Neutron分析（5）—— neutron-l3-agent中的iptables

一.iptables简介

1.iptables数据包处理流程

以本机为目的的包，由上至下，走左边的路
本机产生的包，从local process开始走左边的路
本机转发的包，由上至下走右边的路

简化流程如下：

2.iptables表结构

在neutron中主要用到filter表和nat表
filter表:
Chain INPUT
Chain FORWARD
Chain OUTPUT
filter表用于信息包过滤，它包含INPUT、OUTPUT和FORWARD 链。

nat表:
Chain PREROUTING
Chain OUTPUT
Chain POSTROUTING
nat表用于网络地址转换，PREROUTING链由指定信息包一到达防火墙就改变它们的规则所组成，而 POSTROUTING 链由指定正当信息包打算离开防火墙时改变它们的规则所组成。

More：
Traversing of tables and chains
Linux Firewalls Using iptables

二.l3 agent消息处理

_rpc_loop  ---  _process_router            ---  _router_added
 
                                           ---  process_router
                                 
                                           ---  _router_removed
           
           ---  _process_router_delete     ---  _router_removed

 

在上面几个方法中，会涉及到iptables的处理。

三.iptables_manager初始化

iptables_manager的初始化是在class IptablesManager中完成的，它对iptables的链进行了包装。

源码目录：neutron/neutron/agent/linux/iptables_manager.py

主要操作：

新建一个neutron-filter-top链，这个是没有包装的，加在原生的FORWARD和OUTPUT链上。
对filter表的INPUT，OUTPUT，FORWARD链进行包装，将到达原链的数据包转发到包装链，还增加一个包装的local链。
对于nat表，PREROUTING，OUTPUT，POSTROUTING链进行包装，另外在POSTROUTING链之后加了snat链。

代码分析：

对于l3 agent，binary_name是neturon-l3-agent。

filter表的操作：
增加一个链neutron-filter-top，增加规则：
-A FORWARD -j neutron-filter-top
-A OUTPUT -j neutron-filter-top

增加一个包装链neutron-l3-agent-local，增加规则：
-A neutron-filter-top -j neutron-l3-agent-local

        # Add a neutron-filter-top chain. It's intended to be shared
        # among the various nova components. It sits at the very top
        # of FORWARD and OUTPUT.
        for tables in [self.ipv4, self.ipv6]:
            tables['filter'].add_chain('neutron-filter-top', wrap=False)
            tables['filter'].add_rule('FORWARD', '-j neutron-filter-top',
                                      wrap=False, top=True)
            tables['filter'].add_rule('OUTPUT', '-j neutron-filter-top',
                                      wrap=False, top=True)

            tables['filter'].add_chain('local')
            tables['filter'].add_rule('neutron-filter-top', '-j $local',
                                      wrap=False)

 

包装IPv4和IPv6 filter表的INPUT，OUTPUT，FORWARD链，以及IPv4 nat表的PREROUTING，OUTPUT，POSTROUTING链。

将到达原链的数据包转发到包装链：

        # Wrap the built-in chains
        builtin_chains = {4: {'filter': ['INPUT', 'OUTPUT', 'FORWARD']},
                          6: {'filter': ['INPUT', 'OUTPUT', 'FORWARD']}}

        if not state_less:
            self.ipv4.update(
                {'nat': IptablesTable(binary_name=self.wrap_name)})
            builtin_chains[4].update({'nat': ['PREROUTING',
                                      'OUTPUT', 'POSTROUTING']})

        for ip_version in builtin_chains:
            if ip_version == 4:
                tables = self.ipv4
            elif ip_version == 6:
                tables = self.ipv6

            for table, chains in builtin_chains[ip_version].iteritems():
                for chain in chains:
                    tables[table].add_chain(chain)
                    tables[table].add_rule(chain, '-j $%s' %
                                           (chain), wrap=False)

 

包装链neutron-l3-agent-INPUT，neutron-l3-agent-OUTPUT，neutron-l3-agent-FORWARD，添加规则：
-A INPUT -j neutron-l3-agent-INPUT
-A OUTPUT -j neutron-l3-agent-OUTPUT
-A FORWARD -j neutron-l3-agent-FORWARD

nat表的操作：
（承上面的代码）
包装链neutron-l3-agent-PREROUTING，neutron-l3-agent-OUTPUT，neutron-l3-agent-POSTROUTING，添加规则：
-A PREROUTING -j neutron-l3-agent-PREROUTING
-A OUTPUT -j neutron-l3-agent-OUTPUT
-A POSTROUTING -j neutron-l3-agent-POSTROUTING

nat表中添加neutron-postrouting-bottom链，增加规则：
-A POSTROUTING -j neutron-postrouting-bottom

nat表中添加包装链neutron-l3-agent-snat，增加规则：
-A neutron-postrouting-bottom -j neutron-l3-agent-snat

nat表中添加包装链neutron-l3-agent-float-snat，增加规则：
-A neutron-l3-agent-snat -j neutron-l3-agent-float-snat

