OSDN Git Service

Merge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
authorDavid S. Miller <davem@davemloft.net>
Wed, 29 Mar 2017 21:35:25 +0000 (14:35 -0700)
committerDavid S. Miller <davem@davemloft.net>
Wed, 29 Mar 2017 21:35:25 +0000 (14:35 -0700)
Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contains a rather large update with Netfilter
fixes, specifically targeted to incorrect RCU usage in several spots and
the userspace conntrack helper infrastructure (nfnetlink_cthelper),
more specifically they are:

1) expect_class_max is incorrect set via cthelper, as in kernel semantics
   mandate that this represents the array of expectation classes minus 1.
   Patch from Liping Zhang.

2) Expectation policy updates via cthelper are currently broken for several
   reasons: This code allows illegal changes in the policy such as changing
   the number of expeciation classes, it is leaking the updated policy and
   such update occurs with no RCU protection at all. Fix this by adding a
   new nfnl_cthelper_update_policy() that describes what is really legal on
   the update path.

3) Fix several memory leaks in cthelper, from Jeffy Chen.

4) synchronize_rcu() is missing in the removal path of several modules,
   this may lead to races since CPU may still be running on code that has
   just gone. Also from Liping Zhang.

5) Don't use the helper hashtable from cthelper, it is not safe to walk
   over those bits without the helper mutex. Fix this by introducing a
   new independent list for userspace helpers. From Liping Zhang.

6) nf_ct_extend_unregister() needs synchronize_rcu() to make sure no
   packets are walking on any conntrack extension that is gone after
   module removal, again from Liping.

7) nf_nat_snmp may crash if we fail to unregister the helper due to
   accidental leftover code, from Gao Feng.

8) Fix leak in nfnetlink_queue with secctx support, from Liping Zhang.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/netfilter/nf_nat_snmp_basic.c
net/netfilter/nf_conntrack_ecache.c
net/netfilter/nf_conntrack_extend.c
net/netfilter/nf_conntrack_netlink.c
net/netfilter/nf_nat_core.c
net/netfilter/nfnetlink_cthelper.c
net/netfilter/nfnetlink_cttimeout.c
net/netfilter/nfnetlink_queue.c

index c9b52c3..53e49f5 100644 (file)
@@ -1260,16 +1260,6 @@ static const struct nf_conntrack_expect_policy snmp_exp_policy = {
        .timeout        = 180,
 };
 
-static struct nf_conntrack_helper snmp_helper __read_mostly = {
-       .me                     = THIS_MODULE,
-       .help                   = help,
-       .expect_policy          = &snmp_exp_policy,
-       .name                   = "snmp",
-       .tuple.src.l3num        = AF_INET,
-       .tuple.src.u.udp.port   = cpu_to_be16(SNMP_PORT),
-       .tuple.dst.protonum     = IPPROTO_UDP,
-};
-
 static struct nf_conntrack_helper snmp_trap_helper __read_mostly = {
        .me                     = THIS_MODULE,
        .help                   = help,
@@ -1288,22 +1278,16 @@ static struct nf_conntrack_helper snmp_trap_helper __read_mostly = {
 
 static int __init nf_nat_snmp_basic_init(void)
 {
-       int ret = 0;
-
        BUG_ON(nf_nat_snmp_hook != NULL);
        RCU_INIT_POINTER(nf_nat_snmp_hook, help);
 
