netfilter: factorize ifname_compare()
[safe/jmp/linux-2.6] / net / ipv6 / netfilter / ip6_tables.c
index db0dc96..e59662b 100644 (file)
@@ -28,6 +28,7 @@
 
 #include <linux/netfilter_ipv6/ip6_tables.h>
 #include <linux/netfilter/x_tables.h>
+#include <net/netfilter/nf_log.h>
 
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
@@ -54,7 +55,7 @@ MODULE_DESCRIPTION("IPv6 packet filter");
 do {                                                           \
        if (!(x))                                               \
                printk("IP_NF_ASSERT: %s:%s:%u\n",              \
-                      __FUNCTION__, __FILE__, __LINE__);       \
+                      __func__, __FILE__, __LINE__);   \
 } while(0)
 #else
 #define IP_NF_ASSERT(x)
@@ -89,6 +90,7 @@ ip6t_ext_hdr(u8 nexthdr)
 }
 
 /* Returns whether matches rule or not. */
+/* Performance critical - called for every packet */
 static inline bool
 ip6_packet_match(const struct sk_buff *skb,
                 const char *indev,
@@ -97,11 +99,10 @@ ip6_packet_match(const struct sk_buff *skb,
                 unsigned int *protoff,
                 int *fragoff, bool *hotdrop)
 {
-       size_t i;
        unsigned long ret;
        const struct ipv6hdr *ipv6 = ipv6_hdr(skb);
 
-#define FWINV(bool,invflg) ((bool) ^ !!(ip6info->invflags & invflg))
+#define FWINV(bool, invflg) ((bool) ^ !!(ip6info->invflags & (invflg)))
 
        if (FWINV(ipv6_masked_addr_cmp(&ipv6->saddr, &ip6info->smsk,
                                       &ip6info->src), IP6T_INV_SRCIP)
@@ -118,12 +119,7 @@ ip6_packet_match(const struct sk_buff *skb,
                return false;
        }
 
-       /* Look for ifname matches; this should unroll nicely. */
-       for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
-               ret |= (((const unsigned long *)indev)[i]
-                       ^ ((const unsigned long *)ip6info->iniface)[i])
-                       & ((const unsigned long *)ip6info->iniface_mask)[i];
-       }
+       ret = ifname_compare_aligned(indev, ip6info->iniface, ip6info->iniface_mask);
 
        if (FWINV(ret != 0, IP6T_INV_VIA_IN)) {
                dprintf("VIA in mismatch (%s vs %s).%s\n",
@@ -132,11 +128,7 @@ ip6_packet_match(const struct sk_buff *skb,
                return false;
        }
 
-       for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
-               ret |= (((const unsigned long *)outdev)[i]
-                       ^ ((const unsigned long *)ip6info->outiface)[i])
-                       & ((const unsigned long *)ip6info->outiface_mask)[i];
-       }
+       ret = ifname_compare_aligned(outdev, ip6info->outiface, ip6info->outiface_mask);
 
        if (FWINV(ret != 0, IP6T_INV_VIA_OUT)) {
                dprintf("VIA out mismatch (%s vs %s).%s\n",
@@ -181,7 +173,7 @@ ip6_packet_match(const struct sk_buff *skb,
 }
 
 /* should be ip6 safe */
-static inline bool
+static bool
 ip6_checkentry(const struct ip6t_ip6 *ipv6)
 {
        if (ipv6->flags & ~IP6T_F_MASK) {
@@ -198,31 +190,25 @@ ip6_checkentry(const struct ip6t_ip6 *ipv6)
 }
 
 static unsigned int
-ip6t_error(struct sk_buff *skb,
-         const struct net_device *in,
-         const struct net_device *out,
-         unsigned int hooknum,
-         const struct xt_target *target,
-         const void *targinfo)
+ip6t_error(struct sk_buff *skb, const struct xt_target_param *par)
 {
        if (net_ratelimit())
-               printk("ip6_tables: error: `%s'\n", (char *)targinfo);
+               printk("ip6_tables: error: `%s'\n",
+                      (const char *)par->targinfo);
 
        return NF_DROP;
 }
 
-static inline
-bool do_match(struct ip6t_entry_match *m,
-             const struct sk_buff *skb,
-             const struct net_device *in,
-             const struct net_device *out,
-             int offset,
-             unsigned int protoff,
-             bool *hotdrop)
+/* Performance critical - called for every packet */
+static inline bool
+do_match(struct ip6t_entry_match *m, const struct sk_buff *skb,
+        struct xt_match_param *par)
 {
+       par->match     = m->u.kernel.match;
+       par->matchinfo = m->data;
+
        /* Stop iteration if it doesn't match */
-       if (!m->u.kernel.match->match(skb, in, out, m->u.kernel.match, m->data,
-                                     offset, protoff, hotdrop))
+       if (!m->u.kernel.match->match(skb, par))
                return true;
        else
                return false;
@@ -235,6 +221,7 @@ get_entry(void *base, unsigned int offset)
 }
 
 /* All zeroes == unconditional rule. */
+/* Mildly perf critical (only if packet tracing is on) */
 static inline int
 unconditional(const struct ip6t_ip6 *ipv6)
 {
@@ -250,7 +237,7 @@ unconditional(const struct ip6t_ip6 *ipv6)
 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
 /* This cries for unification! */
-static const char *hooknames[] = {
+static const char *const hooknames[] = {
        [NF_INET_PRE_ROUTING]           = "PREROUTING",
        [NF_INET_LOCAL_IN]              = "INPUT",
        [NF_INET_FORWARD]               = "FORWARD",
@@ -264,7 +251,7 @@ enum nf_ip_trace_comments {
        NF_IP6_TRACE_COMMENT_POLICY,
 };
 
