Commit 46df7b81 authored by Pravin B Shelar's avatar Pravin B Shelar Committed by Jesse Gross

openvswitch: Add support for network namespaces.

Following patch adds support for network namespace to openvswitch.
Since it must release devices when namespaces are destroyed, a
side effect of this patch is that the module no longer keeps a
refcount but instead cleans up any state when it is unloaded.
Signed-off-by: default avatarPravin B Shelar <pshelar@nicira.com>
Signed-off-by: default avatarJesse Gross <jesse@nicira.com>
parent 0d7614f0
......@@ -49,11 +49,28 @@
#include <linux/dmi.h>
#include <linux/workqueue.h>
#include <net/genetlink.h>
#include <net/net_namespace.h>
#include <net/netns/generic.h>
#include "datapath.h"
#include "flow.h"
#include "vport-internal_dev.h"
/**
* struct ovs_net - Per net-namespace data for ovs.
* @dps: List of datapaths to enable dumping them all out.
* Protected by genl_mutex.
*/
struct ovs_net {
struct list_head dps;
};
static int ovs_net_id __read_mostly;
#define REHASH_FLOW_INTERVAL (10 * 60 * HZ)
static void rehash_flow_table(struct work_struct *work);
static DECLARE_DELAYED_WORK(rehash_flow_wq, rehash_flow_table);
/**
* DOC: Locking:
*
......@@ -71,29 +88,21 @@
* each other.
*/
/* Global list of datapaths to enable dumping them all out.
* Protected by genl_mutex.
*/
static LIST_HEAD(dps);
#define REHASH_FLOW_INTERVAL (10 * 60 * HZ)
static void rehash_flow_table(struct work_struct *work);
static DECLARE_DELAYED_WORK(rehash_flow_wq, rehash_flow_table);
static struct vport *new_vport(const struct vport_parms *);
static int queue_gso_packets(int dp_ifindex, struct sk_buff *,
static int queue_gso_packets(struct net *, int dp_ifindex, struct sk_buff *,
const struct dp_upcall_info *);
static int queue_userspace_packet(int dp_ifindex, struct sk_buff *,
static int queue_userspace_packet(struct net *, int dp_ifindex,
struct sk_buff *,
const struct dp_upcall_info *);
/* Must be called with rcu_read_lock, genl_mutex, or RTNL lock. */
static struct datapath *get_dp(int dp_ifindex)
static struct datapath *get_dp(struct net *net, int dp_ifindex)
{
struct datapath *dp = NULL;
struct net_device *dev;
rcu_read_lock();
dev = dev_get_by_index_rcu(&init_net, dp_ifindex);
dev = dev_get_by_index_rcu(net, dp_ifindex);
if (dev) {
struct vport *vport = ovs_internal_dev_get_vport(dev);
if (vport)
......@@ -135,6 +144,7 @@ static void destroy_dp_rcu(struct rcu_head *rcu)
ovs_flow_tbl_destroy((__force struct flow_table *)dp->table);
free_percpu(dp->stats_percpu);
release_net(ovs_dp_get_net(dp));
kfree(dp);
}
......@@ -220,11 +230,12 @@ static struct genl_family dp_packet_genl_family = {
.hdrsize = sizeof(struct ovs_header),
.name = OVS_PACKET_FAMILY,
.version = OVS_PACKET_VERSION,
.maxattr = OVS_PACKET_ATTR_MAX
.maxattr = OVS_PACKET_ATTR_MAX,
.netnsok = true
};
int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
const struct dp_upcall_info *upcall_info)
const struct dp_upcall_info *upcall_info)
{
struct dp_stats_percpu *stats;
int dp_ifindex;
......@@ -242,9 +253,9 @@ int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
}
if (!skb_is_gso(skb))
err = queue_userspace_packet(dp_ifindex, skb, upcall_info);
err = queue_userspace_packet(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
else
err = queue_gso_packets(dp_ifindex, skb, upcall_info);
err = queue_gso_packets(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
if (err)
goto err;
......@@ -260,7 +271,8 @@ err:
return err;
}
static int queue_gso_packets(int dp_ifindex, struct sk_buff *skb,
static int queue_gso_packets(struct net *net, int dp_ifindex,