-       ret = nf_conntrack_helper_register(&snmp_trap_helper);
-       if (ret < 0) {
-               nf_conntrack_helper_unregister(&snmp_helper);
-               return ret;
-       }
-       return ret;
+       return nf_conntrack_helper_register(&snmp_trap_helper);
 }
 
 static void __exit nf_nat_snmp_basic_fini(void)
 {
        RCU_INIT_POINTER(nf_nat_snmp_hook, NULL);
+       synchronize_rcu();
        nf_conntrack_helper_unregister(&snmp_trap_helper);
 }
 
index da9df2d..22fc321 100644 (file)
@@ -290,6 +290,7 @@ void nf_conntrack_unregister_notifier(struct net *net,
        BUG_ON(notify != new);
        RCU_INIT_POINTER(net->ct.nf_conntrack_event_cb, NULL);
        mutex_unlock(&nf_ct_ecache_mutex);
+       /* synchronize_rcu() is called from ctnetlink_exit. */
 }
 EXPORT_SYMBOL_GPL(nf_conntrack_unregister_notifier);
 
@@ -326,6 +327,7 @@ void nf_ct_expect_unregister_notifier(struct net *net,
        BUG_ON(notify != new);
        RCU_INIT_POINTER(net->ct.nf_expect_event_cb, NULL);
        mutex_unlock(&nf_ct_ecache_mutex);
+       /* synchronize_rcu() is called from ctnetlink_exit. */
 }
 EXPORT_SYMBOL_GPL(nf_ct_expect_unregister_notifier);
 
index 02bcf00..008299b 100644 (file)
@@ -53,7 +53,11 @@ nf_ct_ext_create(struct nf_ct_ext **ext, enum nf_ct_ext_id id,
 
        rcu_read_lock();
        t = rcu_dereference(nf_ct_ext_types[id]);
-       BUG_ON(t == NULL);
+       if (!t) {
+               rcu_read_unlock();
+               return NULL;
+       }
+
        off = ALIGN(sizeof(struct nf_ct_ext), t->align);
        len = off + t->len + var_alloc_len;
        alloc_size = t->alloc_size + var_alloc_len;
@@ -88,7 +92,10 @@ void *__nf_ct_ext_add_length(struct nf_conn *ct, enum nf_ct_ext_id id,
 
        rcu_read_lock();
        t = rcu_dereference(nf_ct_ext_types[id]);
-       BUG_ON(t == NULL);
+       if (!t) {
+               rcu_read_unlock();
+               return NULL;
+       }
 
        newoff = ALIGN(old->len, t->align);
        newlen = newoff + t->len + var_alloc_len;
@@ -175,6 +182,6 @@ void nf_ct_extend_unregister(struct nf_ct_ext_type *type)
        RCU_INIT_POINTER(nf_ct_ext_types[type->id], NULL);
        update_alloc_size(type);
        mutex_unlock(&nf_ct_ext_type_mutex);
-       rcu_barrier(); /* Wait for completion of call_rcu()'s */
+       synchronize_rcu();
 }
 EXPORT_SYMBOL_GPL(nf_ct_extend_unregister);
index 6806b5e..908d858 100644 (file)
@@ -3442,6 +3442,7 @@ static void __exit ctnetlink_exit(void)
 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT
        RCU_INIT_POINTER(nfnl_ct_hook, NULL);
 #endif
+       synchronize_rcu();
 }
 
 module_init(ctnetlink_init);
index 94b14c5..82802e4 100644 (file)
@@ -903,6 +903,8 @@ static void __exit nf_nat_cleanup(void)
 #ifdef CONFIG_XFRM
        RCU_INIT_POINTER(nf_nat_decode_session_hook, NULL);
 #endif
+       synchronize_rcu();
+
        for (i = 0; i < NFPROTO_NUMPROTO; i++)
                kfree(nf_nat_l4protos[i]);
 
index de87823..d455581 100644 (file)
@@ -32,6 +32,13 @@ MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
 MODULE_DESCRIPTION("nfnl_cthelper: User-space connection tracking helpers");
 
