_ATTR_8021Q):
+ q_key = nla_data(ovs_key);
+ if (q_key->q_tpid != htons(ETH_P_8021Q))
+ return -EINVAL;
+
+ if (q_key->q_tci & htons(VLAN_TAG_PRESENT))
+ return -EINVAL;
+ break;
+
+ case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_IPV4):
+ if (flow_key->eth.type != htons(ETH_P_IP))
+ return -EINVAL;
+
+ if (!flow_key->ipv4.addr.src || !flow_key->ipv4.addr.dst)
+ return -EINVAL;
+
+ ipv4_key = nla_data(ovs_key);
+ if (ipv4_key->ipv4_proto != flow_key->ip.proto)
+ return -EINVAL;
+
+ if (ipv4_key->ipv4_tos & INET_ECN_MASK)
+ return -EINVAL;
+
+ if (ipv4_key->ipv4_frag !=
+ (flow_key->ip.tos_frag & OVS_FRAG_TYPE_MASK))
+ return -EINVAL;
+
+ break;
+
+ case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_TCP):
+ if (flow_key->ip.proto != IPPROTO_TCP)
+ return -EINVAL;
+
+ if (!flow_key->ipv4.tp.src || !flow_key->ipv4.tp.dst)
+ return -EINVAL;
+
+ break;
+
+ case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_UDP):
+ if (flow_key->ip.proto != IPPROTO_UDP)
+ return -EINVAL;
+
+ if (!flow_key->ipv4.tp.src || !flow_key->ipv4.tp.dst)
+ return -EINVAL;
+ break;
+
+ default:
+ return -EINVAL;
+ }
+#undef ACTION
+ return 0;
+}
+
+static int validate_userspace(const struct nlattr *attr)
+{
+ static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
+ [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
+ [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_U64 },
+ };
+ struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
+ int error;
+
+ error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
+ attr, userspace_policy);
+ if (error)
+ return error;
+
+ if (!a[OVS_USERSPACE_ATTR_PID] ||
+ !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
+ return -EINVAL;
+
+ return 0;
+}
+
+static int validate_actions(const struct nlattr *attr,
+ const struct sw_flow_key *key, int depth)
+{
+ const struct nlattr *a;
+ int rem, err;
+
+ if (depth >= SAMPLE_ACTION_DEPTH)
+ return -EOVERFLOW;
+
nla_for_each_nested(a, attr, rem) {
+ /* Expected argument lengths, (u32)-1 for variable length. */
static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
[OVS_ACTION_ATTR_OUTPUT] = 4,
- [OVS_ACTION_ATTR_USERSPACE] = 8,
- [OVS_ACTION_ATTR_PUSH_VLAN] = 2,
- [OVS_ACTION_ATTR_POP_VLAN] = 0,
- [OVS_ACTION_ATTR_SET_DL_SRC] = ETH_ALEN,
- [OVS_ACTION_ATTR_SET_DL_DST] = ETH_ALEN,
- [OVS_ACTION_ATTR_SET_NW_SRC] = 4,
- [OVS_ACTION_ATTR_SET_NW_DST] = 4,
- [OVS_ACTION_ATTR_SET_NW_TOS] = 1,
- [OVS_ACTION_ATTR_SET_TP_SRC] = 2,
- [OVS_ACTION_ATTR_SET_TP_DST] = 2,
- [OVS_ACTION_ATTR_SET_TUNNEL] = 8,
- [OVS_ACTION_ATTR_SET_PRIORITY] = 4,
- [OVS_ACTION_ATTR_POP_PRIORITY] = 0,
+ [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
+ [OVS_ACTION_ATTR_PUSH] = (u32)-1,
+ [OVS_ACTION_ATTR_POP] = 2,
+ [OVS_ACTION_ATTR_SET] = (u32)-1,
+ [OVS_ACTION_ATTR_SAMPLE] = (u32)-1
};
int type = nla_type(a);
- if (type > OVS_ACTION_ATTR_MAX || nla_len(a) != action_lens[type])
+ if (type > OVS_ACTION_ATTR_MAX ||
+ (action_lens[type] != nla_len(a) &&