-static const char *comments[] = {
+static const char *const comments[] = {
        [NF_IP6_TRACE_COMMENT_RULE]     = "rule",
        [NF_IP6_TRACE_COMMENT_RETURN]   = "return",
        [NF_IP6_TRACE_COMMENT_POLICY]   = "policy",
@@ -280,6 +267,7 @@ static struct nf_loginfo trace_loginfo = {
        },
 };
 
+/* Mildly perf critical (only if packet tracing is on) */
 static inline int
 get_chainname_rulenum(struct ip6t_entry *s, struct ip6t_entry *e,
                      char *hookname, char **chainname,
@@ -315,12 +303,12 @@ static void trace_packet(struct sk_buff *skb,
                         unsigned int hook,
                         const struct net_device *in,
                         const struct net_device *out,
-                        char *tablename,
+                        const char *tablename,
                         struct xt_table_info *private,
                         struct ip6t_entry *e)
 {
        void *table_base;
-       struct ip6t_entry *root;
+       const struct ip6t_entry *root;
        char *hookname, *chainname, *comment;
        unsigned int rulenum = 0;
 
@@ -350,8 +338,6 @@ ip6t_do_table(struct sk_buff *skb,
              struct xt_table *table)
 {
        static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
-       int offset = 0;
-       unsigned int protoff = 0;
        bool hotdrop = false;
        /* Initializing verdict to NF_DROP keeps gcc happy. */
        unsigned int verdict = NF_DROP;
@@ -359,6 +345,8 @@ ip6t_do_table(struct sk_buff *skb,
        void *table_base;
        struct ip6t_entry *e, *back;
        struct xt_table_info *private;
+       struct xt_match_param mtpar;
+       struct xt_target_param tgpar;
 
        /* Initialization */
        indev = in ? in->name : nulldevname;
@@ -369,11 +357,18 @@ ip6t_do_table(struct sk_buff *skb,
         * things we don't know, ie. tcp syn flag or ports).  If the
         * rule is also a fragment-specific rule, non-fragments won't
         * match it. */
+       mtpar.hotdrop = &hotdrop;
+       mtpar.in      = tgpar.in  = in;
+       mtpar.out     = tgpar.out = out;
+       mtpar.family  = tgpar.family = NFPROTO_IPV6;
+       tgpar.hooknum = hook;
 
-       read_lock_bh(&table->lock);
-       private = table->private;
        IP_NF_ASSERT(table->valid_hooks & (1 << hook));
-       table_base = (void *)private->entries[smp_processor_id()];
+
+       rcu_read_lock();
+       private = rcu_dereference(table->private);
+       table_base = rcu_dereference(private->entries[smp_processor_id()]);
+
        e = get_entry(table_base, private->hook_entry[hook]);
 
        /* For return from builtin chain */
@@ -383,12 +378,10 @@ ip6t_do_table(struct sk_buff *skb,
                IP_NF_ASSERT(e);
                IP_NF_ASSERT(back);
                if (ip6_packet_match(skb, indev, outdev, &e->ipv6,
-                       &protoff, &offset, &hotdrop)) {
+                       &mtpar.thoff, &mtpar.fragoff, &hotdrop)) {
                        struct ip6t_entry_target *t;
 
-                       if (IP6T_MATCH_ITERATE(e, do_match,
-                                              skb, in, out,
-                                              offset, protoff, &hotdrop) != 0)
+                       if (IP6T_MATCH_ITERATE(e, do_match, skb, &mtpar) != 0)
                                goto no_match;
 
                        ADD_COUNTER(e->counters,
@@ -436,15 +429,15 @@ ip6t_do_table(struct sk_buff *skb,
                        } else {
                                /* Targets which reenter must return
                                   abs. verdicts */
+                               tgpar.target   = t->u.kernel.target;
+                               tgpar.targinfo = t->data;
+
 #ifdef CONFIG_NETFILTER_DEBUG
                                ((struct ip6t_entry *)table_base)->comefrom
                                        = 0xeeeeeeec;
 #endif
                                verdict = t->u.kernel.target->target(skb,
-                                                                    in, out,
-                                                                    hook,
-                                                                    t->u.kernel.target,
-                                                                    t->data);
+                                                                    &tgpar);
 
 #ifdef CONFIG_NETFILTER_DEBUG
                                if (((struct ip6t_entry *)table_base)->comefrom
@@ -473,7 +466,7 @@ ip6t_do_table(struct sk_buff *skb,
 #ifdef CONFIG_NETFILTER_DEBUG
        ((struct ip6t_entry *)table_base)->comefrom = NETFILTER_LINK_POISON;
 #endif
-       read_unlock_bh(&table->lock);
+       rcu_read_unlock();
 
 #ifdef DEBUG_ALLOW_ALL
        return NF_ACCEPT;
@@ -496,9 +489,7 @@ mark_source_chains(struct xt_table_info *newinfo,
           to 0 as we leave), and comefrom to save source hook bitmask */
        for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
                unsigned int pos = newinfo->hook_entry[hook];
-               struct ip6t_entry *e
-                       = (struct ip6t_entry *)(entry0 + pos);
-               int visited = e->comefrom & (1 << hook);
+               struct ip6t_entry *e = (struct ip6t_entry *)(entry0 + pos);
 
                if (!(valid_hooks & (1 << hook)))
                        continue;
@@ -509,14 +500,14 @@ mark_source_chains(struct xt_table_info *newinfo,
                for (;;) {
                        struct ip6t_standard_target *t
                                = (void *)ip6t_get_target(e);
+                       int visited = e->comefrom & (1 << hook);
 
                        if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
                                printk("iptables: loop hook %u pos %u %08X.\n",
                                       hook, pos, e->comefrom);
                                return 0;
                        }
-                       e->comefrom
-                               |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
+                       e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 
                        /* Unconditional return/END. */
                        if ((e->target_offset == sizeof(struct ip6t_entry)
@@ -596,19 +587,24 @@ mark_source_chains(struct xt_table_info *newinfo,
        return 1;
 }
 
