#define NF_CONNTRACK_VERSION "0.5.0"
-DEFINE_RWLOCK(nf_conntrack_lock);
+DEFINE_SPINLOCK(nf_conntrack_lock);
EXPORT_SYMBOL_GPL(nf_conntrack_lock);
/* nf_conntrack_standalone needs this */
HLIST_HEAD(unconfirmed);
static int nf_conntrack_vmalloc __read_mostly;
static struct kmem_cache *nf_conntrack_cachep __read_mostly;
-static unsigned int nf_conntrack_next_id;
DEFINE_PER_CPU(struct ip_conntrack_stat, nf_conntrack_stat);
EXPORT_PER_CPU_SYMBOL(nf_conntrack_stat);
static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple,
unsigned int size, unsigned int rnd)
{
- unsigned int a, b;
-
- a = jhash2(tuple->src.u3.all, ARRAY_SIZE(tuple->src.u3.all),
- (tuple->src.l3num << 16) | tuple->dst.protonum);
- b = jhash2(tuple->dst.u3.all, ARRAY_SIZE(tuple->dst.u3.all),
- ((__force __u16)tuple->src.u.all << 16) |
- (__force __u16)tuple->dst.u.all);
-
- return jhash_2words(a, b, rnd) % size;
+ unsigned int n;
+ u_int32_t h;
+
+ /* The direction must be ignored, so we hash everything up to the
+ * destination ports (which is a multiple of 4) and treat the last
+ * three bytes manually.
+ */
+ n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32);
+ h = jhash2((u32 *)tuple, n,
+ rnd ^ (((__force __u16)tuple->dst.u.all << 16) |
+ tuple->dst.protonum));
+
+ return ((u64)h * size) >> 32;
}
static inline u_int32_t hash_conntrack(const struct nf_conntrack_tuple *tuple)
clean_from_lists(struct nf_conn *ct)
{
pr_debug("clean_from_lists(%p)\n", ct);
- hlist_del(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnode);
- hlist_del(&ct->tuplehash[IP_CT_DIR_REPLY].hnode);
+ hlist_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnode);
+ hlist_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnode);
/* Destroy all pending expectations */
nf_ct_remove_expectations(ct);
rcu_read_unlock();
- write_lock_bh(&nf_conntrack_lock);
+ spin_lock_bh(&nf_conntrack_lock);
/* Expectations will have been removed in clean_from_lists,
* except TFTP can create an expectation on the first packet,
* before connection is in the list, so we need to clean here,
}
NF_CT_STAT_INC(delete);
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
if (ct->master)
nf_ct_put(ct->master);
rcu_read_unlock();
}
- write_lock_bh(&nf_conntrack_lock);
+ spin_lock_bh(&nf_conntrack_lock);
/* Inside lock so preempt is disabled on module removal path.
* Otherwise we can get spurious warnings. */
NF_CT_STAT_INC(delete_list);
clean_from_lists(ct);
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
nf_ct_put(ct);
}
struct nf_conntrack_tuple_hash *
-__nf_conntrack_find(const struct nf_conntrack_tuple *tuple,
- const struct nf_conn *ignored_conntrack)
+__nf_conntrack_find(const struct nf_conntrack_tuple *tuple)
{
struct nf_conntrack_tuple_hash *h;
struct hlist_node *n;
unsigned int hash = hash_conntrack(tuple);
- hlist_for_each_entry(h, n, &nf_conntrack_hash[hash], hnode) {
- if (nf_ct_tuplehash_to_ctrack(h) != ignored_conntrack &&
- nf_ct_tuple_equal(tuple, &h->tuple)) {
+ hlist_for_each_entry_rcu(h, n, &nf_conntrack_hash[hash], hnode) {
+ if (nf_ct_tuple_equal(tuple, &h->tuple)) {
NF_CT_STAT_INC(found);
return h;
}
nf_conntrack_find_get(const struct nf_conntrack_tuple *tuple)
{
struct nf_conntrack_tuple_hash *h;
+ struct nf_conn *ct;
- read_lock_bh(&nf_conntrack_lock);
- h = __nf_conntrack_find(tuple, NULL);
- if (h)
- atomic_inc(&nf_ct_tuplehash_to_ctrack(h)->ct_general.use);
- read_unlock_bh(&nf_conntrack_lock);
+ rcu_read_lock();
+ h = __nf_conntrack_find(tuple);
+ if (h) {