+ action_lens[type] != (u32)-1))
return -EINVAL;
switch (type) {
return -EINVAL;
case OVS_ACTION_ATTR_USERSPACE:
- case OVS_ACTION_ATTR_POP_VLAN:
- case OVS_ACTION_ATTR_SET_DL_SRC:
- case OVS_ACTION_ATTR_SET_DL_DST:
- case OVS_ACTION_ATTR_SET_NW_SRC:
- case OVS_ACTION_ATTR_SET_NW_DST:
- case OVS_ACTION_ATTR_SET_TP_SRC:
- case OVS_ACTION_ATTR_SET_TP_DST:
- case OVS_ACTION_ATTR_SET_TUNNEL:
- case OVS_ACTION_ATTR_SET_PRIORITY:
- case OVS_ACTION_ATTR_POP_PRIORITY:
- /* No validation needed. */
+ err = validate_userspace(a);
+ if (err)
+ return err;
break;
case OVS_ACTION_ATTR_OUTPUT:
return -EINVAL;
break;
- case OVS_ACTION_ATTR_PUSH_VLAN:
- if (nla_get_be16(a) & htons(VLAN_CFI_MASK))
+
+ case OVS_ACTION_ATTR_POP:
+ if (nla_get_u16(a) != OVS_KEY_ATTR_8021Q)
return -EINVAL;
break;
- case OVS_ACTION_ATTR_SET_NW_TOS:
- if (nla_get_u8(a) & INET_ECN_MASK)
- return -EINVAL;
+ case OVS_ACTION_ATTR_SET:
+ case OVS_ACTION_ATTR_PUSH:
+ err = validate_action_key(a, key);
+ if (err)
+ return err;
+ break;
+
+ case OVS_ACTION_ATTR_SAMPLE:
+ err = validate_sample(a, key, depth);
+ if (err)
+ return err;
break;
default:
- return -EOPNOTSUPP;
+ return -EINVAL;
}
}
return 0;
}
+
static void clear_stats(struct sw_flow *flow)
{
flow->used = 0;
struct sw_flow *flow;
struct datapath *dp;
struct ethhdr *eth;
- bool is_frag;
int len;
int err;
int key_len;
nla_len(a[OVS_PACKET_ATTR_PACKET]) < ETH_HLEN)
goto err;
- err = validate_actions(a[OVS_PACKET_ATTR_ACTIONS]);
- if (err)
- goto err;
-
len = nla_len(a[OVS_PACKET_ATTR_PACKET]);
packet = __dev_alloc_skb(NET_IP_ALIGN + len, GFP_KERNEL);
err = -ENOMEM;
if (IS_ERR(flow))
goto err_kfree_skb;
- err = flow_extract(packet, -1, &flow->key, &key_len, &is_frag);
+ err = flow_extract(packet, -1, &flow->key, &key_len);
if (err)
goto err_flow_put;
- err = flow_metadata_from_nlattrs(&flow->key.eth.in_port,
- &flow->key.eth.tun_id,
+ err = flow_metadata_from_nlattrs(&flow->key.phy.priority,
+ &flow->key.phy.in_port,
+ &flow->key.phy.tun_id,
a[OVS_PACKET_ATTR_KEY]);
if (err)
goto err_flow_put;
+ err = validate_actions(a[OVS_PACKET_ATTR_ACTIONS], &flow->key, 0);
+ if (err)
+ goto err_flow_put;
+
flow->hash = flow_hash(&flow->key, key_len);
acts = flow_actions_alloc(a[OVS_PACKET_ATTR_ACTIONS]);
rcu_assign_pointer(flow->sf_acts, acts);
OVS_CB(packet)->flow = flow;
+ packet->priority = flow->key.phy.priority;
rcu_read_lock();
dp = get_dp(ovs_header->dp_ifindex);
if (!dp)
goto err_unlock;
- if (flow->key.eth.in_port < DP_MAX_PORTS)
+ if (flow->key.phy.in_port < DP_MAX_PORTS)
OVS_CB(packet)->vport = get_vport_protected(dp,
- flow->key.eth.in_port);
+ flow->key.phy.in_port);
+ local_bh_disable();
err = execute_actions(dp, packet);
+ local_bh_enable();
rcu_read_unlock();
flow_put(flow);
stats->n_flows = flow_tbl_count(table);
- stats->n_frags = stats->n_hit = stats->n_missed = stats->n_lost = 0;
+ stats->n_hit = stats->n_missed = stats->n_lost = 0;
for_each_possible_cpu(i) {
const struct dp_stats_percpu *percpu_stats;