代码如下：

        if not state_less:
            # Add a neutron-postrouting-bottom chain. It's intended to be
            # shared among the various nova components. We set it as the last
            # chain of POSTROUTING chain.
            self.ipv4['nat'].add_chain('neutron-postrouting-bottom',
                                       wrap=False)
            self.ipv4['nat'].add_rule('POSTROUTING',
                                      '-j neutron-postrouting-bottom',
                                      wrap=False)

            # We add a snat chain to the shared neutron-postrouting-bottom
            # chain so that it's applied last.
            self.ipv4['nat'].add_chain('snat')
            self.ipv4['nat'].add_rule('neutron-postrouting-bottom',
                                      '-j $snat', wrap=False)

            # And then we add a float-snat chain and jump to first thing in
            # the snat chain.
            self.ipv4['nat'].add_chain('float-snat')
            self.ipv4['nat'].add_rule('snat', '-j $float-snat')

 

四.l3 agent代码中关于iptables的处理

1._router_added

_router_added方法，创建和metadata相关的iptables规则：

    def _router_added(self, router_id, router):
        ri = RouterInfo(router_id, self.root_helper,
                        self.conf.use_namespaces, router)
        self.router_info[router_id] = ri
        if self.conf.use_namespaces:
            self._create_router_namespace(ri)
        for c, r in self.metadata_filter_rules():
            ri.iptables_manager.ipv4['filter'].add_rule(c, r)
        for c, r in self.metadata_nat_rules():
            ri.iptables_manager.ipv4['nat'].add_rule(c, r)
        ri.iptables_manager.apply()
        super(L3NATAgent, self).process_router_add(ri)
        if self.conf.enable_metadata_proxy:
            self._spawn_metadata_proxy(ri.router_id, ri.ns_name)

 

1.metadata_filter_rules方法中，如果enable_metadata_proxy为True，增加规则

    def metadata_filter_rules(self):
        rules = []
        if self.conf.enable_metadata_proxy:
            rules.append(('INPUT', '-s 0.0.0.0/0 -d 127.0.0.1 '
                          '-p tcp -m tcp --dport %s '
                          '-j ACCEPT' % self.conf.metadata_port))
        return rules

 

然后在filter表中增加这条规则，接受所有从外面进来到达metadata_port端口的数据包：
-A neutron-l3-agent-INPUT -s 0.0.0.0/0 -d 127.0.0.1 -p tcp -m tcp –dport 9697 -j ACCEPT

2.metadata_nat_rules方法，如果enable_metadata_proxy为True，增加规则

    def metadata_nat_rules(self):
        rules = []
        if self.conf.enable_metadata_proxy:
            rules.append(('PREROUTING', '-s 0.0.0.0/0 -d 169.254.169.254/32 '
                          '-p tcp -m tcp --dport 80 -j REDIRECT '
                          '--to-port %s' % self.conf.metadata_port))
        return rules

 

然后在nat表中增加这条规则做DNAT转换，在route之前，将虚拟机访问169.254.169.254端口80的数据包重定向到metadat_port端口：
-A neutron-l3-agent-PREROUTING -s 0.0.0.0/0 -d 169.254.169.254/32 -p tcp -m tcp –dport 80 -j REDIRECT –to-port 9697

再调用iptables_manager.apply()方法，应用规则：
iptables-save -c ，获取当前所有iptables信息；
iptables-restore -c ，应用最新的iptables配置；

2.process_router

process_router方法：

1.perform_snat_action，为external gateway处理SNAT规则

    def _handle_router_snat_rules(self, ri, ex_gw_port, internal_cidrs,
                                  interface_name, action):
        # Remove all the rules
        # This is safe because if use_namespaces is set as False
        # then the agent can only configure one router, otherwise
        # each router's SNAT rules will be in their own namespace
        ri.iptables_manager.ipv4['nat'].empty_chain('POSTROUTING')
        ri.iptables_manager.ipv4['nat'].empty_chain('snat')

        # Add back the jump to float-snat
        ri.iptables_manager.ipv4['nat'].add_rule('snat', '-j $float-snat')

        # And add them back if the action if add_rules
        if action == 'add_rules' and ex_gw_port:
            # ex_gw_port should not be None in this case
            # NAT rules are added only if ex_gw_port has an IPv4 address
            for ip_addr in ex_gw_port['fixed_ips']:
                ex_gw_ip = ip_addr['ip_address']
                if netaddr.IPAddress(ex_gw_ip).version == 4:
                    rules = self.external_gateway_nat_rules(ex_gw_ip,
                                                            internal_cidrs,
                                                            interface_name)
                    for rule in rules:
                        ri.iptables_manager.ipv4['nat'].add_rule(*rule)
                    break
        ri.iptables_manager.apply()