+struct nfnl_cthelper {
+       struct list_head                list;
+       struct nf_conntrack_helper      helper;
+};
+
+static LIST_HEAD(nfnl_cthelper_list);
+
 static int
 nfnl_userspace_cthelper(struct sk_buff *skb, unsigned int protoff,
                        struct nf_conn *ct, enum ip_conntrack_info ctinfo)
@@ -161,6 +168,7 @@ nfnl_cthelper_parse_expect_policy(struct nf_conntrack_helper *helper,
        int i, ret;
        struct nf_conntrack_expect_policy *expect_policy;
        struct nlattr *tb[NFCTH_POLICY_SET_MAX+1];
+       unsigned int class_max;
 
        ret = nla_parse_nested(tb, NFCTH_POLICY_SET_MAX, attr,
                               nfnl_cthelper_expect_policy_set);
@@ -170,19 +178,18 @@ nfnl_cthelper_parse_expect_policy(struct nf_conntrack_helper *helper,
        if (!tb[NFCTH_POLICY_SET_NUM])
                return -EINVAL;
 
-       helper->expect_class_max =
-               ntohl(nla_get_be32(tb[NFCTH_POLICY_SET_NUM]));
-
-       if (helper->expect_class_max != 0 &&
-           helper->expect_class_max > NF_CT_MAX_EXPECT_CLASSES)
+       class_max = ntohl(nla_get_be32(tb[NFCTH_POLICY_SET_NUM]));
+       if (class_max == 0)
+               return -EINVAL;
+       if (class_max > NF_CT_MAX_EXPECT_CLASSES)
                return -EOVERFLOW;
 
        expect_policy = kzalloc(sizeof(struct nf_conntrack_expect_policy) *
-                               helper->expect_class_max, GFP_KERNEL);
+                               class_max, GFP_KERNEL);
        if (expect_policy == NULL)
                return -ENOMEM;
 
-       for (i=0; i<helper->expect_class_max; i++) {
+       for (i = 0; i < class_max; i++) {
                if (!tb[NFCTH_POLICY_SET+i])
                        goto err;
 
@@ -191,6 +198,8 @@ nfnl_cthelper_parse_expect_policy(struct nf_conntrack_helper *helper,
                if (ret < 0)
                        goto err;
        }
+
+       helper->expect_class_max = class_max - 1;
        helper->expect_policy = expect_policy;
        return 0;
 err:
@@ -203,18 +212,20 @@ nfnl_cthelper_create(const struct nlattr * const tb[],
                     struct nf_conntrack_tuple *tuple)
 {
        struct nf_conntrack_helper *helper;
+       struct nfnl_cthelper *nfcth;
        int ret;
 
        if (!tb[NFCTH_TUPLE] || !tb[NFCTH_POLICY] || !tb[NFCTH_PRIV_DATA_LEN])
                return -EINVAL;
 
-       helper = kzalloc(sizeof(struct nf_conntrack_helper), GFP_KERNEL);
-       if (helper == NULL)
+       nfcth = kzalloc(sizeof(*nfcth), GFP_KERNEL);
+       if (nfcth == NULL)
                return -ENOMEM;
+       helper = &nfcth->helper;
 
        ret = nfnl_cthelper_parse_expect_policy(helper, tb[NFCTH_POLICY]);
        if (ret < 0)
-               goto err;
+               goto err1;
 
        strncpy(helper->name, nla_data(tb[NFCTH_NAME]), NF_CT_HELPER_NAME_LEN);
        helper->data_len = ntohl(nla_get_be32(tb[NFCTH_PRIV_DATA_LEN]));
@@ -245,15 +256,101 @@ nfnl_cthelper_create(const struct nlattr * const tb[],
 
        ret = nf_conntrack_helper_register(helper);
        if (ret < 0)
-               goto err;
+               goto err2;
 