-static inline int
+static int
 cleanup_match(struct ip6t_entry_match *m, unsigned int *i)
 {
+       struct xt_mtdtor_param par;
+
        if (i && (*i)-- == 0)
                return 1;
 
-       if (m->u.kernel.match->destroy)
-               m->u.kernel.match->destroy(m->u.kernel.match, m->data);
-       module_put(m->u.kernel.match->me);
+       par.match     = m->u.kernel.match;
+       par.matchinfo = m->data;
+       par.family    = NFPROTO_IPV6;
+       if (par.match->destroy != NULL)
+               par.match->destroy(&par);
+       module_put(par.match->me);
        return 0;
 }
 
-static inline int
+static int
 check_entry(struct ip6t_entry *e, const char *name)
 {
        struct ip6t_entry_target *t;
@@ -629,41 +625,35 @@ check_entry(struct ip6t_entry *e, const char *name)
        return 0;
 }
 
-static inline int check_match(struct ip6t_entry_match *m, const char *name,
-                             const struct ip6t_ip6 *ipv6,
-                             unsigned int hookmask, unsigned int *i)
+static int check_match(struct ip6t_entry_match *m, struct xt_mtchk_param *par,
+                      unsigned int *i)
 {
-       struct xt_match *match;
+       const struct ip6t_ip6 *ipv6 = par->entryinfo;
        int ret;
 
-       match = m->u.kernel.match;
-       ret = xt_check_match(match, AF_INET6, m->u.match_size - sizeof(*m),
-                            name, hookmask, ipv6->proto,
-                            ipv6->invflags & IP6T_INV_PROTO);
-       if (!ret && m->u.kernel.match->checkentry
-           && !m->u.kernel.match->checkentry(name, ipv6, match, m->data,
-                                             hookmask)) {
+       par->match     = m->u.kernel.match;
+       par->matchinfo = m->data;
+
+       ret = xt_check_match(par, m->u.match_size - sizeof(*m),
+                            ipv6->proto, ipv6->invflags & IP6T_INV_PROTO);
+       if (ret < 0) {
                duprintf("ip_tables: check failed for `%s'.\n",
-                        m->u.kernel.match->name);
-               ret = -EINVAL;
+                        par.match->name);
+               return ret;
        }
-       if (!ret)
-               (*i)++;
-       return ret;
+       ++*i;
+       return 0;
 }
 
-static inline int
-find_check_match(struct ip6t_entry_match *m,
-                const char *name,
-                const struct ip6t_ip6 *ipv6,
-                unsigned int hookmask,
+static int
+find_check_match(struct ip6t_entry_match *m, struct xt_mtchk_param *par,
                 unsigned int *i)
 {
        struct xt_match *match;
        int ret;
 
        match = try_then_request_module(xt_find_match(AF_INET6, m->u.user.name,
-                                       m->u.user.revision),
+                                                     m->u.user.revision),
                                        "ip6t_%s", m->u.user.name);
        if (IS_ERR(match) || !match) {
                duprintf("find_check_match: `%s' not found\n", m->u.user.name);
@@ -671,7 +661,7 @@ find_check_match(struct ip6t_entry_match *m,
        }
        m->u.kernel.match = match;
 
-       ret = check_match(m, name, ipv6, hookmask, i);
+       ret = check_match(m, par, i);
        if (ret)
                goto err;
 
@@ -681,28 +671,31 @@ err:
        return ret;
 }
 
-static inline int check_target(struct ip6t_entry *e, const char *name)
+static int check_target(struct ip6t_entry *e, const char *name)
 {
-       struct ip6t_entry_target *t;
-       struct xt_target *target;
+       struct ip6t_entry_target *t = ip6t_get_target(e);
+       struct xt_tgchk_param par = {
+               .table     = name,
+               .entryinfo = e,
+               .target    = t->u.kernel.target,
+               .targinfo  = t->data,
+               .hook_mask = e->comefrom,
+               .family    = NFPROTO_IPV6,
+       };
        int ret;
 
        t = ip6t_get_target(e);
-       target = t->u.kernel.target;
-       ret = xt_check_target(target, AF_INET6, t->u.target_size - sizeof(*t),
-                             name, e->comefrom, e->ipv6.proto,
-                             e->ipv6.invflags & IP6T_INV_PROTO);
-       if (!ret && t->u.kernel.target->checkentry
-           && !t->u.kernel.target->checkentry(name, e, target, t->data,
-                                              e->comefrom)) {
+       ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
+             e->ipv6.proto, e->ipv6.invflags & IP6T_INV_PROTO);
+       if (ret < 0) {
                duprintf("ip_tables: check failed for `%s'.\n",
                         t->u.kernel.target->name);
-               ret = -EINVAL;
+               return ret;
        }
-       return ret;
+       return 0;
 }
 
-static inline int
+static int
 find_check_entry(struct ip6t_entry *e, const char *name, unsigned int size,
                 unsigned int *i)
 {
@@ -710,14 +703,18 @@ find_check_entry(struct ip6t_entry *e, const char *name, unsigned int size,
        struct xt_target *target;
        int ret;
        unsigned int j;
+       struct xt_mtchk_param mtpar;
 
        ret = check_entry(e, name);
        if (ret)
                return ret;
 
        j = 0;
-       ret = IP6T_MATCH_ITERATE(e, find_check_match, name, &e->ipv6,
-                                e->comefrom, &j);
+       mtpar.table     = name;
+       mtpar.entryinfo = &e->ipv6;
+       mtpar.hook_mask = e->comefrom;
+       mtpar.family    = NFPROTO_IPV6;
+       ret = IP6T_MATCH_ITERATE(e, find_check_match, &mtpar, &j);
        if (ret != 0)
                goto cleanup_matches;
 