struct sk_buff *skb,
const struct dp_upcall_info *upcall_info)
{
unsigned short gso_type = skb_shinfo(skb)->gso_type;
......@@ -276,7 +288,7 @@ static int queue_gso_packets(int dp_ifindex, struct sk_buff *skb,
/* Queue all of the segments. */
skb = segs;
do {
err = queue_userspace_packet(dp_ifindex, skb, upcall_info);
err = queue_userspace_packet(net, dp_ifindex, skb, upcall_info);
if (err)
break;
......@@ -306,7 +318,8 @@ static int queue_gso_packets(int dp_ifindex, struct sk_buff *skb,
return err;
}
static int queue_userspace_packet(int dp_ifindex, struct sk_buff *skb,
static int queue_userspace_packet(struct net *net, int dp_ifindex,
struct sk_buff *skb,
const struct dp_upcall_info *upcall_info)
{
struct ovs_header *upcall;
......@@ -362,7 +375,7 @@ static int queue_userspace_packet(int dp_ifindex, struct sk_buff *skb,
skb_copy_and_csum_dev(skb, nla_data(nla));
err = genlmsg_unicast(&init_net, user_skb, upcall_info->pid);
err = genlmsg_unicast(net, user_skb, upcall_info->pid);
out:
kfree_skb(nskb);
......@@ -370,15 +383,10 @@ out:
}
/* Called with genl_mutex. */
static int flush_flows(int dp_ifindex)
static int flush_flows(struct datapath *dp)
{
struct flow_table *old_table;
struct flow_table *new_table;
struct datapath *dp;
dp = get_dp(dp_ifindex);
if (!dp)
return -ENODEV;
old_table = genl_dereference(dp->table);
new_table = ovs_flow_tbl_alloc(TBL_MIN_BUCKETS);
......@@ -668,7 +676,7 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
packet->priority = flow->key.phy.priority;
rcu_read_lock();
dp = get_dp(ovs_header->dp_ifindex);
dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
err = -ENODEV;
if (!dp)
goto err_unlock;
......@@ -742,7 +750,8 @@ static struct genl_family dp_flow_genl_family = {
.hdrsize = sizeof(struct ovs_header),
.name = OVS_FLOW_FAMILY,
.version = OVS_FLOW_VERSION,
.maxattr = OVS_FLOW_ATTR_MAX
.maxattr = OVS_FLOW_ATTR_MAX,
.netnsok = true
};
static struct genl_multicast_group ovs_dp_flow_multicast_group = {
......@@ -894,7 +903,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
goto error;
}
dp = get_dp(ovs_header->dp_ifindex);
dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
error = -ENODEV;
if (!dp)
goto error;
......@@ -995,7 +1004,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
ovs_dp_flow_multicast_group.id, info->nlhdr,
GFP_KERNEL);
else
netlink_set_err(init_net.genl_sock, 0,
netlink_set_err(sock_net(skb->sk)->genl_sock, 0,
ovs_dp_flow_multicast_group.id, PTR_ERR(reply));
return 0;
......@@ -1023,7 +1032,7 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
if (err)
return err;
dp = get_dp(ovs_header->dp_ifindex);
dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
if (!dp)
return -ENODEV;
......@@ -1052,16 +1061,17 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
int err;
int key_len;
dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
if (!dp)
return -ENODEV;
if (!a[OVS_FLOW_ATTR_KEY])
return flush_flows(ovs_header->dp_ifindex);
return flush_flows(dp);
err = ovs_flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]);
if (err)
return err;
dp = get_dp(ovs_header->dp_ifindex);
if (!dp)
return -ENODEV;
table = genl_dereference(dp->table);
flow = ovs_flow_tbl_lookup(table, &key, key_len);
if (!flow)
......@@ -1090,7 +1100,7 @@ static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
struct datapath *dp;
struct flow_table *table;
dp = get_dp(ovs_header->dp_ifindex);
dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
if (!dp)
return -ENODEV;
......@@ -1152,7 +1162,8 @@ static struct genl_family dp_datapath_genl_family = {