+       list_add_tail(&nfcth->list, &nfnl_cthelper_list);
        return 0;
-err:
-       kfree(helper);
+err2:
+       kfree(helper->expect_policy);
+err1:
+       kfree(nfcth);
        return ret;
 }
 
 static int
+nfnl_cthelper_update_policy_one(const struct nf_conntrack_expect_policy *policy,
+                               struct nf_conntrack_expect_policy *new_policy,
+                               const struct nlattr *attr)
+{
+       struct nlattr *tb[NFCTH_POLICY_MAX + 1];
+       int err;
+
+       err = nla_parse_nested(tb, NFCTH_POLICY_MAX, attr,
+                              nfnl_cthelper_expect_pol);
+       if (err < 0)
+               return err;
+
+       if (!tb[NFCTH_POLICY_NAME] ||
+           !tb[NFCTH_POLICY_EXPECT_MAX] ||
+           !tb[NFCTH_POLICY_EXPECT_TIMEOUT])
+               return -EINVAL;
+
+       if (nla_strcmp(tb[NFCTH_POLICY_NAME], policy->name))
+               return -EBUSY;
+
+       new_policy->max_expected =
+               ntohl(nla_get_be32(tb[NFCTH_POLICY_EXPECT_MAX]));
+       new_policy->timeout =
+               ntohl(nla_get_be32(tb[NFCTH_POLICY_EXPECT_TIMEOUT]));
+
+       return 0;
+}
+
+static int nfnl_cthelper_update_policy_all(struct nlattr *tb[],
+                                          struct nf_conntrack_helper *helper)
+{
+       struct nf_conntrack_expect_policy new_policy[helper->expect_class_max + 1];
+       struct nf_conntrack_expect_policy *policy;
+       int i, err;
+
+       /* Check first that all policy attributes are well-formed, so we don't
+        * leave things in inconsistent state on errors.
+        */
+       for (i = 0; i < helper->expect_class_max + 1; i++) {
+
+               if (!tb[NFCTH_POLICY_SET + i])
+                       return -EINVAL;
+
+               err = nfnl_cthelper_update_policy_one(&helper->expect_policy[i],
+                                                     &new_policy[i],
+                                                     tb[NFCTH_POLICY_SET + i]);
+               if (err < 0)
+                       return err;
+       }
+       /* Now we can safely update them. */
+       for (i = 0; i < helper->expect_class_max + 1; i++) {
+               policy = (struct nf_conntrack_expect_policy *)
+                               &helper->expect_policy[i];
+               policy->max_expected = new_policy->max_expected;
+               policy->timeout = new_policy->timeout;
+       }
+
+       return 0;
+}
+
+static int nfnl_cthelper_update_policy(struct nf_conntrack_helper *helper,
+                                      const struct nlattr *attr)
+{
+       struct nlattr *tb[NFCTH_POLICY_SET_MAX + 1];
+       unsigned int class_max;
+       int err;
+
+       err = nla_parse_nested(tb, NFCTH_POLICY_SET_MAX, attr,
+                              nfnl_cthelper_expect_policy_set);
+       if (err < 0)
+               return err;
+
+       if (!tb[NFCTH_POLICY_SET_NUM])
+               return -EINVAL;
+
+       class_max = ntohl(nla_get_be32(tb[NFCTH_POLICY_SET_NUM]));
+       if (helper->expect_class_max + 1 != class_max)
+               return -EBUSY;
+
+       return nfnl_cthelper_update_policy_all(tb, helper);
+}
+
+static int
 nfnl_cthelper_update(const struct nlattr * const tb[],
                     struct nf_conntrack_helper *helper)
 {
@@ -263,8 +360,7 @@ nfnl_cthelper_update(const struct nlattr * const tb[],
                return -EBUSY;
 
        if (tb[NFCTH_POLICY]) {
-               ret = nfnl_cthelper_parse_expect_policy(helper,
-                                                       tb[NFCTH_POLICY]);
+               ret = nfnl_cthelper_update_policy(helper, tb[NFCTH_POLICY]);
                if (ret < 0)
                        return ret;
        }
@@ -293,7 +389,8 @@ static int nfnl_cthelper_new(struct net *net, struct sock *nfnl,
        const char *helper_name;
        struct nf_conntrack_helper *cur, *helper = NULL;
        struct nf_conntrack_tuple tuple;
-       int ret = 0, i;
+       struct nfnl_cthelper *nlcth;
+       int ret = 0;
 
        if (!tb[NFCTH_NAME] || !tb[NFCTH_TUPLE])
                return -EINVAL;
@@ -304,31 +401,22 @@ static int nfnl_cthelper_new(struct net *net, struct sock *nfnl,
        if (ret < 0)
                return ret;
 