+ ct = nf_ct_tuplehash_to_ctrack(h);
+ if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use)))
+ h = NULL;
+ }
+ rcu_read_unlock();
return h;
}
unsigned int hash,
unsigned int repl_hash)
{
- ct->id = ++nf_conntrack_next_id;
- hlist_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnode,
- &nf_conntrack_hash[hash]);
- hlist_add_head(&ct->tuplehash[IP_CT_DIR_REPLY].hnode,
- &nf_conntrack_hash[repl_hash]);
+ hlist_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnode,
+ &nf_conntrack_hash[hash]);
+ hlist_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnode,
+ &nf_conntrack_hash[repl_hash]);
}
void nf_conntrack_hash_insert(struct nf_conn *ct)
hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple);
repl_hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_REPLY].tuple);
- write_lock_bh(&nf_conntrack_lock);
+ spin_lock_bh(&nf_conntrack_lock);
__nf_conntrack_hash_insert(ct, hash, repl_hash);
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
}
EXPORT_SYMBOL_GPL(nf_conntrack_hash_insert);
/* Confirm a connection given skb; places it in hash table */
int
-__nf_conntrack_confirm(struct sk_buff **pskb)
+__nf_conntrack_confirm(struct sk_buff *skb)
{
unsigned int hash, repl_hash;
struct nf_conntrack_tuple_hash *h;
struct hlist_node *n;
enum ip_conntrack_info ctinfo;
- ct = nf_ct_get(*pskb, &ctinfo);
+ ct = nf_ct_get(skb, &ctinfo);
/* ipt_REJECT uses nf_conntrack_attach to attach related
ICMP/TCP RST packets in other direction. Actual packet
NF_CT_ASSERT(!nf_ct_is_confirmed(ct));
pr_debug("Confirming conntrack %p\n", ct);
- write_lock_bh(&nf_conntrack_lock);
+ spin_lock_bh(&nf_conntrack_lock);
/* See if there's one in the list already, including reverse:
NAT could have grabbed it without realizing, since we're
atomic_inc(&ct->ct_general.use);
set_bit(IPS_CONFIRMED_BIT, &ct->status);
NF_CT_STAT_INC(insert);
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
help = nfct_help(ct);
if (help && help->helper)
- nf_conntrack_event_cache(IPCT_HELPER, *pskb);
+ nf_conntrack_event_cache(IPCT_HELPER, skb);
#ifdef CONFIG_NF_NAT_NEEDED
if (test_bit(IPS_SRC_NAT_DONE_BIT, &ct->status) ||
test_bit(IPS_DST_NAT_DONE_BIT, &ct->status))
- nf_conntrack_event_cache(IPCT_NATINFO, *pskb);
+ nf_conntrack_event_cache(IPCT_NATINFO, skb);
#endif
nf_conntrack_event_cache(master_ct(ct) ?
- IPCT_RELATED : IPCT_NEW, *pskb);
+ IPCT_RELATED : IPCT_NEW, skb);
return NF_ACCEPT;
out:
NF_CT_STAT_INC(insert_failed);
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
return NF_DROP;
}
EXPORT_SYMBOL_GPL(__nf_conntrack_confirm);
const struct nf_conn *ignored_conntrack)
{
struct nf_conntrack_tuple_hash *h;
+ struct hlist_node *n;
+ unsigned int hash = hash_conntrack(tuple);
- read_lock_bh(&nf_conntrack_lock);
- h = __nf_conntrack_find(tuple, ignored_conntrack);
- read_unlock_bh(&nf_conntrack_lock);
+ rcu_read_lock();
+ hlist_for_each_entry_rcu(h, n, &nf_conntrack_hash[hash], hnode) {
+ if (nf_ct_tuplehash_to_ctrack(h) != ignored_conntrack &&
+ nf_ct_tuple_equal(tuple, &h->tuple)) {
+ NF_CT_STAT_INC(found);
+ rcu_read_unlock();
+ return 1;
+ }
+ NF_CT_STAT_INC(searched);
+ }
+ rcu_read_unlock();
- return h != NULL;
+ return 0;
}
EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken);
unsigned int i, cnt = 0;
int dropped = 0;
- read_lock_bh(&nf_conntrack_lock);
+ rcu_read_lock();
for (i = 0; i < nf_conntrack_htable_size; i++) {
- hlist_for_each_entry(h, n, &nf_conntrack_hash[hash], hnode) {
+ hlist_for_each_entry_rcu(h, n, &nf_conntrack_hash[hash],