struct dp_stats_percpu local_stats;
local_stats = *percpu_stats;
} while (read_seqcount_retry(&percpu_stats->seqlock, seqcount));
- stats->n_frags += local_stats.n_frags;
stats->n_hit += local_stats.n_hit;
stats->n_missed += local_stats.n_missed;
stats->n_lost += local_stats.n_lost;
.id = GENL_ID_GENERATE,
.hdrsize = sizeof(struct ovs_header),
.name = OVS_FLOW_FAMILY,
- .version = 1,
+ .version = OVS_FLOW_VERSION,
.maxattr = OVS_FLOW_ATTR_MAX
};
/* Called with genl_lock. */
static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
- struct sk_buff *skb, u32 pid, u32 seq, u32 flags, u8 cmd)
+ struct sk_buff *skb, u32 pid,
+ u32 seq, u32 flags, u8 cmd)
{
const int skb_orig_len = skb->len;
const struct sw_flow_actions *sf_acts;
if (!ovs_header)
return -EMSGSIZE;
- ovs_header->dp_ifindex = dp->dp_ifindex;
+ ovs_header->dp_ifindex = get_dpifindex(dp);
nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY);
if (!nla)
NLA_PUT_U64(skb, OVS_FLOW_ATTR_USED, flow_used_time(used));
if (stats.n_packets)
- NLA_PUT(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats);
+ NLA_PUT(skb, OVS_FLOW_ATTR_STATS,
+ sizeof(struct ovs_flow_stats), &stats);
if (tcp_flags)
NLA_PUT_U8(skb, OVS_FLOW_ATTR_TCP_FLAGS, tcp_flags);
sf_acts = rcu_dereference_protected(flow->sf_acts,
lockdep_genl_is_held());
- len = nla_total_size(FLOW_BUFSIZE); /* OVS_FLOW_ATTR_KEY */
- len += nla_total_size(sf_acts->actions_len); /* OVS_FLOW_ATTR_ACTIONS */
- len += nla_total_size(sizeof(struct ovs_flow_stats)); /* OVS_FLOW_ATTR_STATS */
- len += nla_total_size(1); /* OVS_FLOW_ATTR_TCP_FLAGS */
- len += nla_total_size(8); /* OVS_FLOW_ATTR_USED */
- return genlmsg_new(NLMSG_ALIGN(sizeof(struct ovs_header)) + len, GFP_KERNEL);
+ /* OVS_FLOW_ATTR_KEY */
+ len = nla_total_size(FLOW_BUFSIZE);
+ /* OVS_FLOW_ATTR_ACTIONS */
+ len += nla_total_size(sf_acts->actions_len);
+ /* OVS_FLOW_ATTR_STATS */
+ len += nla_total_size(sizeof(struct ovs_flow_stats));
+ /* OVS_FLOW_ATTR_TCP_FLAGS */
+ len += nla_total_size(1);
+ /* OVS_FLOW_ATTR_USED */
+ len += nla_total_size(8);
+
+ len += NLMSG_ALIGN(sizeof(struct ovs_header));
+
+ return genlmsg_new(len, GFP_KERNEL);
}
-static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow, struct datapath *dp,
+static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow,
+ struct datapath *dp,
u32 pid, u32 seq, u8 cmd)
{
struct sk_buff *skb;
/* Validate actions. */
if (a[OVS_FLOW_ATTR_ACTIONS]) {
- error = validate_actions(a[OVS_FLOW_ATTR_ACTIONS]);
+ error = validate_actions(a[OVS_FLOW_ATTR_ACTIONS], &key, 0);
if (error)
goto error;
} else if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW) {
flow_tbl_insert(table, flow);
reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid,
- info->snd_seq, OVS_FLOW_CMD_NEW);
+ info->snd_seq,
+ OVS_FLOW_CMD_NEW);
} else {
/* We found a matching flow. */
struct sw_flow_actions *old_acts;
+ struct nlattr *acts_attrs;
/* Bail out if we're not allowed to modify an existing flow.