@@ -746,7 +743,7 @@ find_check_entry(struct ip6t_entry *e, const char *name, unsigned int size,
        return ret;
 }
 
-static inline int
+static int
 check_entry_size_and_hooks(struct ip6t_entry *e,
                           struct xt_table_info *newinfo,
                           unsigned char *base,
@@ -789,9 +786,10 @@ check_entry_size_and_hooks(struct ip6t_entry *e,
        return 0;
 }
 
-static inline int
+static int
 cleanup_entry(struct ip6t_entry *e, unsigned int *i)
 {
+       struct xt_tgdtor_param par;
        struct ip6t_entry_target *t;
 
        if (i && (*i)-- == 0)
@@ -800,9 +798,13 @@ cleanup_entry(struct ip6t_entry *e, unsigned int *i)
        /* Cleanup all matches */
        IP6T_MATCH_ITERATE(e, cleanup_match, NULL);
        t = ip6t_get_target(e);
-       if (t->u.kernel.target->destroy)
-               t->u.kernel.target->destroy(t->u.kernel.target, t->data);
-       module_put(t->u.kernel.target->me);
+
+       par.target   = t->u.kernel.target;
+       par.targinfo = t->data;
+       par.family   = NFPROTO_IPV6;
+       if (par.target->destroy != NULL)
+               par.target->destroy(&par);
+       module_put(par.target->me);
        return 0;
 }
 
@@ -885,7 +887,7 @@ translate_table(const char *name,
                        memcpy(newinfo->entries[i], entry0, newinfo->size);
        }
 
-       return 0;
+       return ret;
 }
 
 /* Gets counters. */
@@ -945,11 +947,64 @@ get_counters(const struct xt_table_info *t,
        }
 }
 
-static inline struct xt_counters *alloc_counters(struct xt_table *table)
+/* We're lazy, and add to the first CPU; overflow works its fey magic
+ * and everything is OK. */
+static int
+add_counter_to_entry(struct ip6t_entry *e,
+                    const struct xt_counters addme[],
+                    unsigned int *i)
+{
+       ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
+
+       (*i)++;
+       return 0;
+}
+
+/* Take values from counters and add them back onto the current cpu */
+static void put_counters(struct xt_table_info *t,
+                        const struct xt_counters counters[])
+{
+       unsigned int i, cpu;
+
+       local_bh_disable();
+       cpu = smp_processor_id();
+       i = 0;
+       IP6T_ENTRY_ITERATE(t->entries[cpu],
+                          t->size,
+                          add_counter_to_entry,
+                          counters,
+                          &i);
+       local_bh_enable();
+}
+
+static inline int
+zero_entry_counter(struct ip6t_entry *e, void *arg)
+{
+       e->counters.bcnt = 0;
+       e->counters.pcnt = 0;
+       return 0;
+}
+
+static void
+clone_counters(struct xt_table_info *newinfo, const struct xt_table_info *info)
+{
+       unsigned int cpu;
+       const void *loc_cpu_entry = info->entries[raw_smp_processor_id()];
+
+       memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
+       for_each_possible_cpu(cpu) {
+               memcpy(newinfo->entries[cpu], loc_cpu_entry, info->size);
+               IP6T_ENTRY_ITERATE(newinfo->entries[cpu], newinfo->size,
+                                  zero_entry_counter, NULL);
+       }
+}
+
+static struct xt_counters *alloc_counters(struct xt_table *table)
 {
        unsigned int countersize;
        struct xt_counters *counters;
        struct xt_table_info *private = table->private;
+       struct xt_table_info *info;
 
        /* We need atomic snapshot of counters: rest doesn't change
           (other than comefrom, which userspace doesn't care
@@ -958,14 +1013,28 @@ static inline struct xt_counters *alloc_counters(struct xt_table *table)
        counters = vmalloc_node(countersize, numa_node_id());
 
        if (counters == NULL)
-               return ERR_PTR(-ENOMEM);
+               goto nomem;
 
-       /* First, sum counters... */
-       write_lock_bh(&table->lock);
-       get_counters(private, counters);
-       write_unlock_bh(&table->lock);
+       info = xt_alloc_table_info(private->size);
+       if (!info)
+               goto free_counters;
 
-       return counters;
+       clone_counters(info, private);
+
+       mutex_lock(&table->lock);
+       xt_table_entry_swap_rcu(private, info);
+       synchronize_net();      /* Wait until smoke has cleared */
+
+       get_counters(info, counters);
+       put_counters(private, counters);
+       mutex_unlock(&table->lock);
+
+       xt_free_table_info(info);
+
+ free_counters:
+       vfree(counters);
+ nomem:
+       return ERR_PTR(-ENOMEM);
 }
 
 static int
@@ -976,15 +1045,18 @@ copy_entries_to_user(unsigned int total_size,
        unsigned int off, num;
        struct ip6t_entry *e;
        struct xt_counters *counters;
-       struct xt_table_info *private = table->private;
+       const struct xt_table_info *private = table->private;
        int ret = 0;
-       void *loc_cpu_entry;
+       const void *loc_cpu_entry;
 
        counters = alloc_counters(table);
        if (IS_ERR(counters))
                return PTR_ERR(counters);
 
