Merge tag 'jfs-3.10' of git://github.com/kleikamp/linux-shaggy
[firefly-linux-kernel-4.4.55.git] / net / netfilter / nfnetlink_log.c
index f248db572972f71c3a40431e31fc1201eee5b78c..faf1e9300d8adc358fa857ff4296892351060193 100644 (file)
@@ -3,6 +3,7 @@
  * nfetlink.
  *
  * (C) 2005 by Harald Welte <laforge@netfilter.org>
+ * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
  *
  * Based on the old ipv4-only ipt_ULOG.c:
  * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
@@ -19,7 +20,7 @@
 #include <linux/ipv6.h>
 #include <linux/netdevice.h>
 #include <linux/netfilter.h>
-#include <linux/netlink.h>
+#include <net/netlink.h>
 #include <linux/netfilter/nfnetlink.h>
 #include <linux/netfilter/nfnetlink_log.h>
 #include <linux/spinlock.h>
@@ -32,6 +33,7 @@
 #include <linux/slab.h>
 #include <net/sock.h>
 #include <net/netfilter/nf_log.h>
+#include <net/netns/generic.h>
 #include <net/netfilter/nfnetlink_log.h>
 
 #include <linux/atomic.h>
@@ -56,6 +58,7 @@ struct nfulnl_instance {
        unsigned int qlen;              /* number of nlmsgs in skb */
        struct sk_buff *skb;            /* pre-allocatd skb */
        struct timer_list timer;
+       struct net *net;
        struct user_namespace *peer_user_ns;    /* User namespace of the peer process */
        int peer_portid;                        /* PORTID of the peer process */
 
@@ -71,25 +74,34 @@ struct nfulnl_instance {
        struct rcu_head rcu;
 };
 
-static DEFINE_SPINLOCK(instances_lock);
-static atomic_t global_seq;
-
 #define INSTANCE_BUCKETS       16
-static struct hlist_head instance_table[INSTANCE_BUCKETS];
 static unsigned int hash_init;
 
+static int nfnl_log_net_id __read_mostly;
+
+struct nfnl_log_net {
+       spinlock_t instances_lock;
+       struct hlist_head instance_table[INSTANCE_BUCKETS];
+       atomic_t global_seq;
+};
+
+static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
+{
+       return net_generic(net, nfnl_log_net_id);
+}
+
 static inline u_int8_t instance_hashfn(u_int16_t group_num)
 {
        return ((group_num & 0xff) % INSTANCE_BUCKETS);
 }
 
 static struct nfulnl_instance *
-__instance_lookup(u_int16_t group_num)
+__instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
 {
        struct hlist_head *head;
        struct nfulnl_instance *inst;
 
-       head = &instance_table[instance_hashfn(group_num)];
+       head = &log->instance_table[instance_hashfn(group_num)];
        hlist_for_each_entry_rcu(inst, head, hlist) {
                if (inst->group_num == group_num)
                        return inst;
@@ -104,12 +116,12 @@ instance_get(struct nfulnl_instance *inst)
 }
 
 static struct nfulnl_instance *
-instance_lookup_get(u_int16_t group_num)
+instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
 {
        struct nfulnl_instance *inst;
 
        rcu_read_lock_bh();
-       inst = __instance_lookup(group_num);
+       inst = __instance_lookup(log, group_num);
        if (inst && !atomic_inc_not_zero(&inst->use))
                inst = NULL;
        rcu_read_unlock_bh();
@@ -119,7 +131,11 @@ instance_lookup_get(u_int16_t group_num)
 
 static void nfulnl_instance_free_rcu(struct rcu_head *head)
 {
-       kfree(container_of(head, struct nfulnl_instance, rcu));
+       struct nfulnl_instance *inst =
+               container_of(head, struct nfulnl_instance, rcu);
+
+       put_net(inst->net);
+       kfree(inst);
        module_put(THIS_MODULE);
 }
 