* We accept NLM_F_CREATE in place of the intended NLM_F_EXCL
/* Update actions. */
old_acts = rcu_dereference_protected(flow->sf_acts,
lockdep_genl_is_held());
- if (a[OVS_FLOW_ATTR_ACTIONS] &&
- (old_acts->actions_len != nla_len(a[OVS_FLOW_ATTR_ACTIONS]) ||
- memcmp(old_acts->actions, nla_data(a[OVS_FLOW_ATTR_ACTIONS]),
- old_acts->actions_len))) {
+ acts_attrs = a[OVS_FLOW_ATTR_ACTIONS];
+ if (acts_attrs &&
+ (old_acts->actions_len != nla_len(acts_attrs) ||
+ memcmp(old_acts->actions, nla_data(acts_attrs),
+ old_acts->actions_len))) {
struct sw_flow_actions *new_acts;
- new_acts = flow_actions_alloc(a[OVS_FLOW_ATTR_ACTIONS]);
+ new_acts = flow_actions_alloc(acts_attrs);
error = PTR_ERR(new_acts);
if (IS_ERR(new_acts))
goto error;
}
reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid,
- info->snd_seq, OVS_FLOW_CMD_NEW);
+ info->snd_seq, OVS_FLOW_CMD_NEW);
/* Clear stats. */
if (a[OVS_FLOW_ATTR_CLEAR]) {
if (!IS_ERR(reply))
genl_notify(reply, genl_info_net(info), info->snd_pid,
- dp_flow_multicast_group.id, info->nlhdr, GFP_KERNEL);
+ dp_flow_multicast_group.id, info->nlhdr, GFP_KERNEL);
else
netlink_set_err(INIT_NET_GENL_SOCK, 0,
dp_flow_multicast_group.id, PTR_ERR(reply));
if (!flow)
return -ENOENT;
- reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid, info->snd_seq, OVS_FLOW_CMD_NEW);
+ reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid,
+ info->snd_seq, OVS_FLOW_CMD_NEW);
if (IS_ERR(reply))
return PTR_ERR(reply);
dp = get_dp(ovs_header->dp_ifindex);
if (!dp)
- return -ENODEV;
+ return -ENODEV;
table = get_table_protected(dp);
flow = flow_tbl_lookup(table, &key, key_len);
if (!flow)
break;
- if (ovs_flow_cmd_fill_info(flow, dp, skb, NETLINK_CB(cb->skb).pid,
+ if (ovs_flow_cmd_fill_info(flow, dp, skb,
+ NETLINK_CB(cb->skb).pid,
cb->nlh->nlmsg_seq, NLM_F_MULTI,
OVS_FLOW_CMD_NEW) < 0)
break;
#ifdef HAVE_NLA_NUL_STRING
[OVS_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
#endif
- [OVS_DP_ATTR_IPV4_FRAGS] = { .type = NLA_U32 },
- [OVS_DP_ATTR_SAMPLING] = { .type = NLA_U32 },
+ [OVS_DP_ATTR_UPCALL_PID] = { .type = NLA_U32 },
};
static struct genl_family dp_datapath_genl_family = {
.id = GENL_ID_GENERATE,
.hdrsize = sizeof(struct ovs_header),
.name = OVS_DATAPATH_FAMILY,
- .version = 1,
+ .version = OVS_DATAPATH_VERSION,
.maxattr = OVS_DP_ATTR_MAX
};
if (!ovs_header)
goto error;
- ovs_header->dp_ifindex = dp->dp_ifindex;
+ ovs_header->dp_ifindex = get_dpifindex(dp);
rcu_read_lock();
err = nla_put_string(skb, OVS_DP_ATTR_NAME, dp_name(dp));
goto nla_put_failure;
get_dp_stats(dp, nla_data(nla));
- NLA_PUT_U32(skb, OVS_DP_ATTR_IPV4_FRAGS,
- dp->drop_frags ? OVS_DP_FRAG_DROP : OVS_DP_FRAG_ZERO);
-