-       rcu_read_lock();
-       for (i = 0; i < nf_ct_helper_hsize && !helper; i++) {
-               hlist_for_each_entry_rcu(cur, &nf_ct_helper_hash[i], hnode) {
+       list_for_each_entry(nlcth, &nfnl_cthelper_list, list) {
+               cur = &nlcth->helper;
 
-                       /* skip non-userspace conntrack helpers. */
-                       if (!(cur->flags & NF_CT_HELPER_F_USERSPACE))
-                               continue;
+               if (strncmp(cur->name, helper_name, NF_CT_HELPER_NAME_LEN))
+                       continue;
 
-                       if (strncmp(cur->name, helper_name,
-                                       NF_CT_HELPER_NAME_LEN) != 0)
-                               continue;
+               if ((tuple.src.l3num != cur->tuple.src.l3num ||
+                    tuple.dst.protonum != cur->tuple.dst.protonum))
+                       continue;
 
-                       if ((tuple.src.l3num != cur->tuple.src.l3num ||
-                            tuple.dst.protonum != cur->tuple.dst.protonum))
-                               continue;
+               if (nlh->nlmsg_flags & NLM_F_EXCL)
+                       return -EEXIST;
 
-                       if (nlh->nlmsg_flags & NLM_F_EXCL) {
-                               ret = -EEXIST;
-                               goto err;
-                       }
-                       helper = cur;
-                       break;
-               }
+               helper = cur;
+               break;
        }
-       rcu_read_unlock();
 
        if (helper == NULL)
                ret = nfnl_cthelper_create(tb, &tuple);
@@ -336,9 +424,6 @@ static int nfnl_cthelper_new(struct net *net, struct sock *nfnl,
                ret = nfnl_cthelper_update(tb, helper);
 
        return ret;
-err:
-       rcu_read_unlock();
-       return ret;
 }
 
 static int
@@ -377,10 +462,10 @@ nfnl_cthelper_dump_policy(struct sk_buff *skb,
                goto nla_put_failure;
 
        if (nla_put_be32(skb, NFCTH_POLICY_SET_NUM,
-                        htonl(helper->expect_class_max)))
+                        htonl(helper->expect_class_max + 1)))
                goto nla_put_failure;
 
-       for (i=0; i<helper->expect_class_max; i++) {
+       for (i = 0; i < helper->expect_class_max + 1; i++) {
                nest_parms2 = nla_nest_start(skb,
                                (NFCTH_POLICY_SET+i) | NLA_F_NESTED);
                if (nest_parms2 == NULL)
@@ -502,11 +587,12 @@ static int nfnl_cthelper_get(struct net *net, struct sock *nfnl,
                             struct sk_buff *skb, const struct nlmsghdr *nlh,
                             const struct nlattr * const tb[])
 {
-       int ret = -ENOENT, i;
+       int ret = -ENOENT;
        struct nf_conntrack_helper *cur;
        struct sk_buff *skb2;
        char *helper_name = NULL;
        struct nf_conntrack_tuple tuple;
+       struct nfnl_cthelper *nlcth;
        bool tuple_set = false;
 
        if (nlh->nlmsg_flags & NLM_F_DUMP) {
@@ -527,45 +613,39 @@ static int nfnl_cthelper_get(struct net *net, struct sock *nfnl,
                tuple_set = true;
        }
 