.hdrsize = sizeof(struct ovs_header),
.name = OVS_DATAPATH_FAMILY,
.version = OVS_DATAPATH_VERSION,
.maxattr = OVS_DP_ATTR_MAX
.maxattr = OVS_DP_ATTR_MAX,
.netnsok = true
};
static struct genl_multicast_group ovs_dp_datapath_multicast_group = {
......@@ -1210,18 +1221,19 @@ static struct sk_buff *ovs_dp_cmd_build_info(struct datapath *dp, u32 pid,
}
/* Called with genl_mutex and optionally with RTNL lock also. */
static struct datapath *lookup_datapath(struct ovs_header *ovs_header,
static struct datapath *lookup_datapath(struct net *net,
struct ovs_header *ovs_header,
struct nlattr *a[OVS_DP_ATTR_MAX + 1])
{
struct datapath *dp;
if (!a[OVS_DP_ATTR_NAME])
dp = get_dp(ovs_header->dp_ifindex);
dp = get_dp(net, ovs_header->dp_ifindex);
else {
struct vport *vport;
rcu_read_lock();
vport = ovs_vport_locate(nla_data(a[OVS_DP_ATTR_NAME]));
vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME]));
dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL;
rcu_read_unlock();
}
......@@ -1235,6 +1247,7 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
struct sk_buff *reply;
struct datapath *dp;
struct vport *vport;
struct ovs_net *ovs_net;
int err;
err = -EINVAL;
......@@ -1242,15 +1255,14 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
goto err;
rtnl_lock();
err = -ENODEV;
if (!try_module_get(THIS_MODULE))
goto err_unlock_rtnl;
err = -ENOMEM;
dp = kzalloc(sizeof(*dp), GFP_KERNEL);
if (dp == NULL)
goto err_put_module;
goto err_unlock_rtnl;
INIT_LIST_HEAD(&dp->port_list);
ovs_dp_set_net(dp, hold_net(sock_net(skb->sk)));
/* Allocate table. */
err = -ENOMEM;
......@@ -1287,7 +1299,8 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
if (IS_ERR(reply))
goto err_destroy_local_port;
list_add_tail(&dp->list_node, &dps);
ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id);
list_add_tail(&dp->list_node, &ovs_net->dps);
rtnl_unlock();
genl_notify(reply, genl_info_net(info), info->snd_pid,
......@@ -1302,34 +1315,20 @@ err_destroy_percpu:
err_destroy_table:
ovs_flow_tbl_destroy(genl_dereference(dp->table));
err_free_dp:
release_net(ovs_dp_get_net(dp));
kfree(dp);
err_put_module:
module_put(THIS_MODULE);
err_unlock_rtnl:
rtnl_unlock();
err:
return err;
}
static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info)
/* Called with genl_mutex. */
static void __dp_destroy(struct datapath *dp)
{
struct vport *vport, *next_vport;
struct sk_buff *reply;
struct datapath *dp;
int err;
rtnl_lock();
dp = lookup_datapath(info->userhdr, info->attrs);
err = PTR_ERR(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);
err = PTR_ERR(reply);
if (IS_ERR(reply))
goto exit_unlock;
list_for_each_entry_safe(vport, next_vport, &dp->port_list, node)
if (vport->port_no != OVSP_LOCAL)
ovs_dp_detach_port(vport);
......@@ -1345,17 +1344,32 @@ static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info)
rtnl_unlock();
call_rcu(&dp->rcu, destroy_dp_rcu);
module_put(THIS_MODULE);
}
static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info)
{
struct sk_buff *reply;
struct datapath *dp;
int err;
dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
err = PTR_ERR(dp);
if (IS_ERR(dp))
return err;
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))
return err;
__dp_destroy(dp);
genl_notify(reply, genl_info_net(info), info->snd_pid,
ovs_dp_datapath_multicast_group.id, info->nlhdr,
GFP_KERNEL);
return 0;
exit_unlock:
rtnl_unlock();
return err;
}
static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info)
......@@ -1364,7 +1378,7 @@ static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info)
struct datapath *dp;
int err;