@@ -133,13 +149,15 @@ instance_put(struct nfulnl_instance *inst)
 static void nfulnl_timer(unsigned long data);
 
 static struct nfulnl_instance *
-instance_create(u_int16_t group_num, int portid, struct user_namespace *user_ns)
+instance_create(struct net *net, u_int16_t group_num,
+               int portid, struct user_namespace *user_ns)
 {
        struct nfulnl_instance *inst;
+       struct nfnl_log_net *log = nfnl_log_pernet(net);
        int err;
 
-       spin_lock_bh(&instances_lock);
-       if (__instance_lookup(group_num)) {
+       spin_lock_bh(&log->instances_lock);
+       if (__instance_lookup(log, group_num)) {
                err = -EEXIST;
                goto out_unlock;
        }
@@ -163,6 +181,7 @@ instance_create(u_int16_t group_num, int portid, struct user_namespace *user_ns)
 
        setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
 
+       inst->net = get_net(net);
        inst->peer_user_ns = user_ns;
        inst->peer_portid = portid;
        inst->group_num = group_num;
@@ -174,14 +193,15 @@ instance_create(u_int16_t group_num, int portid, struct user_namespace *user_ns)
        inst->copy_range        = NFULNL_COPY_RANGE_MAX;
 
        hlist_add_head_rcu(&inst->hlist,
-                      &instance_table[instance_hashfn(group_num)]);
+                      &log->instance_table[instance_hashfn(group_num)]);
+
 
-       spin_unlock_bh(&instances_lock);
+       spin_unlock_bh(&log->instances_lock);
 
        return inst;
 
 out_unlock:
-       spin_unlock_bh(&instances_lock);
+       spin_unlock_bh(&log->instances_lock);
        return ERR_PTR(err);
 }
 
@@ -210,11 +230,12 @@ __instance_destroy(struct nfulnl_instance *inst)
 }
 
 static inline void
-instance_destroy(struct nfulnl_instance *inst)
+instance_destroy(struct nfnl_log_net *log,
+                struct nfulnl_instance *inst)
 {
-       spin_lock_bh(&instances_lock);
+       spin_lock_bh(&log->instances_lock);
        __instance_destroy(inst);
-       spin_unlock_bh(&instances_lock);
+       spin_unlock_bh(&log->instances_lock);
 }
 
 static int
@@ -298,7 +319,7 @@ nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
 }
 
 static struct sk_buff *
-nfulnl_alloc_skb(unsigned int inst_size, unsigned int pkt_size)
+nfulnl_alloc_skb(u32 peer_portid, unsigned int inst_size, unsigned int pkt_size)
 {
        struct sk_buff *skb;
        unsigned int n;
@@ -307,13 +328,14 @@ nfulnl_alloc_skb(unsigned int inst_size, unsigned int pkt_size)
         * message.  WARNING: has to be <= 128k due to slab restrictions */
 
        n = max(inst_size, pkt_size);
-       skb = alloc_skb(n, GFP_ATOMIC);
+       skb = nfnetlink_alloc_skb(&init_net, n, peer_portid, GFP_ATOMIC);
        if (!skb) {
                if (n > pkt_size) {
                        /* try to allocate only as much as we need for current
                         * packet */
 
-                       skb = alloc_skb(pkt_size, GFP_ATOMIC);
+                       skb = nfnetlink_alloc_skb(&init_net, pkt_size,
+                                                 peer_portid, GFP_ATOMIC);
                        if (!skb)
                                pr_err("nfnetlink_log: can't even alloc %u bytes\n",
                                       pkt_size);
@@ -336,7 +358,7 @@ __nfulnl_send(struct nfulnl_instance *inst)
                if (!nlh)
                        goto out;
        }
-       status = nfnetlink_unicast(inst->skb, &init_net, inst->peer_portid,
+       status = nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
                                   MSG_DONTWAIT);
 