- if (dp->sflow_probability)
- NLA_PUT_U32(skb, OVS_DP_ATTR_SAMPLING, dp->sflow_probability);
-
- nla = nla_nest_start(skb, OVS_DP_ATTR_MCGROUPS);
- if (!nla)
- goto nla_put_failure;
- NLA_PUT_U32(skb, OVS_PACKET_CMD_MISS, packet_mc_group(dp, OVS_PACKET_CMD_MISS));
- NLA_PUT_U32(skb, OVS_PACKET_CMD_ACTION, packet_mc_group(dp, OVS_PACKET_CMD_ACTION));
- NLA_PUT_U32(skb, OVS_PACKET_CMD_SAMPLE, packet_mc_group(dp, OVS_PACKET_CMD_SAMPLE));
- nla_nest_end(skb, nla);
-
return genlmsg_end(skb, ovs_header);
nla_put_failure:
static int ovs_dp_cmd_validate(struct nlattr *a[OVS_DP_ATTR_MAX + 1])
{
- if (a[OVS_DP_ATTR_IPV4_FRAGS]) {
- u32 frags = nla_get_u32(a[OVS_DP_ATTR_IPV4_FRAGS]);
-
- if (frags != OVS_DP_FRAG_ZERO && frags != OVS_DP_FRAG_DROP)
- return -EINVAL;
- }
-
return CHECK_NUL_STRING(a[OVS_DP_ATTR_NAME], IFNAMSIZ - 1);
}
/* Called with genl_mutex and optionally with RTNL lock also. */
-static struct datapath *lookup_datapath(struct ovs_header *ovs_header, struct nlattr *a[OVS_DP_ATTR_MAX + 1])
+static struct datapath *lookup_datapath(struct ovs_header *ovs_header,
+ struct nlattr *a[OVS_DP_ATTR_MAX + 1])
{
struct datapath *dp;
return dp ? dp : ERR_PTR(-ENODEV);
}
-/* Called with genl_mutex. */
-static void change_datapath(struct datapath *dp, struct nlattr *a[OVS_DP_ATTR_MAX + 1])
-{
- if (a[OVS_DP_ATTR_IPV4_FRAGS])
- dp->drop_frags = nla_get_u32(a[OVS_DP_ATTR_IPV4_FRAGS]) == OVS_DP_FRAG_DROP;
- if (a[OVS_DP_ATTR_SAMPLING])
- dp->sflow_probability = nla_get_u32(a[OVS_DP_ATTR_SAMPLING]);
-}
-
static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
{
struct nlattr **a = info->attrs;
int err;
err = -EINVAL;
- if (!a[OVS_DP_ATTR_NAME])
+ if (!a[OVS_DP_ATTR_NAME] || !a[OVS_DP_ATTR_UPCALL_PID])
goto err;
err = ovs_dp_cmd_validate(a);
if (!dp->table)
goto err_free_dp;
+ dp->stats_percpu = alloc_percpu(struct dp_stats_percpu);
+ if (!dp->stats_percpu) {
+ err = -ENOMEM;
+ goto err_destroy_table;
+ }
+
/* Set up our datapath device. */
parms.name = nla_data(a[OVS_DP_ATTR_NAME]);
parms.type = OVS_VPORT_TYPE_INTERNAL;
parms.options = NULL;
parms.dp = dp;
parms.port_no = OVSP_LOCAL;
+ parms.upcall_pid = nla_get_u32(a[OVS_DP_ATTR_UPCALL_PID]);
+
vport = new_vport(&parms);
if (IS_ERR(vport)) {
err = PTR_ERR(vport);
if (err == -EBUSY)
err = -EEXIST;
- goto err_destroy_table;
+ goto err_destroy_percpu;
}
- dp->dp_ifindex = vport_get_ifindex(vport);
-
- dp->drop_frags = 0;
- dp->stats_percpu = alloc_percpu(struct dp_stats_percpu);
- if (!dp->stats_percpu) {
- err = -ENOMEM;
- goto err_destroy_local_port;
- }
-
- change_datapath(dp, a);
- reply = ovs_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, OVS_DP_CMD_NEW);