-       /* choose the copy that is on ourc node/cpu */
+       /* choose the copy that is on our node/cpu, ...
+        * This choice is lazy (because current thread is
+        * allowed to migrate to another cpu)
+        */
        loc_cpu_entry = private->entries[raw_smp_processor_id()];
        if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
                ret = -EFAULT;
@@ -995,8 +1067,8 @@ copy_entries_to_user(unsigned int total_size,
        /* ... then go back and fix counters and names */
        for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
                unsigned int i;
-               struct ip6t_entry_match *m;
-               struct ip6t_entry_target *t;
+               const struct ip6t_entry_match *m;
+               const struct ip6t_entry_target *t;
 
                e = (struct ip6t_entry *)(loc_cpu_entry + off);
                if (copy_to_user(userptr + off
@@ -1112,14 +1184,14 @@ static int compat_table_info(const struct xt_table_info *info,
 }
 #endif
 
-static int get_info(void __user *user, int *len, int compat)
+static int get_info(struct net *net, void __user *user, int *len, int compat)
 {
        char name[IP6T_TABLE_MAXNAMELEN];
        struct xt_table *t;
        int ret;
 
        if (*len != sizeof(struct ip6t_getinfo)) {
-               duprintf("length %u != %u\n", *len,
+               duprintf("length %u != %zu\n", *len,
                         sizeof(struct ip6t_getinfo));
                return -EINVAL;
        }
@@ -1132,11 +1204,11 @@ static int get_info(void __user *user, int *len, int compat)
        if (compat)
                xt_compat_lock(AF_INET6);
 #endif
-       t = try_then_request_module(xt_find_table_lock(AF_INET6, name),
+       t = try_then_request_module(xt_find_table_lock(net, AF_INET6, name),
                                    "ip6table_%s", name);
        if (t && !IS_ERR(t)) {
                struct ip6t_getinfo info;
-               struct xt_table_info *private = t->private;
+               const struct xt_table_info *private = t->private;
 
 #ifdef CONFIG_COMPAT
                if (compat) {
@@ -1153,7 +1225,7 @@ static int get_info(void __user *user, int *len, int compat)
                       sizeof(info.underflow));
                info.num_entries = private->number;
                info.size = private->size;
-               memcpy(info.name, name, sizeof(info.name));
+               strcpy(info.name, name);
 
                if (copy_to_user(user, &info, *len) != 0)
                        ret = -EFAULT;
@@ -1172,25 +1244,25 @@ static int get_info(void __user *user, int *len, int compat)
 }
 
 static int
-get_entries(struct ip6t_get_entries __user *uptr, int *len)
+get_entries(struct net *net, struct ip6t_get_entries __user *uptr, int *len)
 {
        int ret;
        struct ip6t_get_entries get;
        struct xt_table *t;
 
        if (*len < sizeof(get)) {
-               duprintf("get_entries: %u < %u\n", *len, sizeof(get));
+               duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
                return -EINVAL;
        }
        if (copy_from_user(&get, uptr, sizeof(get)) != 0)
                return -EFAULT;
        if (*len != sizeof(struct ip6t_get_entries) + get.size) {
-               duprintf("get_entries: %u != %u\n", *len,
-                        sizeof(struct ip6t_get_entries) + get.size);
+               duprintf("get_entries: %u != %zu\n",
+                        *len, sizeof(get) + get.size);
                return -EINVAL;
        }
 
-       t = xt_find_table_lock(AF_INET6, get.name);
+       t = xt_find_table_lock(net, AF_INET6, get.name);
        if (t && !IS_ERR(t)) {
                struct xt_table_info *private = t->private;
                duprintf("t->private->number = %u\n", private->number);
@@ -1199,8 +1271,8 @@ get_entries(struct ip6t_get_entries __user *uptr, int *len)
                                                   t, uptr->entrytable);
                else {
                        duprintf("get_entries: I've got %u not %u!\n",
-                                private->size, entries->size);
-                       ret = -EINVAL;
+                                private->size, get.size);
+                       ret = -EAGAIN;
                }
                module_put(t->me);
                xt_table_unlock(t);
@@ -1211,7 +1283,7 @@ get_entries(struct ip6t_get_entries __user *uptr, int *len)
 }
 
 static int
-__do_replace(const char *name, unsigned int valid_hooks,
+__do_replace(struct net *net, const char *name, unsigned int valid_hooks,
             struct xt_table_info *newinfo, unsigned int num_counters,
             void __user *counters_ptr)
 {
@@ -1219,7 +1291,7 @@ __do_replace(const char *name, unsigned int valid_hooks,
        struct xt_table *t;
        struct xt_table_info *oldinfo;
        struct xt_counters *counters;
-       void *loc_cpu_old_entry;
+       const void *loc_cpu_old_entry;
 
        ret = 0;
        counters = vmalloc_node(num_counters * sizeof(struct xt_counters),
@@ -1229,7 +1301,7 @@ __do_replace(const char *name, unsigned int valid_hooks,
                goto out;
        }
 
-       t = try_then_request_module(xt_find_table_lock(AF_INET6, name),
+       t = try_then_request_module(xt_find_table_lock(net, AF_INET6, name),
                                    "ip6table_%s", name);
        if (!t || IS_ERR(t)) {
                ret = t ? PTR_ERR(t) : -ENOENT;
@@ -1282,7 +1354,7 @@ __do_replace(const char *name, unsigned int valid_hooks,
 }
 
 static int
-do_replace(void __user *user, unsigned int len)
+do_replace(struct net *net, void __user *user, unsigned int len)
 {
        int ret;
        struct ip6t_replace tmp;
@@ -1316,7 +1388,7 @@ do_replace(void __user *user, unsigned int len)
 
        duprintf("ip_tables: Translated table\n");
 
-       ret = __do_replace(tmp.name, tmp.valid_hooks, newinfo,
+       ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
                           tmp.num_counters, tmp.counters);
        if (ret)
                goto free_newinfo_untrans;
@@ -1329,30 +1401,9 @@ do_replace(void __user *user, unsigned int len)
        return ret;
 }
 