        inst->qlen = 0;
@@ -370,7 +392,8 @@ nfulnl_timer(unsigned long data)
 /* This is an inline function, we don't really care about a long
  * list of arguments */
 static inline int
-__build_packet_message(struct nfulnl_instance *inst,
+__build_packet_message(struct nfnl_log_net *log,
+                       struct nfulnl_instance *inst,
                        const struct sk_buff *skb,
                        unsigned int data_len,
                        u_int8_t pf,
@@ -536,7 +559,7 @@ __build_packet_message(struct nfulnl_instance *inst,
        /* global sequence number */
        if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
            nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
-                        htonl(atomic_inc_return(&global_seq))))
+                        htonl(atomic_inc_return(&log->global_seq))))
                goto nla_put_failure;
 
        if (data_len) {
@@ -592,13 +615,15 @@ nfulnl_log_packet(u_int8_t pf,
        const struct nf_loginfo *li;
        unsigned int qthreshold;
        unsigned int plen;
+       struct net *net = dev_net(in ? in : out);
+       struct nfnl_log_net *log = nfnl_log_pernet(net);
 
        if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
                li = li_user;
        else
                li = &default_loginfo;
 
-       inst = instance_lookup_get(li->u.ulog.group);
+       inst = instance_lookup_get(log, li->u.ulog.group);
        if (!inst)
                return;
 
@@ -609,7 +634,7 @@ nfulnl_log_packet(u_int8_t pf,
        /* FIXME: do we want to make the size calculation conditional based on
         * what is actually present?  way more branches and checks, but more
         * memory efficient... */
-       size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
+       size =    nlmsg_total_size(sizeof(struct nfgenmsg))
                + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
@@ -673,14 +698,15 @@ nfulnl_log_packet(u_int8_t pf,
        }
 
        if (!inst->skb) {
-               inst->skb = nfulnl_alloc_skb(inst->nlbufsiz, size);
+               inst->skb = nfulnl_alloc_skb(inst->peer_portid, inst->nlbufsiz,
+                                            size);
                if (!inst->skb)
                        goto alloc_failure;
        }
 
        inst->qlen++;
 
-       __build_packet_message(inst, skb, data_len, pf,
+       __build_packet_message(log, inst, skb, data_len, pf,
                                hooknum, in, out, prefix, plen);
 
        if (inst->qlen >= qthreshold)
@@ -709,24 +735,24 @@ nfulnl_rcv_nl_event(struct notifier_block *this,
                   unsigned long event, void *ptr)
 {
        struct netlink_notify *n = ptr;
+       struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 
        if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
                int i;
 
                /* destroy all instances for this portid */
-               spin_lock_bh(&instances_lock);
+               spin_lock_bh(&log->instances_lock);
                for  (i = 0; i < INSTANCE_BUCKETS; i++) {
                        struct hlist_node *t2;
                        struct nfulnl_instance *inst;
-                       struct hlist_head *head = &instance_table[i];
+                       struct hlist_head *head = &log->instance_table[i];
 
                        hlist_for_each_entry_safe(inst, t2, head, hlist) {
-                               if ((net_eq(n->net, &init_net)) &&
-                                   (n->portid == inst->peer_portid))
+                               if (n->portid == inst->peer_portid)
                                        __instance_destroy(inst);
                        }
                }
-               spin_unlock_bh(&instances_lock);
+               spin_unlock_bh(&log->instances_lock);
        }
        return NOTIFY_DONE;
 }
@@ -767,6 +793,8 @@ nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
        u_int16_t group_num = ntohs(nfmsg->res_id);
        struct nfulnl_instance *inst;
        struct nfulnl_msg_config_cmd *cmd = NULL;
+       struct net *net = sock_net(ctnl);
+       struct nfnl_log_net *log = nfnl_log_pernet(net);
        int ret = 0;
 
        if (nfula[NFULA_CFG_CMD]) {
@@ -776,14 +804,14 @@ nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
                /* Commands without queue context */
                switch (cmd->command) {
                case NFULNL_CFG_CMD_PF_BIND:
-                       return nf_log_bind_pf(pf, &nfulnl_logger);
+                       return nf_log_bind_pf(net, pf, &nfulnl_logger);
                case NFULNL_CFG_CMD_PF_UNBIND:
-                       nf_log_unbind_pf(pf);
+                       nf_log_unbind_pf(net, pf);
                        return 0;
                }
        }
 