dp = lookup_datapath(info->userhdr, info->attrs);
dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
if (IS_ERR(dp))
return PTR_ERR(dp);
......@@ -1372,7 +1386,7 @@ static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info)
info->snd_seq, OVS_DP_CMD_NEW);
if (IS_ERR(reply)) {
err = PTR_ERR(reply);
netlink_set_err(init_net.genl_sock, 0,
netlink_set_err(sock_net(skb->sk)->genl_sock, 0,
ovs_dp_datapath_multicast_group.id, err);
return 0;
}
......@@ -1389,7 +1403,7 @@ static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
struct sk_buff *reply;
struct datapath *dp;
dp = lookup_datapath(info->userhdr, info->attrs);
dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs);
if (IS_ERR(dp))
return PTR_ERR(dp);
......@@ -1403,11 +1417,12 @@ static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info)
static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
{
struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
struct datapath *dp;
int skip = cb->args[0];
int i = 0;
list_for_each_entry(dp, &dps, list_node) {
list_for_each_entry(dp, &ovs_net->dps, list_node) {
if (i >= skip &&
ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).pid,
cb->nlh->nlmsg_seq, NLM_F_MULTI,
......@@ -1459,7 +1474,8 @@ static struct genl_family dp_vport_genl_family = {
.hdrsize = sizeof(struct ovs_header),
.name = OVS_VPORT_FAMILY,
.version = OVS_VPORT_VERSION,
.maxattr = OVS_VPORT_ATTR_MAX
.maxattr = OVS_VPORT_ATTR_MAX,
.netnsok = true
};
struct genl_multicast_group ovs_dp_vport_multicast_group = {
......@@ -1525,14 +1541,15 @@ struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 pid,
}
/* Called with RTNL lock or RCU read lock. */
static struct vport *lookup_vport(struct ovs_header *ovs_header,
static struct vport *lookup_vport(struct net *net,
struct ovs_header *ovs_header,
struct nlattr *a[OVS_VPORT_ATTR_MAX + 1])
{
struct datapath *dp;
struct vport *vport;
if (a[OVS_VPORT_ATTR_NAME]) {
vport = ovs_vport_locate(nla_data(a[OVS_VPORT_ATTR_NAME]));
vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME]));
if (!vport)
return ERR_PTR(-ENODEV);
if (ovs_header->dp_ifindex &&
......@@ -1545,7 +1562,7 @@ static struct vport *lookup_vport(struct ovs_header *ovs_header,
if (port_no >= DP_MAX_PORTS)
return ERR_PTR(-EFBIG);
dp = get_dp(ovs_header->dp_ifindex);
dp = get_dp(net, ovs_header->dp_ifindex);
if (!dp)
return ERR_PTR(-ENODEV);
......@@ -1574,7 +1591,7 @@ static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info)
goto exit;
rtnl_lock();
dp = get_dp(ovs_header->dp_ifindex);
dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
err = -ENODEV;
if (!dp)
goto exit_unlock;
......@@ -1638,7 +1655,7 @@ static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info)
int err;
rtnl_lock();
vport = lookup_vport(info->userhdr, a);
vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
err = PTR_ERR(vport);
if (IS_ERR(vport))
goto exit_unlock;
......@@ -1658,7 +1675,7 @@ static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info)
reply = ovs_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq,
OVS_VPORT_CMD_NEW);
if (IS_ERR(reply)) {
netlink_set_err(init_net.genl_sock, 0,
netlink_set_err(sock_net(skb->sk)->genl_sock, 0,
ovs_dp_vport_multicast_group.id, PTR_ERR(reply));
goto exit_unlock;
}
......@@ -1679,7 +1696,7 @@ static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info)
int err;
rtnl_lock();
vport = lookup_vport(info->userhdr, a);
vport = lookup_vport(sock_net(skb->sk), info->userhdr, a);
err = PTR_ERR(vport);
if (IS_ERR(vport))
goto exit_unlock;
......@@ -1714,7 +1731,7 @@ static int ovs_vport_cmd_get(struct sk_buff *skb, struct genl_info *info)
int err;
rcu_read_lock();
vport = lookup_vport(ovs_header, a);
vport = lookup_vport(sock_net(skb->sk), ovs_header, a);
err = PTR_ERR(vport);
if (IS_ERR(vport))
goto exit_unlock;
......@@ -1741,7 +1758,7 @@ static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
u32 port_no;
int retval;
dp = get_dp(ovs_header->dp_ifindex);
dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
if (!dp)
return -ENODEV;
......@@ -1766,28 +1783,6 @@ static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
return retval;
}
static void rehash_flow_table(struct work_struct *work)
{
struct datapath *dp;
genl_lock();
list_for_each_entry(dp, &dps, list_node) {
struct flow_table *old_table = genl_dereference(dp->table);
struct flow_table *new_table;
new_table = ovs_flow_tbl_rehash(old_table);
if (!IS_ERR(new_table)) {
rcu_assign_pointer(dp->table, new_table);
ovs_flow_tbl_deferred_destroy(old_table);
}
}
genl_unlock();
schedule_delayed_work(&rehash_flow_wq, REHASH_FLOW_INTERVAL);
}
static struct genl_ops dp_vport_genl_ops[] = {
{ .cmd = OVS_VPORT_CMD_NEW,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
......@@ -1872,6 +1867,59 @@ error:
return err;
}
static void rehash_flow_table(struct work_struct *work)
{
struct datapath *dp;
struct net *net;
genl_lock();
rtnl_lock();
for_each_net(net) {
struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
list_for_each_entry(dp, &ovs_net->dps, list_node) {
struct flow_table *old_table = genl_dereference(dp->table);
struct flow_table *new_table;
new_table = ovs_flow_tbl_rehash(old_table);
if (!IS_ERR(new_table)) {
rcu_assign_pointer(dp->table, new_table);
ovs_flow_tbl_deferred_destroy(old_table);
}
}
}
rtnl_unlock();
genl_unlock();
schedule_delayed_work(&rehash_flow_wq, REHASH_FLOW_INTERVAL);
}
static int __net_init ovs_init_net(struct net *net)
{
struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
INIT_LIST_HEAD(&ovs_net->dps);
return 0;
}
static void __net_exit ovs_exit_net(struct net *net)
{
struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
struct datapath *dp, *dp_next;
genl_lock();
list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node)
__dp_destroy(dp);
genl_unlock();
}
static struct pernet_operations ovs_net_ops = {
.init = ovs_init_net,
.exit = ovs_exit_net,
.id = &ovs_net_id,
.size = sizeof(struct ovs_net),
};
static int __init dp_init(void)
{
struct sk_buff *dummy_skb;
......@@ -1889,10 +1937,14 @@ static int __init dp_init(void)
if (err)
goto error_flow_exit;
err = register_netdevice_notifier(&ovs_dp_device_notifier);
err = register_pernet_device(&ovs_net_ops);
if (err)
goto error_vport_exit;
err = register_netdevice_notifier(&ovs_dp_device_notifier);
if (err)
goto error_netns_exit;
err = dp_register_genl();
if (err < 0)
goto error_unreg_notifier;
......@@ -1903,6 +1955,8 @@ static int __init dp_init(void)
error_unreg_notifier:
unregister_netdevice_notifier(&ovs_dp_device_notifier);
error_netns_exit:
unregister_pernet_device(&ovs_net_ops);
error_vport_exit:
ovs_vport_exit();
error_flow_exit:
......@@ -1914,9 +1968,10 @@ error:
static void dp_cleanup(void)
{
cancel_delayed_work_sync(&rehash_flow_wq);
rcu_barrier();
dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
unregister_netdevice_notifier(&ovs_dp_device_notifier);
unregister_pernet_device(&ovs_net_ops);
rcu_barrier();
ovs_vport_exit();
ovs_flow_exit();
}
......
......@@ -27,8 +27,7 @@
#include <linux/u64_stats_sync.h>
#include "flow.h"
struct vport;
#include "vport.h"
#define DP_MAX_PORTS 1024
#define SAMPLE_ACTION_DEPTH 3
......@@ -63,6 +62,7 @@ struct dp_stats_percpu {