+ hnode) {
tmp = nf_ct_tuplehash_to_ctrack(h);
if (!test_bit(IPS_ASSURED_BIT, &tmp->status))
ct = tmp;
cnt++;
}
+
+ if (ct && unlikely(!atomic_inc_not_zero(&ct->ct_general.use)))
+ ct = NULL;
if (ct || cnt >= NF_CT_EVICTION_RANGE)
break;
hash = (hash + 1) % nf_conntrack_htable_size;
}
- if (ct)
- atomic_inc(&ct->ct_general.use);
- read_unlock_bh(&nf_conntrack_lock);
+ rcu_read_unlock();
if (!ct)
return dropped;
/* Don't set timer yet: wait for confirmation */
setup_timer(&conntrack->timeout, death_by_timeout,
(unsigned long)conntrack);
+ INIT_RCU_HEAD(&conntrack->rcu);
return conntrack;
}
EXPORT_SYMBOL_GPL(nf_conntrack_alloc);
-void nf_conntrack_free(struct nf_conn *conntrack)
+static void nf_conntrack_free_rcu(struct rcu_head *head)
{
- nf_ct_ext_free(conntrack);
- kmem_cache_free(nf_conntrack_cachep, conntrack);
+ struct nf_conn *ct = container_of(head, struct nf_conn, rcu);
+
+ nf_ct_ext_free(ct);
+ kmem_cache_free(nf_conntrack_cachep, ct);
atomic_dec(&nf_conntrack_count);
}
+
+void nf_conntrack_free(struct nf_conn *conntrack)
+{
+ call_rcu(&conntrack->rcu, nf_conntrack_free_rcu);
+}
EXPORT_SYMBOL_GPL(nf_conntrack_free);
/* Allocate a new conntrack: we return -ENOMEM if classification
return NULL;
}
- write_lock_bh(&nf_conntrack_lock);
+ spin_lock_bh(&nf_conntrack_lock);
exp = nf_ct_find_expectation(tuple);
if (exp) {
pr_debug("conntrack: expectation arrives ct=%p exp=%p\n",
hlist_add_head(&conntrack->tuplehash[IP_CT_DIR_ORIGINAL].hnode,
&unconfirmed);
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
if (exp) {
if (exp->expectfn)
}
unsigned int
-nf_conntrack_in(int pf, unsigned int hooknum, struct sk_buff **pskb)
+nf_conntrack_in(int pf, unsigned int hooknum, struct sk_buff *skb)
{
struct nf_conn *ct;
enum ip_conntrack_info ctinfo;
int ret;
/* Previously seen (loopback or untracked)? Ignore. */
- if ((*pskb)->nfct) {
+ if (skb->nfct) {
NF_CT_STAT_INC_ATOMIC(ignore);
return NF_ACCEPT;
}
/* rcu_read_lock()ed by nf_hook_slow */
l3proto = __nf_ct_l3proto_find((u_int16_t)pf);
- ret = l3proto->get_l4proto(*pskb, skb_network_offset(*pskb),
+ ret = l3proto->get_l4proto(skb, skb_network_offset(skb),
&dataoff, &protonum);
if (ret <= 0) {
pr_debug("not prepared to track yet or error occured\n");
* inverse of the return code tells to the netfilter
* core what to do with the packet. */
if (l4proto->error != NULL &&
- (ret = l4proto->error(*pskb, dataoff, &ctinfo, pf, hooknum)) <= 0) {
+ (ret = l4proto->error(skb, dataoff, &ctinfo, pf, hooknum)) <= 0) {
NF_CT_STAT_INC_ATOMIC(error);
NF_CT_STAT_INC_ATOMIC(invalid);
return -ret;
}
- ct = resolve_normal_ct(*pskb, dataoff, pf, protonum, l3proto, l4proto,
+ ct = resolve_normal_ct(skb, dataoff, pf, protonum, l3proto, l4proto,
&set_reply, &ctinfo);
if (!ct) {
/* Not valid part of a connection */
return NF_DROP;
}
- NF_CT_ASSERT((*pskb)->nfct);
+ NF_CT_ASSERT(skb->nfct);
- ret = l4proto->packet(ct, *pskb, dataoff, ctinfo, pf, hooknum);
+ ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum);
if (ret < 0) {
/* Invalid: inverse of the return code tells
* the netfilter core what to do */
pr_debug("nf_conntrack_in: Can't track with proto module\n");
- nf_conntrack_put((*pskb)->nfct);
- (*pskb)->nfct = NULL;
+ nf_conntrack_put(skb->nfct);
+ skb->nfct = NULL;
NF_CT_STAT_INC_ATOMIC(invalid);
return -ret;
}
if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status))
- nf_conntrack_event_cache(IPCT_STATUS, *pskb);
+ nf_conntrack_event_cache(IPCT_STATUS, skb);