-       inst = instance_lookup_get(group_num);
+       inst = instance_lookup_get(log, group_num);
        if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
                ret = -EPERM;
                goto out_put;
@@ -797,9 +825,9 @@ nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
                                goto out_put;
                        }
 
-                       inst = instance_create(group_num,
+                       inst = instance_create(net, group_num,
                                               NETLINK_CB(skb).portid,
-                                              sk_user_ns(NETLINK_CB(skb).ssk));
+                                              sk_user_ns(NETLINK_CB(skb).sk));
                        if (IS_ERR(inst)) {
                                ret = PTR_ERR(inst);
                                goto out;
@@ -811,7 +839,7 @@ nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
                                goto out;
                        }
 
-                       instance_destroy(inst);
+                       instance_destroy(log, inst);
                        goto out_put;
                default:
                        ret = -ENOTSUPP;
@@ -894,55 +922,68 @@ static const struct nfnetlink_subsystem nfulnl_subsys = {
 
 #ifdef CONFIG_PROC_FS
 struct iter_state {
+       struct seq_net_private p;
        unsigned int bucket;
 };
 
-static struct hlist_node *get_first(struct iter_state *st)
+static struct hlist_node *get_first(struct net *net, struct iter_state *st)
 {
+       struct nfnl_log_net *log;
        if (!st)
                return NULL;
 
+       log = nfnl_log_pernet(net);
+
        for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
-               if (!hlist_empty(&instance_table[st->bucket]))
-                       return rcu_dereference_bh(hlist_first_rcu(&instance_table[st->bucket]));
+               struct hlist_head *head = &log->instance_table[st->bucket];
+
+               if (!hlist_empty(head))
+                       return rcu_dereference_bh(hlist_first_rcu(head));
        }
        return NULL;
 }
 
-static struct hlist_node *get_next(struct iter_state *st, struct hlist_node *h)
+static struct hlist_node *get_next(struct net *net, struct iter_state *st,
+                                  struct hlist_node *h)
 {
        h = rcu_dereference_bh(hlist_next_rcu(h));
        while (!h) {
+               struct nfnl_log_net *log;
+               struct hlist_head *head;
+
                if (++st->bucket >= INSTANCE_BUCKETS)
                        return NULL;
 
-               h = rcu_dereference_bh(hlist_first_rcu(&instance_table[st->bucket]));
+               log = nfnl_log_pernet(net);
+               head = &log->instance_table[st->bucket];
+               h = rcu_dereference_bh(hlist_first_rcu(head));
        }
        return h;
 }
 
-static struct hlist_node *get_idx(struct iter_state *st, loff_t pos)
+static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
+                                 loff_t pos)
 {
        struct hlist_node *head;
-       head = get_first(st);
+       head = get_first(net, st);
 
        if (head)
-               while (pos && (head = get_next(st, head)))
+               while (pos && (head = get_next(net, st, head)))
                        pos--;
        return pos ? NULL : head;
 }
 
-static void *seq_start(struct seq_file *seq, loff_t *pos)
+static void *seq_start(struct seq_file *s, loff_t *pos)
        __acquires(rcu_bh)
 {
        rcu_read_lock_bh();
-       return get_idx(seq->private, *pos);
+       return get_idx(seq_file_net(s), s->private, *pos);
 }
 