-/* We're lazy, and add to the first CPU; overflow works its fey magic
- * and everything is OK. */
-static inline int
-add_counter_to_entry(struct ip6t_entry *e,
-                    const struct xt_counters addme[],
-                    unsigned int *i)
-{
-#if 0
-       duprintf("add_counter: Entry %u %lu/%lu + %lu/%lu\n",
-                *i,
-                (long unsigned int)e->counters.pcnt,
-                (long unsigned int)e->counters.bcnt,
-                (long unsigned int)addme[*i].pcnt,
-                (long unsigned int)addme[*i].bcnt);
-#endif
-
-       ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
-
-       (*i)++;
-       return 0;
-}
-
 static int
-do_add_counters(void __user *user, unsigned int len, int compat)
+do_add_counters(struct net *net, void __user *user, unsigned int len,
+               int compat)
 {
        unsigned int i;
        struct xt_counters_info tmp;
@@ -1361,10 +1412,10 @@ do_add_counters(void __user *user, unsigned int len, int compat)
        char *name;
        int size;
        void *ptmp;
-       struct xt_table_info *private;
        struct xt_table *t;
+       const struct xt_table_info *private;
        int ret = 0;
-       void *loc_cpu_entry;
+       const void *loc_cpu_entry;
 #ifdef CONFIG_COMPAT
        struct compat_xt_counters_info compat_tmp;
 
@@ -1404,29 +1455,31 @@ do_add_counters(void __user *user, unsigned int len, int compat)
                goto free;
        }
 
-       t = xt_find_table_lock(AF_INET6, name);
+       t = xt_find_table_lock(net, AF_INET6, name);
        if (!t || IS_ERR(t)) {
                ret = t ? PTR_ERR(t) : -ENOENT;
                goto free;
        }
 
-       write_lock_bh(&t->lock);
+       mutex_lock(&t->lock);
        private = t->private;
        if (private->number != num_counters) {
                ret = -EINVAL;
                goto unlock_up_free;
        }
 
+       preempt_disable();
        i = 0;
        /* Choose the copy that is on our node */
-       loc_cpu_entry = private->entries[smp_processor_id()];
+       loc_cpu_entry = private->entries[raw_smp_processor_id()];
        IP6T_ENTRY_ITERATE(loc_cpu_entry,
                          private->size,
                          add_counter_to_entry,
                          paddc,
                          &i);
+       preempt_enable();
  unlock_up_free:
-       write_unlock_bh(&t->lock);
+       mutex_unlock(&t->lock);
        xt_table_unlock(t);
        module_put(t->me);
  free:
@@ -1450,7 +1503,7 @@ struct compat_ip6t_replace {
 
 static int
 compat_copy_entry_to_user(struct ip6t_entry *e, void __user **dstptr,
-                         compat_uint_t *size, struct xt_counters *counters,
+                         unsigned int *size, struct xt_counters *counters,
                          unsigned int *i)
 {
        struct ip6t_entry_target *t;
@@ -1492,12 +1545,12 @@ out:
        return ret;
 }
 
-static inline int
+static int
 compat_find_calc_match(struct ip6t_entry_match *m,
                       const char *name,
                       const struct ip6t_ip6 *ipv6,
                       unsigned int hookmask,
-                      int *size, int *i)
+                      int *size, unsigned int *i)
 {
        struct xt_match *match;
 
@@ -1516,7 +1569,7 @@ compat_find_calc_match(struct ip6t_entry_match *m,
        return 0;
 }
 
-static inline int
+static int
 compat_release_match(struct ip6t_entry_match *m, unsigned int *i)
 {
        if (i && (*i)-- == 0)
@@ -1526,7 +1579,7 @@ compat_release_match(struct ip6t_entry_match *m, unsigned int *i)
        return 0;
 }
 
-static inline int
+static int
 compat_release_entry(struct compat_ip6t_entry *e, unsigned int *i)
 {
        struct ip6t_entry_target *t;
@@ -1541,7 +1594,7 @@ compat_release_entry(struct compat_ip6t_entry *e, unsigned int *i)
        return 0;
 }
 
-static inline int
+static int
 check_compat_entry_size_and_hooks(struct compat_ip6t_entry *e,
                                  struct xt_table_info *newinfo,
                                  unsigned int *size,
@@ -1555,7 +1608,8 @@ check_compat_entry_size_and_hooks(struct compat_ip6t_entry *e,
        struct ip6t_entry_target *t;
        struct xt_target *target;
        unsigned int entry_offset;
-       int ret, off, h, j;
+       unsigned int j;
+       int ret, off, h;
 
        duprintf("check_compat_entry_size_and_hooks %p\n", e);
        if ((unsigned long)e % __alignof__(struct compat_ip6t_entry) != 0
@@ -1664,14 +1718,19 @@ compat_copy_entry_from_user(struct compat_ip6t_entry *e, void **dstptr,
        return ret;
 }
 
-static inline int compat_check_entry(struct ip6t_entry *e, const char *name,
+static int compat_check_entry(struct ip6t_entry *e, const char *name,
                                     unsigned int *i)
 {
-       int j, ret;
+       unsigned int j;
+       int ret;
+       struct xt_mtchk_param mtpar;
 
        j = 0;
-       ret = IP6T_MATCH_ITERATE(e, check_match, name, &e->ipv6,
-                                e->comefrom, &j);
+       mtpar.table     = name;
+       mtpar.entryinfo = &e->ipv6;
+       mtpar.hook_mask = e->comefrom;
+       mtpar.family    = NFPROTO_IPV6;
+       ret = IP6T_MATCH_ITERATE(e, check_match, &mtpar, &j);
        if (ret)
                goto cleanup_matches;
 