+ reply = ovs_dp_cmd_build_info(dp, info->snd_pid,
+ info->snd_seq, OVS_DP_CMD_NEW);
err = PTR_ERR(reply);
if (IS_ERR(reply))
goto err_destroy_local_port;
err_destroy_local_port:
dp_detach_port(get_vport_protected(dp, OVSP_LOCAL));
+err_destroy_percpu:
+ free_percpu(dp->stats_percpu);
err_destroy_table:
flow_tbl_destroy(get_table_protected(dp));
err_free_dp:
if (IS_ERR(dp))
goto exit_unlock;
- reply = ovs_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, OVS_DP_CMD_DEL);
+ reply = ovs_dp_cmd_build_info(dp, info->snd_pid,
+ info->snd_seq, OVS_DP_CMD_DEL);
err = PTR_ERR(reply);
if (IS_ERR(reply))
goto exit_unlock;
- list_for_each_entry_safe (vport, next_vport, &dp->port_list, node)
+ list_for_each_entry_safe(vport, next_vport, &dp->port_list, node)
if (vport->port_no != OVSP_LOCAL)
dp_detach_port(vport);
if (IS_ERR(dp))
return PTR_ERR(dp);
- change_datapath(dp, info->attrs);
-
- reply = ovs_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, OVS_DP_CMD_NEW);
+ reply = ovs_dp_cmd_build_info(dp, info->snd_pid,
+ info->snd_seq, OVS_DP_CMD_NEW);
if (IS_ERR(reply)) {
err = PTR_ERR(reply);
netlink_set_err(INIT_NET_GENL_SOCK, 0,
if (IS_ERR(dp))
return PTR_ERR(dp);
- reply = ovs_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, OVS_DP_CMD_NEW);
+ reply = ovs_dp_cmd_build_info(dp, info->snd_pid,
+ info->snd_seq, OVS_DP_CMD_NEW);
if (IS_ERR(reply))
return PTR_ERR(reply);
int skip = cb->args[0];
int i = 0;
- list_for_each_entry (dp, &dps, list_node) {
+ list_for_each_entry(dp, &dps, list_node) {
if (i < skip)
continue;
if (ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).pid,
static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = {
#ifdef HAVE_NLA_NUL_STRING
[OVS_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 },
- [OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
- [OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
[OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct ovs_vport_stats) },
[OVS_VPORT_ATTR_ADDRESS] = { .len = ETH_ALEN },
#else
[OVS_VPORT_ATTR_STATS] = { .minlen = sizeof(struct ovs_vport_stats) },
[OVS_VPORT_ATTR_ADDRESS] = { .minlen = ETH_ALEN },
#endif
+ [OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 },
+ [OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 },
+ [OVS_VPORT_ATTR_UPCALL_PID] = { .type = NLA_U32 },
[OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED },
};
.id = GENL_ID_GENERATE,
.hdrsize = sizeof(struct ovs_header),
.name = OVS_VPORT_FAMILY,
- .version = 1,
+ .version = OVS_VPORT_VERSION,
.maxattr = OVS_VPORT_ATTR_MAX
};
{
struct ovs_header *ovs_header;
struct nlattr *nla;
- int ifindex;
int err;
ovs_header = genlmsg_put(skb, pid, seq, &dp_vport_genl_family,
