[NETFILTER]: ip_tables: per-netns FILTER/MANGLE/RAW tables for real
authorAlexey Dobriyan <adobriyan@sw.ru>
Mon, 14 Apr 2008 07:56:02 +0000 (09:56 +0200)
committerPatrick McHardy <kaber@trash.net>
Mon, 14 Apr 2008 07:56:02 +0000 (09:56 +0200)
Commit 9335f047fe61587ec82ff12fbb1220bcfdd32006 aka
"[NETFILTER]: ip_tables: per-netns FILTER, MANGLE, RAW"
added per-netns _view_ of iptables rules. They were shown to user, but
ignored by filtering code. Now that it's possible to at least ping loopback,
per-netns tables can affect filtering decisions.

netns is taken in case of
PRE_ROUTING, LOCAL_IN -- from in device,
POST_ROUTING, LOCAL_OUT -- from out device,
FORWARD -- from in device which should be equal to out device's netns.
   This code is relatively new, so BUG_ON was plugged.

Wrappers were added to a) keep code the same from CONFIG_NET_NS=n users
(overwhelming majority), b) consolidate code in one place -- similar
changes will be done in ipv6 and arp netfilter code.

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Signed-off-by: Patrick McHardy <kaber@trash.net>
include/linux/netfilter.h
net/ipv4/netfilter/iptable_filter.c
net/ipv4/netfilter/iptable_mangle.c
net/ipv4/netfilter/iptable_raw.c

index 89e6c72ad2959411523648fc53ec76f58cb557aa..66bc52060fd63f3bf759505cd5cc923965a11332 100644 (file)
@@ -6,11 +6,13 @@
 #include <linux/types.h>
 #include <linux/skbuff.h>
 #include <linux/net.h>
+#include <linux/netdevice.h>
 #include <linux/if.h>
 #include <linux/in.h>
 #include <linux/in6.h>
 #include <linux/wait.h>
 #include <linux/list.h>
+#include <net/net_namespace.h>
 #endif
 #include <linux/compiler.h>
 
@@ -76,7 +78,6 @@ extern void netfilter_init(void);
 #define NF_MAX_HOOKS 8
 
 struct sk_buff;
-struct net_device;
 
 typedef unsigned int nf_hookfn(unsigned int hooknum,
                               struct sk_buff *skb,
@@ -320,5 +321,56 @@ extern void (*nf_ct_destroy)(struct nf_conntrack *);
 static inline void nf_ct_attach(struct sk_buff *new, struct sk_buff *skb) {}
 #endif
 
+static inline struct net *nf_pre_routing_net(const struct net_device *in,
+                                            const struct net_device *out)
+{
+#ifdef CONFIG_NET_NS
+       return in->nd_net;
+#else
+       return &init_net;
+#endif
+}
+
+static inline struct net *nf_local_in_net(const struct net_device *in,
+                                         const struct net_device *out)
+{
+#ifdef CONFIG_NET_NS
+       return in->nd_net;
+#else
+       return &init_net;
+#endif
+}
+
+static inline struct net *nf_forward_net(const struct net_device *in,
+                                        const struct net_device *out)
+{
+#ifdef CONFIG_NET_NS
+       BUG_ON(in->nd_net != out->nd_net);
+       return in->nd_net;
+#else
+       return &init_net;
+#endif
+}
+
+static inline struct net *nf_local_out_net(const struct net_device *in,
+                                          const struct net_device *out)
+{
+#ifdef CONFIG_NET_NS
+       return out->nd_net;
+#else
+       return &init_net;
+#endif
+}
+
+static inline struct net *nf_post_routing_net(const struct net_device *in,
+                                             const struct net_device *out)
+{
+#ifdef CONFIG_NET_NS
+       return out->nd_net;
+#else
+       return &init_net;
+#endif
+}
+
 #endif /*__KERNEL__*/
 #endif /*__LINUX_NETFILTER_H*/
index 69f3d7e6e96f794a4cecd47c8896e473c3dd59df..7fcf60adbbedea74dbccfb8fd47e4abf659676b5 100644 (file)
@@ -62,6 +62,17 @@ static struct xt_table packet_filter = {
 };
 
 /* The work comes in here from netfilter.c. */
+static unsigned int
+ipt_local_in_hook(unsigned int hook,
+                 struct sk_buff *skb,
+                 const struct net_device *in,
+                 const struct net_device *out,
+                 int (*okfn)(struct sk_buff *))
+{
+       return ipt_do_table(skb, hook, in, out,
+                           nf_local_in_net(in, out)->ipv4.iptable_filter);
+}
+
 static unsigned int
 ipt_hook(unsigned int hook,
         struct sk_buff *skb,
@@ -69,7 +80,8 @@ ipt_hook(unsigned int hook,
         const struct net_device *out,
         int (*okfn)(struct sk_buff *))
 {
-       return ipt_do_table(skb, hook, in, out, init_net.ipv4.iptable_filter);
+       return ipt_do_table(skb, hook, in, out,
+                           nf_forward_net(in, out)->ipv4.iptable_filter);
 }
 
 static unsigned int
@@ -88,12 +100,13 @@ ipt_local_out_hook(unsigned int hook,
                return NF_ACCEPT;
        }
 