@@ -1809,7 +1868,7 @@ out_unlock:
 }
 
 static int
-compat_do_replace(void __user *user, unsigned int len)
+compat_do_replace(struct net *net, void __user *user, unsigned int len)
 {
        int ret;
        struct compat_ip6t_replace tmp;
@@ -1829,7 +1888,7 @@ compat_do_replace(void __user *user, unsigned int len)
        if (!newinfo)
                return -ENOMEM;
 
-       /* choose the copy that is our node/cpu */
+       /* choose the copy that is on our node/cpu */
        loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
        if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
                           tmp.size) != 0) {
@@ -1846,7 +1905,7 @@ compat_do_replace(void __user *user, unsigned int len)
 
        duprintf("compat_do_replace: Translated table\n");
 
-       ret = __do_replace(tmp.name, tmp.valid_hooks, newinfo,
+       ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
                           tmp.num_counters, compat_ptr(tmp.counters));
        if (ret)
                goto free_newinfo_untrans;
@@ -1870,11 +1929,11 @@ compat_do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user,
 
        switch (cmd) {
        case IP6T_SO_SET_REPLACE:
-               ret = compat_do_replace(user, len);
+               ret = compat_do_replace(sock_net(sk), user, len);
                break;
 
        case IP6T_SO_SET_ADD_COUNTERS:
-               ret = do_add_counters(user, len, 1);
+               ret = do_add_counters(sock_net(sk), user, len, 1);
                break;
 
        default:
@@ -1896,11 +1955,11 @@ compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
                            void __user *userptr)
 {
        struct xt_counters *counters;
-       struct xt_table_info *private = table->private;
+       const struct xt_table_info *private = table->private;
        void __user *pos;
        unsigned int size;
        int ret = 0;
-       void *loc_cpu_entry;
+       const void *loc_cpu_entry;
        unsigned int i = 0;
 
        counters = alloc_counters(table);
@@ -1923,15 +1982,15 @@ compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
 }
 
 static int
-compat_get_entries(struct compat_ip6t_get_entries __user *uptr, int *len)
+compat_get_entries(struct net *net, struct compat_ip6t_get_entries __user *uptr,
+                  int *len)
 {
        int ret;
        struct compat_ip6t_get_entries get;
        struct xt_table *t;
 
        if (*len < sizeof(get)) {
-               duprintf("compat_get_entries: %u < %u\n",
-                        *len, (unsigned int)sizeof(get));
+               duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
                return -EINVAL;
        }
 
@@ -1939,28 +1998,25 @@ compat_get_entries(struct compat_ip6t_get_entries __user *uptr, int *len)
                return -EFAULT;
 
        if (*len != sizeof(struct compat_ip6t_get_entries) + get.size) {
-               duprintf("compat_get_entries: %u != %u\n", *len,
-                        (unsigned int)(sizeof(struct compat_ip6t_get_entries) +
-                        get.size));
+               duprintf("compat_get_entries: %u != %zu\n",
+                        *len, sizeof(get) + get.size);
                return -EINVAL;
        }
 
        xt_compat_lock(AF_INET6);
-       t = xt_find_table_lock(AF_INET6, get.name);
+       t = xt_find_table_lock(net, AF_INET6, get.name);
        if (t && !IS_ERR(t)) {
-               struct xt_table_info *private = t->private;
+               const struct xt_table_info *private = t->private;
                struct xt_table_info info;
-               duprintf("t->private->number = %u\n",
-                        private->number);
+               duprintf("t->private->number = %u\n", private->number);
                ret = compat_table_info(private, &info);
                if (!ret && get.size == info.size) {
                        ret = compat_copy_entries_to_user(private->size,
                                                          t, uptr->entrytable);
                } else if (!ret) {
                        duprintf("compat_get_entries: I've got %u not %u!\n",
-                                private->size,
-                                get.size);
-                       ret = -EINVAL;
+                                private->size, get.size);
+                       ret = -EAGAIN;
                }
                xt_compat_flush_offsets(AF_INET6);
                module_put(t->me);
@@ -1984,10 +2040,10 @@ compat_do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
 
        switch (cmd) {
        case IP6T_SO_GET_INFO:
-               ret = get_info(user, len, 1);
+               ret = get_info(sock_net(sk), user, len, 1);
                break;
        case IP6T_SO_GET_ENTRIES:
-               ret = compat_get_entries(user, len);
+               ret = compat_get_entries(sock_net(sk), user, len);
                break;
        default:
                ret = do_ip6t_get_ctl(sk, cmd, user, len);
@@ -2006,11 +2062,11 @@ do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
 
        switch (cmd) {
        case IP6T_SO_SET_REPLACE:
-               ret = do_replace(user, len);
+               ret = do_replace(sock_net(sk), user, len);
                break;
 
        case IP6T_SO_SET_ADD_COUNTERS:
-               ret = do_add_counters(user, len, 0);
+               ret = do_add_counters(sock_net(sk), user, len, 0);
                break;
 
        default:
@@ -2031,11 +2087,11 @@ do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
 
        switch (cmd) {
        case IP6T_SO_GET_INFO:
-               ret = get_info(user, len, 0);
+               ret = get_info(sock_net(sk), user, len, 0);
                break;
 
        case IP6T_SO_GET_ENTRIES:
-               ret = get_entries(user, len);
+               ret = get_entries(sock_net(sk), user, len);
                break;
 
        case IP6T_SO_GET_REVISION_MATCH:
@@ -2072,20 +2128,23 @@ do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
        return ret;
 }
 
-int ip6t_register_table(struct xt_table *table,
-                       const struct ip6t_replace *repl)
+struct xt_table *ip6t_register_table(struct net *net, struct xt_table *table,
+                                    const struct ip6t_replace *repl)
 {
        int ret;
        struct xt_table_info *newinfo;
        struct xt_table_info bootstrap
                = { 0, 0, 0, { 0 }, { 0 }, { } };
        void *loc_cpu_entry;
+       struct xt_table *new_table;
 
        newinfo = xt_alloc_table_info(repl->size);
-       if (!newinfo)
-               return -ENOMEM;
+       if (!newinfo) {
+               ret = -ENOMEM;
+               goto out;
+       }
 