return ret;
}
struct nf_conn_help *help = nfct_help(ct);
struct nf_conntrack_helper *helper;
- write_lock_bh(&nf_conntrack_lock);
/* Should be unconfirmed, so not in hash table yet */
NF_CT_ASSERT(!nf_ct_is_confirmed(ct));
ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply;
if (ct->master || (help && help->expecting != 0))
- goto out;
+ return;
+ rcu_read_lock();
helper = __nf_ct_helper_find(newreply);
if (helper == NULL) {
if (help)
rcu_assign_pointer(help->helper, helper);
out:
- write_unlock_bh(&nf_conntrack_lock);
+ rcu_read_unlock();
}
EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply);
NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct);
NF_CT_ASSERT(skb);
- write_lock_bh(&nf_conntrack_lock);
+ spin_lock_bh(&nf_conntrack_lock);
/* Only update if this is not a fixed timeout */
- if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) {
- write_unlock_bh(&nf_conntrack_lock);
- return;
- }
+ if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status))
+ goto acct;
/* If not in hash table, timer will not be active yet */
if (!nf_ct_is_confirmed(ct)) {
}
}
+acct:
#ifdef CONFIG_NF_CT_ACCT
if (do_acct) {
ct->counters[CTINFO2DIR(ctinfo)].packets++;
}
#endif
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
/* must be unlocked when calling event cache */
if (event)
#include <linux/netfilter/nfnetlink_conntrack.h>
#include <linux/mutex.h>
-
/* Generic function for tcp/udp/sctp/dccp and alike. This needs to be
* in ip_conntrack_core, since we don't want the protocols to autoload
* or depend on ctnetlink */
-int nf_ct_port_tuple_to_nfattr(struct sk_buff *skb,
+int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb,
const struct nf_conntrack_tuple *tuple)
{
- NFA_PUT(skb, CTA_PROTO_SRC_PORT, sizeof(u_int16_t),
- &tuple->src.u.tcp.port);
- NFA_PUT(skb, CTA_PROTO_DST_PORT, sizeof(u_int16_t),
- &tuple->dst.u.tcp.port);
+ NLA_PUT_BE16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port);
+ NLA_PUT_BE16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port);
return 0;
-nfattr_failure:
+nla_put_failure:
return -1;
}
-EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nfattr);
+EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr);
-static const size_t cta_min_proto[CTA_PROTO_MAX] = {
- [CTA_PROTO_SRC_PORT-1] = sizeof(u_int16_t),
- [CTA_PROTO_DST_PORT-1] = sizeof(u_int16_t)
+const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = {
+ [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 },
+ [CTA_PROTO_DST_PORT] = { .type = NLA_U16 },
};
+EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy);
-int nf_ct_port_nfattr_to_tuple(struct nfattr *tb[],
+int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[],
struct nf_conntrack_tuple *t)
{
- if (!tb[CTA_PROTO_SRC_PORT-1] || !tb[CTA_PROTO_DST_PORT-1])
- return -EINVAL;
-
- if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
+ if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT])
return -EINVAL;
- t->src.u.tcp.port = *(__be16 *)NFA_DATA(tb[CTA_PROTO_SRC_PORT-1]);
- t->dst.u.tcp.port = *(__be16 *)NFA_DATA(tb[CTA_PROTO_DST_PORT-1]);
+ t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]);
+ t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]);
return 0;
}
-EXPORT_SYMBOL_GPL(nf_ct_port_nfattr_to_tuple);
+EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple);
#endif
/* Used by ipt_REJECT and ip6t_REJECT. */
-void __nf_conntrack_attach(struct sk_buff *nskb, struct sk_buff *skb)
+static void nf_conntrack_attach(struct sk_buff *nskb, struct sk_buff *skb)