-       for (i = 0; i < nf_ct_helper_hsize; i++) {
-               hlist_for_each_entry_rcu(cur, &nf_ct_helper_hash[i], hnode) {
+       list_for_each_entry(nlcth, &nfnl_cthelper_list, list) {
+               cur = &nlcth->helper;
+               if (helper_name &&
+                   strncmp(cur->name, helper_name, NF_CT_HELPER_NAME_LEN))
+                       continue;
 
-                       /* skip non-userspace conntrack helpers. */
-                       if (!(cur->flags & NF_CT_HELPER_F_USERSPACE))
-                               continue;
+               if (tuple_set &&
+                   (tuple.src.l3num != cur->tuple.src.l3num ||
+                    tuple.dst.protonum != cur->tuple.dst.protonum))
+                       continue;
 
-                       if (helper_name && strncmp(cur->name, helper_name,
-                                               NF_CT_HELPER_NAME_LEN) != 0) {
-                               continue;
-                       }
-                       if (tuple_set &&
-                           (tuple.src.l3num != cur->tuple.src.l3num ||
-                            tuple.dst.protonum != cur->tuple.dst.protonum))
-                               continue;
-
-                       skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
-                       if (skb2 == NULL) {
-                               ret = -ENOMEM;
-                               break;
-                       }
+               skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
+               if (skb2 == NULL) {
+                       ret = -ENOMEM;
+                       break;
+               }
 
-                       ret = nfnl_cthelper_fill_info(skb2, NETLINK_CB(skb).portid,
-                                               nlh->nlmsg_seq,
-                                               NFNL_MSG_TYPE(nlh->nlmsg_type),
-                                               NFNL_MSG_CTHELPER_NEW, cur);
-                       if (ret <= 0) {
-                               kfree_skb(skb2);
-                               break;
-                       }
+               ret = nfnl_cthelper_fill_info(skb2, NETLINK_CB(skb).portid,
+                                             nlh->nlmsg_seq,
+                                             NFNL_MSG_TYPE(nlh->nlmsg_type),
+                                             NFNL_MSG_CTHELPER_NEW, cur);
+               if (ret <= 0) {
+                       kfree_skb(skb2);
+                       break;
+               }
 
-                       ret = netlink_unicast(nfnl, skb2, NETLINK_CB(skb).portid,
-                                               MSG_DONTWAIT);
-                       if (ret > 0)
-                               ret = 0;
+               ret = netlink_unicast(nfnl, skb2, NETLINK_CB(skb).portid,
+                                     MSG_DONTWAIT);
+               if (ret > 0)
+                       ret = 0;
 
-                       /* this avoids a loop in nfnetlink. */
-                       return ret == -EAGAIN ? -ENOBUFS : ret;
-               }
+               /* this avoids a loop in nfnetlink. */
+               return ret == -EAGAIN ? -ENOBUFS : ret;
        }
        return ret;
 }
@@ -576,10 +656,10 @@ static int nfnl_cthelper_del(struct net *net, struct sock *nfnl,
 {
        char *helper_name = NULL;
        struct nf_conntrack_helper *cur;
-       struct hlist_node *tmp;
        struct nf_conntrack_tuple tuple;
        bool tuple_set = false, found = false;
-       int i, j = 0, ret;
+       struct nfnl_cthelper *nlcth, *n;
+       int j = 0, ret;
 
        if (tb[NFCTH_NAME])
                helper_name = nla_data(tb[NFCTH_NAME]);
@@ -592,28 +672,27 @@ static int nfnl_cthelper_del(struct net *net, struct sock *nfnl,
                tuple_set = true;
        }
 
-       for (i = 0; i < nf_ct_helper_hsize; i++) {
-               hlist_for_each_entry_safe(cur, tmp, &nf_ct_helper_hash[i],
-                                                               hnode) {
-                       /* skip non-userspace conntrack helpers. */
-                       if (!(cur->flags & NF_CT_HELPER_F_USERSPACE))
-                               continue;
+       list_for_each_entry_safe(nlcth, n, &nfnl_cthelper_list, list) {
+               cur = &nlcth->helper;
+               j++;
 