-       return ipt_do_table(skb, hook, in, out, init_net.ipv4.iptable_filter);
+       return ipt_do_table(skb, hook, in, out,
+                           nf_local_out_net(in, out)->ipv4.iptable_filter);
 }
 
 static struct nf_hook_ops ipt_ops[] __read_mostly = {
        {
-               .hook           = ipt_hook,
+               .hook           = ipt_local_in_hook,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
                .hooknum        = NF_INET_LOCAL_IN,
index c55a210853a7a5e92d10cecf30dabe645a6355c3..ba827035b691d82eb823099e81069c826c29ffd4 100644 (file)
@@ -74,13 +74,47 @@ static struct xt_table packet_mangler = {
 
 /* The work comes in here from netfilter.c. */
 static unsigned int
-ipt_route_hook(unsigned int hook,
+ipt_pre_routing_hook(unsigned int hook,
+                    struct sk_buff *skb,
+                    const struct net_device *in,
+                    const struct net_device *out,
+                    int (*okfn)(struct sk_buff *))
+{
+       return ipt_do_table(skb, hook, in, out,
+                           nf_pre_routing_net(in, out)->ipv4.iptable_mangle);
+}
+
+static unsigned int
+ipt_post_routing_hook(unsigned int hook,
+                     struct sk_buff *skb,
+                     const struct net_device *in,
+                     const struct net_device *out,
+                     int (*okfn)(struct sk_buff *))
+{
+       return ipt_do_table(skb, hook, in, out,
+                           nf_post_routing_net(in, out)->ipv4.iptable_mangle);
+}
+
+static unsigned int
+ipt_local_in_hook(unsigned int hook,
+                 struct sk_buff *skb,
+                 const struct net_device *in,
+                 const struct net_device *out,
+                 int (*okfn)(struct sk_buff *))
+{
+       return ipt_do_table(skb, hook, in, out,
+                           nf_local_in_net(in, out)->ipv4.iptable_mangle);
+}
+
+static unsigned int
+ipt_forward_hook(unsigned int hook,
         struct sk_buff *skb,
         const struct net_device *in,
         const struct net_device *out,
         int (*okfn)(struct sk_buff *))
 {
-       return ipt_do_table(skb, hook, in, out, init_net.ipv4.iptable_mangle);
+       return ipt_do_table(skb, hook, in, out,
+                           nf_forward_net(in, out)->ipv4.iptable_mangle);
 }
 
 static unsigned int
@@ -112,7 +146,8 @@ ipt_local_hook(unsigned int hook,
        daddr = iph->daddr;
        tos = iph->tos;
 
-       ret = ipt_do_table(skb, hook, in, out, init_net.ipv4.iptable_mangle);
+       ret = ipt_do_table(skb, hook, in, out,
+                          nf_local_out_net(in, out)->ipv4.iptable_mangle);
        /* Reroute for ANY change. */
        if (ret != NF_DROP && ret != NF_STOLEN && ret != NF_QUEUE) {
                iph = ip_hdr(skb);
@@ -130,21 +165,21 @@ ipt_local_hook(unsigned int hook,
 
 static struct nf_hook_ops ipt_ops[] __read_mostly = {
        {
-               .hook           = ipt_route_hook,
+               .hook           = ipt_pre_routing_hook,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
                .hooknum        = NF_INET_PRE_ROUTING,
                .priority       = NF_IP_PRI_MANGLE,
        },
        {
-               .hook           = ipt_route_hook,
+               .hook           = ipt_local_in_hook,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
                .hooknum        = NF_INET_LOCAL_IN,
                .priority       = NF_IP_PRI_MANGLE,
        },
        {
-               .hook           = ipt_route_hook,
+               .hook           = ipt_forward_hook,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
                .hooknum        = NF_INET_FORWARD,
@@ -158,7 +193,7 @@ static struct nf_hook_ops ipt_ops[] __read_mostly = {
                .priority       = NF_IP_PRI_MANGLE,
        },
        {
-               .hook           = ipt_route_hook,
+               .hook           = ipt_post_routing_hook,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
                .hooknum        = NF_INET_POST_ROUTING,
index e41fe8ca4e1c87b9538961411504e1e7777c27b3..4b689742d58b105e5dc0244c42869aa5ac1f7c73 100644 (file)
@@ -52,7 +52,8 @@ ipt_hook(unsigned int hook,
         const struct net_device *out,
         int (*okfn)(struct sk_buff *))
 {
-       return ipt_do_table(skb, hook, in, out, init_net.ipv4.iptable_raw);
+       return ipt_do_table(skb, hook, in, out,
+                           nf_pre_routing_net(in, out)->ipv4.iptable_raw);
 }
 
 static unsigned int
@@ -70,7 +71,8 @@ ipt_local_hook(unsigned int hook,
                               "packet.\n");
                return NF_ACCEPT;
        }
-       return ipt_do_table(skb, hook, in, out, init_net.ipv4.iptable_raw);
+       return ipt_do_table(skb, hook, in, out,
+                           nf_local_out_net(in, out)->ipv4.iptable_raw);
 }
 
 /* 'raw' is the very first table. */