{
struct nf_conn *ct;
enum ip_conntrack_info ctinfo;
nskb->nfctinfo = ctinfo;
nf_conntrack_get(nskb->nfct);
}
-EXPORT_SYMBOL_GPL(__nf_conntrack_attach);
static inline int
do_iter(const struct nf_conntrack_tuple_hash *i,
struct nf_conn *ct;
struct hlist_node *n;
- write_lock_bh(&nf_conntrack_lock);
+ spin_lock_bh(&nf_conntrack_lock);
for (; *bucket < nf_conntrack_htable_size; (*bucket)++) {
hlist_for_each_entry(h, n, &nf_conntrack_hash[*bucket], hnode) {
ct = nf_ct_tuplehash_to_ctrack(h);
if (iter(ct, data))
set_bit(IPS_DYING_BIT, &ct->status);
}
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
return NULL;
found:
atomic_inc(&ct->ct_general.use);
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
return ct;
}
return 1;
}
-void nf_ct_free_hashtable(struct hlist_head *hash, int vmalloced, int size)
+void nf_ct_free_hashtable(struct hlist_head *hash, int vmalloced, unsigned int size)
{
if (vmalloced)
vfree(hash);
nf_conntrack_expect_fini();
}
-struct hlist_head *nf_ct_alloc_hashtable(int *sizep, int *vmalloced)
+struct hlist_head *nf_ct_alloc_hashtable(unsigned int *sizep, int *vmalloced)
{
struct hlist_head *hash;
unsigned int size, i;
*vmalloced = 0;
size = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_head));
- hash = (void*)__get_free_pages(GFP_KERNEL,
+ hash = (void*)__get_free_pages(GFP_KERNEL|__GFP_NOWARN,
get_order(sizeof(struct hlist_head)
* size));
if (!hash) {
}
EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable);
-int set_hashsize(const char *val, struct kernel_param *kp)
+int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp)
{
- int i, bucket, hashsize, vmalloced;
- int old_vmalloced, old_size;
+ int i, bucket, vmalloced, old_vmalloced;
+ unsigned int hashsize, old_size;
int rnd;
struct hlist_head *hash, *old_hash;
struct nf_conntrack_tuple_hash *h;
if (!nf_conntrack_htable_size)
return param_set_uint(val, kp);
- hashsize = simple_strtol(val, NULL, 0);
+ hashsize = simple_strtoul(val, NULL, 0);
if (!hashsize)
return -EINVAL;
* use a newrandom seed */
get_random_bytes(&rnd, 4);
- write_lock_bh(&nf_conntrack_lock);
+ /* Lookups in the old hash might happen in parallel, which means we
+ * might get false negatives during connection lookup. New connections
+ * created because of a false negative won't make it into the hash
+ * though since that required taking the lock.
+ */
+ spin_lock_bh(&nf_conntrack_lock);
for (i = 0; i < nf_conntrack_htable_size; i++) {
while (!hlist_empty(&nf_conntrack_hash[i])) {
h = hlist_entry(nf_conntrack_hash[i].first,
struct nf_conntrack_tuple_hash, hnode);
- hlist_del(&h->hnode);
+ hlist_del_rcu(&h->hnode);
bucket = __hash_conntrack(&h->tuple, hashsize, rnd);
hlist_add_head(&h->hnode, &hash[bucket]);
}
nf_conntrack_vmalloc = vmalloced;
nf_conntrack_hash = hash;
nf_conntrack_hash_rnd = rnd;
- write_unlock_bh(&nf_conntrack_lock);
+ spin_unlock_bh(&nf_conntrack_lock);
nf_ct_free_hashtable(old_hash, old_vmalloced, old_size);
return 0;
}
+EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize);
-module_param_call(hashsize, set_hashsize, param_get_uint,
+module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
&nf_conntrack_htable_size, 0600);
int __init nf_conntrack_init(void)
goto out_fini_expect;
/* For use by REJECT target */
- rcu_assign_pointer(ip_ct_attach, __nf_conntrack_attach);
+ rcu_assign_pointer(ip_ct_attach, nf_conntrack_attach);
rcu_assign_pointer(nf_ct_destroy, destroy_conntrack);
/* Set up fake conntrack: