上一节我们讲了ovs-vswitchd,其中虚拟网桥初始化的时候,对调用内核模块来添加虚拟网卡。
我们从openvswitch内核模块的加载过程,来看这个过程。
在datapath/datapath.c中会调用module_init(dp_init);来初始化内核模块。
static
int __init dp_init(void)
{
int err;
BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > FIELD_SIZEOF(struct sk_buff, cb));
pr_info("Open vSwitch switching datapath %s\n", VERSION);
err = compat_init();
if (err)
goto error;
err = action_fifos_init();
if (err)
goto error_compat_exit;
err = ovs_internal_dev_rtnl_link_register();
if (err)
goto error_action_fifos_exit;
err = ovs_flow_init();
if (err)
goto error_unreg_rtnl_link;
err = ovs_vport_init();
if (err)
goto error_flow_exit;
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 = ovs_netdev_init();
if (err)
goto error_unreg_notifier;
err = dp_register_genl();
if (err < 0)
goto error_unreg_netdev;
return 0;
error_unreg_netdev:
ovs_netdev_exit();
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:
ovs_flow_exit();
error_unreg_rtnl_link:
ovs_internal_dev_rtnl_link_unregister();
error_action_fifos_exit:
action_fifos_exit();
error_compat_exit:
compat_exit();
error:
return err;
}
|
其中比较重要的是调用了dp_register_genl(),这个就是注册netlink函数,从而ovs-vswitchd可以通过netlink调用内核。
static
int dp_register_genl(void)
{
int err;
int i;
for (i = 0; i < ARRAY_SIZE(dp_genl_families); i++) {
err = genl_register_family(dp_genl_families[i]);
if (err)
goto error;
}
return 0;
error:
dp_unregister_genl(i);
return err;
}
|
这里dp_genl_families由四个netlink的family组成
static
struct genl_family *dp_genl_families[] = {
&dp_datapath_genl_family,
&dp_vport_genl_family,
&dp_flow_genl_family,
&dp_packet_genl_family,
};
|
其中分别定义了以下的操作:
Family名称 | .name | .ops |
dp_datapath_genl_family | static
struct genl_family dp_datapath_genl_family = {
.id = GENL_ID_GENERATE,
.hdrsize = sizeof(struct ovs_header),
.name = OVS_DATAPATH_FAMILY,
.version = OVS_DATAPATH_VERSION,
.maxattr = OVS_DP_ATTR_MAX,
.netnsok = true,
.parallel_ops = true,
.ops = dp_datapath_genl_ops,
.n_ops = ARRAY_SIZE(dp_datapath_genl_ops),
.mcgrps = &ovs_dp_datapath_multicast_group,
.n_mcgrps = 1,
};
| static
struct genl_ops dp_datapath_genl_ops[] = {
{ .cmd = OVS_DP_CMD_NEW,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = datapath_policy,
.doit = ovs_dp_cmd_new
},
{ .cmd = OVS_DP_CMD_DEL,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = datapath_policy,
.doit = ovs_dp_cmd_del
},
{ .cmd = OVS_DP_CMD_GET,
.flags = 0, /* OK for unprivileged users. */
.policy = datapath_policy,
.doit = ovs_dp_cmd_get,
.dumpit = ovs_dp_cmd_dump
},
{ .cmd = OVS_DP_CMD_SET,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = datapath_policy,
.doit = ovs_dp_cmd_set,
},
};
|
dp_vport_genl_family | struct genl_family dp_vport_genl_family = {
.id = GENL_ID_GENERATE,
.hdrsize = sizeof(struct ovs_header),
.name = OVS_VPORT_FAMILY,
.version = OVS_VPORT_VERSION,
.maxattr = OVS_VPORT_ATTR_MAX,
.netnsok = true,
.parallel_ops = true,
.ops = dp_vport_genl_ops,