 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 {
        (*pos)++;
-       return get_next(s->private, v);
+       return get_next(seq_file_net(s), s->private, v);
 }
 
 static void seq_stop(struct seq_file *s, void *v)
@@ -971,8 +1012,8 @@ static const struct seq_operations nful_seq_ops = {
 
 static int nful_open(struct inode *inode, struct file *file)
 {
-       return seq_open_private(file, &nful_seq_ops,
-                       sizeof(struct iter_state));
+       return seq_open_net(inode, file, &nful_seq_ops,
+                           sizeof(struct iter_state));
 }
 
 static const struct file_operations nful_file_ops = {
@@ -980,17 +1021,43 @@ static const struct file_operations nful_file_ops = {
        .open    = nful_open,
        .read    = seq_read,
        .llseek  = seq_lseek,
-       .release = seq_release_private,
+       .release = seq_release_net,
 };
 
 #endif /* PROC_FS */
 
-static int __init nfnetlink_log_init(void)
+static int __net_init nfnl_log_net_init(struct net *net)
 {
-       int i, status = -ENOMEM;
+       unsigned int i;
+       struct nfnl_log_net *log = nfnl_log_pernet(net);
 
        for (i = 0; i < INSTANCE_BUCKETS; i++)
-               INIT_HLIST_HEAD(&instance_table[i]);
+               INIT_HLIST_HEAD(&log->instance_table[i]);
+       spin_lock_init(&log->instances_lock);
+
+#ifdef CONFIG_PROC_FS
+       if (!proc_create("nfnetlink_log", 0440,
+                        net->nf.proc_netfilter, &nful_file_ops))
+               return -ENOMEM;
+#endif
+       return 0;
+}
+
+static void __net_exit nfnl_log_net_exit(struct net *net)
+{
+       remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
+}
+
+static struct pernet_operations nfnl_log_net_ops = {
+       .init   = nfnl_log_net_init,
+       .exit   = nfnl_log_net_exit,
+       .id     = &nfnl_log_net_id,
+       .size   = sizeof(struct nfnl_log_net),
+};
+
+static int __init nfnetlink_log_init(void)
+{
+       int status = -ENOMEM;
 
        /* it's not really all that important to have a random value, so
         * we can do this from the init function, even if there hasn't
@@ -1000,29 +1067,25 @@ static int __init nfnetlink_log_init(void)
        netlink_register_notifier(&nfulnl_rtnl_notifier);
        status = nfnetlink_subsys_register(&nfulnl_subsys);
        if (status < 0) {
-               printk(KERN_ERR "log: failed to create netlink socket\n");
+               pr_err("log: failed to create netlink socket\n");
                goto cleanup_netlink_notifier;
        }
 
        status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
        if (status < 0) {
-               printk(KERN_ERR "log: failed to register logger\n");
+               pr_err("log: failed to register logger\n");
                goto cleanup_subsys;
        }
 
-#ifdef CONFIG_PROC_FS
-       if (!proc_create("nfnetlink_log", 0440,
-                        proc_net_netfilter, &nful_file_ops)) {
-               status = -ENOMEM;
+       status = register_pernet_subsys(&nfnl_log_net_ops);
+       if (status < 0) {
+               pr_err("log: failed to register pernet ops\n");
                goto cleanup_logger;
        }
-#endif
        return status;
 
-#ifdef CONFIG_PROC_FS
 cleanup_logger:
        nf_log_unregister(&nfulnl_logger);
-#endif
 cleanup_subsys:
        nfnetlink_subsys_unregister(&nfulnl_subsys);
 cleanup_netlink_notifier:
@@ -1032,10 +1095,8 @@ cleanup_netlink_notifier:
 
 static void __exit nfnetlink_log_fini(void)
 {
+       unregister_pernet_subsys(&nfnl_log_net_ops);
        nf_log_unregister(&nfulnl_logger);
-#ifdef CONFIG_PROC_FS
-       remove_proc_entry("nfnetlink_log", proc_net_netfilter);
-#endif
        nfnetlink_subsys_unregister(&nfulnl_subsys);
        netlink_unregister_notifier(&nfulnl_rtnl_notifier);
 }