-       /* choose the copy on our node/cpu */
+       /* choose the copy on our node/cpu, but dont care about preemption */
        loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
        memcpy(loc_cpu_entry, repl->entries, repl->size);
 
@@ -2094,30 +2153,35 @@ int ip6t_register_table(struct xt_table *table,
                              repl->num_entries,
                              repl->hook_entry,
                              repl->underflow);
-       if (ret != 0) {
-               xt_free_table_info(newinfo);
-               return ret;
-       }
+       if (ret != 0)
+               goto out_free;
 
-       ret = xt_register_table(table, &bootstrap, newinfo);
-       if (ret != 0) {
-               xt_free_table_info(newinfo);
-               return ret;
+       new_table = xt_register_table(net, table, &bootstrap, newinfo);
+       if (IS_ERR(new_table)) {
+               ret = PTR_ERR(new_table);
+               goto out_free;
        }
+       return new_table;
 
-       return 0;
+out_free:
+       xt_free_table_info(newinfo);
+out:
+       return ERR_PTR(ret);
 }
 
 void ip6t_unregister_table(struct xt_table *table)
 {
        struct xt_table_info *private;
        void *loc_cpu_entry;
+       struct module *table_owner = table->me;
 
        private = xt_unregister_table(table);
 
        /* Decrease module usage counts and free resources */
        loc_cpu_entry = private->entries[raw_smp_processor_id()];
        IP6T_ENTRY_ITERATE(loc_cpu_entry, private->size, cleanup_entry, NULL);
+       if (private->number > private->initial_entries)
+               module_put(table_owner);
        xt_free_table_info(private);
 }
 
@@ -2132,28 +2196,23 @@ icmp6_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
 }
 
 static bool
-icmp6_match(const struct sk_buff *skb,
-          const struct net_device *in,
-          const struct net_device *out,
-          const struct xt_match *match,
-          const void *matchinfo,
-          int offset,
-          unsigned int protoff,
-          bool *hotdrop)
-{
-       struct icmp6hdr _icmp, *ic;
-       const struct ip6t_icmp *icmpinfo = matchinfo;
+icmp6_match(const struct sk_buff *skb, const struct xt_match_param *par)
+{
+       const struct icmp6hdr *ic;
+       struct icmp6hdr _icmph;
+       const struct ip6t_icmp *icmpinfo = par->matchinfo;
 
        /* Must not be a fragment. */
-       if (offset)
+       if (par->fragoff != 0)
                return false;
 
-       ic = skb_header_pointer(skb, protoff, sizeof(_icmp), &_icmp);
+       ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
        if (ic == NULL) {
                /* We've been asked to examine this packet, and we
-                  can't.  Hence, no choice but to drop. */
+                * can't.  Hence, no choice but to drop.
+                */
                duprintf("Dropping evil ICMP tinygram.\n");
-               *hotdrop = true;
+               *par->hotdrop = true;
                return false;
        }
 
@@ -2165,14 +2224,9 @@ icmp6_match(const struct sk_buff *skb,
 }
 
 /* Called when user tries to insert an entry of this type. */
-static bool
-icmp6_checkentry(const char *tablename,
-          const void *entry,
-          const struct xt_match *match,
-          void *matchinfo,
-          unsigned int hook_mask)
+static bool icmp6_checkentry(const struct xt_mtchk_param *par)
 {
-       const struct ip6t_icmp *icmpinfo = matchinfo;
+       const struct ip6t_icmp *icmpinfo = par->matchinfo;
 
        /* Must specify no unknown invflags */
        return !(icmpinfo->invflags & ~IP6T_ICMP_INV);
@@ -2216,18 +2270,33 @@ static struct nf_sockopt_ops ip6t_sockopts = {
 
 static struct xt_match icmp6_matchstruct __read_mostly = {
        .name           = "icmp6",
-       .match          = &icmp6_match,
+       .match          = icmp6_match,
        .matchsize      = sizeof(struct ip6t_icmp),
        .checkentry     = icmp6_checkentry,
        .proto          = IPPROTO_ICMPV6,
        .family         = AF_INET6,
 };
 
+static int __net_init ip6_tables_net_init(struct net *net)
+{
+       return xt_proto_init(net, AF_INET6);
+}
+
+static void __net_exit ip6_tables_net_exit(struct net *net)
+{
+       xt_proto_fini(net, AF_INET6);
+}
+
+static struct pernet_operations ip6_tables_net_ops = {
+       .init = ip6_tables_net_init,
+       .exit = ip6_tables_net_exit,
+};
+
 static int __init ip6_tables_init(void)
 {
        int ret;
 
-       ret = xt_proto_init(AF_INET6);
+       ret = register_pernet_subsys(&ip6_tables_net_ops);
        if (ret < 0)
                goto err1;
 
@@ -2257,7 +2326,7 @@ err4:
 err3:
        xt_unregister_target(&ip6t_standard_target);
 err2:
-       xt_proto_fini(AF_INET6);
+       unregister_pernet_subsys(&ip6_tables_net_ops);
 err1:
        return ret;
 }
@@ -2265,10 +2334,12 @@ err1:
 static void __exit ip6_tables_fini(void)
 {
        nf_unregister_sockopt(&ip6t_sockopts);
+
        xt_unregister_match(&icmp6_matchstruct);
        xt_unregister_target(&ip6t_error_target);
        xt_unregister_target(&ip6t_standard_target);
-       xt_proto_fini(AF_INET6);
+
+       unregister_pernet_subsys(&ip6_tables_net_ops);
 }
 
 /*