-                       j++;
+               if (helper_name &&
+                   strncmp(cur->name, helper_name, NF_CT_HELPER_NAME_LEN))
+                       continue;
 
-                       if (helper_name && strncmp(cur->name, helper_name,
-                                               NF_CT_HELPER_NAME_LEN) != 0) {
-                               continue;
-                       }
-                       if (tuple_set &&
-                           (tuple.src.l3num != cur->tuple.src.l3num ||
-                            tuple.dst.protonum != cur->tuple.dst.protonum))
-                               continue;
+               if (tuple_set &&
+                   (tuple.src.l3num != cur->tuple.src.l3num ||
+                    tuple.dst.protonum != cur->tuple.dst.protonum))
+                       continue;
 
-                       found = true;
-                       nf_conntrack_helper_unregister(cur);
-               }
+               found = true;
+               nf_conntrack_helper_unregister(cur);
+               kfree(cur->expect_policy);
+
+               list_del(&nlcth->list);
+               kfree(nlcth);
        }
+
        /* Make sure we return success if we flush and there is no helpers */
        return (found || j == 0) ? 0 : -ENOENT;
 }
@@ -662,20 +741,16 @@ err_out:
 static void __exit nfnl_cthelper_exit(void)
 {
        struct nf_conntrack_helper *cur;
-       struct hlist_node *tmp;
-       int i;
+       struct nfnl_cthelper *nlcth, *n;
 
        nfnetlink_subsys_unregister(&nfnl_cthelper_subsys);
 
-       for (i=0; i<nf_ct_helper_hsize; i++) {
-               hlist_for_each_entry_safe(cur, tmp, &nf_ct_helper_hash[i],
-                                                                       hnode) {
-                       /* skip non-userspace conntrack helpers. */
-                       if (!(cur->flags & NF_CT_HELPER_F_USERSPACE))
-                               continue;
+       list_for_each_entry_safe(nlcth, n, &nfnl_cthelper_list, list) {
+               cur = &nlcth->helper;
 
-                       nf_conntrack_helper_unregister(cur);
-               }
+               nf_conntrack_helper_unregister(cur);
+               kfree(cur->expect_policy);
+               kfree(nlcth);
        }
 }
 
index 139e086..47d6656 100644 (file)
@@ -646,8 +646,8 @@ static void __exit cttimeout_exit(void)
 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
        RCU_INIT_POINTER(nf_ct_timeout_find_get_hook, NULL);
        RCU_INIT_POINTER(nf_ct_timeout_put_hook, NULL);
+       synchronize_rcu();
 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */
-       rcu_barrier();
 }
 
 module_init(cttimeout_init);
index 3ee0b8a..933509e 100644 (file)
@@ -443,7 +443,7 @@ nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
        skb = alloc_skb(size, GFP_ATOMIC);
        if (!skb) {
                skb_tx_error(entskb);
-               return NULL;
+               goto nlmsg_failure;
        }
 
        nlh = nlmsg_put(skb, 0, 0,
@@ -452,7 +452,7 @@ nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
        if (!nlh) {
                skb_tx_error(entskb);
                kfree_skb(skb);
-               return NULL;
+               goto nlmsg_failure;
        }
        nfmsg = nlmsg_data(nlh);
        nfmsg->nfgen_family = entry->state.pf;
@@ -598,12 +598,17 @@ nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
        }
 
        nlh->nlmsg_len = skb->len;
+       if (seclen)
+               security_release_secctx(secdata, seclen);
        return skb;
 
 nla_put_failure:
        skb_tx_error(entskb);
        kfree_skb(skb);
        net_err_ratelimited("nf_queue: error creating packet message\n");
+nlmsg_failure:
+       if (seclen)
+               security_release_secctx(secdata, seclen);
        return NULL;
 }