 

先清空nat表的neutron-l3-agent-POSTROUTING链和neutron-l3-agent-snat链；

再在nat表的neutron-l3-agent-snat链添加规则：

-A neutron-l3-agent-snat -j neutron-l3-agent-float-snat

然后对应add_rules操作，则处理external_gateway_nat_rules，处理完后在nat表中添加规则：

    def external_gateway_nat_rules(self, ex_gw_ip, internal_cidrs,
                                   interface_name):
        rules = [('POSTROUTING', '! -i %(interface_name)s '
                  '! -o %(interface_name)s -m conntrack ! '
                  '--ctstate DNAT -j ACCEPT' %
                  {'interface_name': interface_name})]
        for cidr in internal_cidrs:
            rules.extend(self.internal_network_nat_rules(ex_gw_ip, cidr))
        return rules

 

规则命令如下：

-A neutron-l3-agent-POSTROUTING ! -i qg-XXX ! -o qg-XXX -m conntrack ! –ctstate DNAT -j ACCEPT

这条命令的意思是除了出口和入口都为qg-XXX，（qg即是router上的外部网关接口）匹配除了DNAT之外的其他状态。

然后处理internal_network_nat_rules：

    def internal_network_nat_rules(self, ex_gw_ip, internal_cidr):
        rules = [('snat', '-s %s -j SNAT --to-source %s' %
                 (internal_cidr, ex_gw_ip))]
        return rules

 

规则命令如下：

-A neutron-l3-agent-snat -s internal_cidr -j SNAT –to-source ex_gw_ip

2.process_router_floating_ip_nat_rules方法，处理floating ip，作SNAT/DNAT转换。

    def process_router_floating_ip_nat_rules(self, ri):
        """Configure NAT rules for the router's floating IPs.

        Configures iptables rules for the floating ips of the given router
        """
        # Clear out all iptables rules for floating ips
        ri.iptables_manager.ipv4['nat'].clear_rules_by_tag('floating_ip')

        # Loop once to ensure that floating ips are configured.
        for fip in ri.router.get(l3_constants.FLOATINGIP_KEY, []):
            # Rebuild iptables rules for the floating ip.
            fixed = fip['fixed_ip_address']
            fip_ip = fip['floating_ip_address']
            for chain, rule in self.floating_forward_rules(fip_ip, fixed):
                ri.iptables_manager.ipv4['nat'].add_rule(chain, rule,
                                                         tag='floating_ip')

        ri.iptables_manager.apply()

   def floating_forward_rules(self, floating_ip, fixed_ip):
        return [('PREROUTING', '-d %s -j DNAT --to %s' %
                 (floating_ip, fixed_ip)),
                ('OUTPUT', '-d %s -j DNAT --to %s' %
                 (floating_ip, fixed_ip)),
                ('float-snat', '-s %s -j SNAT --to %s' %
                 (fixed_ip, floating_ip))]

 

先清理nat表所有的floationg ip规则；然后floating_forward_rules方法,在nat表中处理floating ip和fixed ip的NAT转换：

 

具体规则如下：
-A neutron-l3-agent-PREROUTING -d floating_ip -j DNAT –to fixed_ip
-A neutron-l3-agent-OUTPUT -d floating_ip -j DNAT –to fixed_ip
-A neutron-l3-agent-float-snat -s fixed_ip -j SNAT –to floating_ip

3._router_removed

_router_removed方法，删除和metadata相关的规则：

    def _router_removed(self, router_id):
        ri = self.router_info.get(router_id)
        if ri is None:
            LOG.warn(_("Info for router %s were not found. "
                       "Skipping router removal"), router_id)
            return
        ri.router['gw_port'] = None
        ri.router[l3_constants.INTERFACE_KEY] = []
        ri.router[l3_constants.FLOATINGIP_KEY] = []
        self.process_router(ri)
        for c, r in self.metadata_filter_rules():
            ri.iptables_manager.ipv4['filter'].remove_rule(c, r)
        for c, r in self.metadata_nat_rules():
            ri.iptables_manager.ipv4['nat'].remove_rule(c, r)
        ri.iptables_manager.apply()
        if self.conf.enable_metadata_proxy:
            self._destroy_metadata_proxy(ri.router_id, ri.ns_name)
        del self.router_info[router_id]
        self._destroy_router_namespace(ri.ns_name)

五.总结

l3 agent初始化完成后，iptables处理流程如下：