.n_ops = ARRAY_SIZE(dp_vport_genl_ops),
.mcgrps = &ovs_dp_vport_multicast_group,
.n_mcgrps = 1,
};
| static
struct genl_ops dp_vport_genl_ops[] = {
{ .cmd = OVS_VPORT_CMD_NEW,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = vport_policy,
.doit = ovs_vport_cmd_new
},
{ .cmd = OVS_VPORT_CMD_DEL,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = vport_policy,
.doit = ovs_vport_cmd_del
},
{ .cmd = OVS_VPORT_CMD_GET,
.flags = 0, /* OK for unprivileged users. */
.policy = vport_policy,
.doit = ovs_vport_cmd_get,
.dumpit = ovs_vport_cmd_dump
},
{ .cmd = OVS_VPORT_CMD_SET,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = vport_policy,
.doit = ovs_vport_cmd_set,
},
};
|
dp_flow_genl_family | static
struct genl_family dp_flow_genl_family = {
.id = GENL_ID_GENERATE,
.hdrsize = sizeof(struct ovs_header),
.name = OVS_FLOW_FAMILY,
.version = OVS_FLOW_VERSION,
.maxattr = OVS_FLOW_ATTR_MAX,
.netnsok = true,
.parallel_ops = true,
.ops = dp_flow_genl_ops,
.n_ops = ARRAY_SIZE(dp_flow_genl_ops),
.mcgrps = &ovs_dp_flow_multicast_group,
.n_mcgrps = 1,
};
| static
struct genl_ops dp_flow_genl_ops[] = {
{ .cmd = OVS_FLOW_CMD_NEW,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = flow_policy,
.doit = ovs_flow_cmd_new
},
{ .cmd = OVS_FLOW_CMD_DEL,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = flow_policy,
.doit = ovs_flow_cmd_del
},
{ .cmd = OVS_FLOW_CMD_GET,
.flags = 0, /* OK for unprivileged users. */
.policy = flow_policy,
.doit = ovs_flow_cmd_get,
.dumpit = ovs_flow_cmd_dump
},
{ .cmd = OVS_FLOW_CMD_SET,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = flow_policy,
.doit = ovs_flow_cmd_set,
},
};
|
dp_packet_genl_family | static
struct genl_family dp_packet_genl_family = {
.id = GENL_ID_GENERATE,
.hdrsize = sizeof(struct ovs_header),
.name = OVS_PACKET_FAMILY,
.version = OVS_PACKET_VERSION,
.maxattr = OVS_PACKET_ATTR_MAX,
.netnsok = true,
.parallel_ops = true,
.ops = dp_packet_genl_ops,
.n_ops = ARRAY_SIZE(dp_packet_genl_ops),
};
| static
struct genl_ops dp_packet_genl_ops[] = {
{ .cmd = OVS_PACKET_CMD_EXECUTE,
.flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */
.policy = packet_policy,
.doit = ovs_packet_cmd_execute
}
};
|
如上一节中,ovs-vswitchd启动的时候,将虚拟网卡添加到虚拟交换机上的时候,会调用netlink的OVS_VPORT_CMD_NEW命令,因而会调用函数ovs_vport_cmd_new。
会调用static struct vport *new_vport(const struct vport_parms *parms)
会调用struct vport *ovs_vport_add(const struct vport_parms *parms)里面会调用vport = ops->create(parms);
ops是什么呢?在dp_init函数中会调用ovs_netdev_init,它会调用ovs_vport_ops_register(&ovs_netdev_vport_ops);
static
struct vport_ops ovs_netdev_vport_ops = {
.type = OVS_VPORT_TYPE_NETDEV,
.create = netdev_create,
.destroy = netdev_destroy,
.send = dev_queue_xmit,
};
|
所以ops->create会调用netdev_create,它会调用ovs_netdev_link,其中有下面的代码:
err = netdev_rx_handler_register(vport->dev, netdev_frame_hook,
vport);
|
注册一个方法叫做netdev_frame_hook,每当网卡收到包的时候,就调用这个方法。
在下面的章节中,我们会从这个函数开始,解析整个网络包的处理过程。