if (!ovs_header)
return -EMSGSIZE;
- ovs_header->dp_ifindex = vport->dp->dp_ifindex;
+ ovs_header->dp_ifindex = get_dpifindex(vport->dp);
NLA_PUT_U32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no);
NLA_PUT_U32(skb, OVS_VPORT_ATTR_TYPE, vport_get_type(vport));
NLA_PUT_STRING(skb, OVS_VPORT_ATTR_NAME, vport_get_name(vport));
+ NLA_PUT_U32(skb, OVS_VPORT_ATTR_UPCALL_PID, vport->upcall_pid);
- nla = nla_reserve(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats));
+ nla = nla_reserve(skb, OVS_VPORT_ATTR_STATS,
+ sizeof(struct ovs_vport_stats));
if (!nla)
goto nla_put_failure;
if (err == -EMSGSIZE)
goto error;
- ifindex = vport_get_ifindex(vport);
- if (ifindex > 0)
- NLA_PUT_U32(skb, OVS_VPORT_ATTR_IFINDEX, ifindex);
-
return genlmsg_end(skb, ovs_header);
nla_put_failure:
}
/* Called with RTNL lock. */
-static int change_vport(struct vport *vport, struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
+static int change_vport(struct vport *vport,
+ struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
{
int err = 0;
int err;
err = -EINVAL;
- if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE])
+ if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE] ||
+ !a[OVS_VPORT_ATTR_UPCALL_PID])
goto exit;
err = ovs_vport_cmd_validate(a);
parms.options = a[OVS_VPORT_ATTR_OPTIONS];
parms.dp = dp;
parms.port_no = port_no;
+ parms.upcall_pid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
vport = new_vport(&parms);
err = PTR_ERR(vport);
if (IS_ERR(vport))
goto exit_unlock;
- dp_sysfs_add_if(vport);
+ dp_sysfs_add_if(vport);
err = change_vport(vport, a);
if (!err) {
reply = ovs_vport_cmd_build_info(vport, info->snd_pid,
- info->snd_seq, OVS_VPORT_CMD_NEW);
+ info->snd_seq,
+ OVS_VPORT_CMD_NEW);
if (IS_ERR(reply))
err = PTR_ERR(reply);
}
goto exit_unlock;
err = 0;
- if (a[OVS_VPORT_ATTR_OPTIONS])
+ if (a[OVS_VPORT_ATTR_TYPE] &&
+ nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport_get_type(vport))
+ err = -EINVAL;
+
+ if (!err && a[OVS_VPORT_ATTR_OPTIONS])
err = vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]);
if (!err)
err = change_vport(vport, a);
+ if (!err && a[OVS_VPORT_ATTR_UPCALL_PID])
+ vport->upcall_pid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]);
reply = ovs_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq,
OVS_VPORT_CMD_NEW);
}
}
- err = packet_register_mc_groups();
- if (err)
- goto error;
return 0;
error:
BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > sizeof(dummy_skb->cb));
- printk("Open vSwitch %s, built "__DATE__" "__TIME__"\n", VERSION BUILDNR);
+ pr_info("Open vSwitch %s, built "__DATE__" "__TIME__"\n",
+ VERSION BUILDNR);
err = tnl_init();
if (err)