netfilter: nf_conntrack: use safer way to lock all buckets
authorSasha Levin <sasha.levin@oracle.com>
Tue, 19 Jan 2016 00:23:51 +0000 (19:23 -0500)
committerPablo Neira Ayuso <pablo@netfilter.org>
Wed, 20 Jan 2016 13:15:31 +0000 (14:15 +0100)
When we need to lock all buckets in the connection hashtable we'd attempt to
lock 1024 spinlocks, which is way more preemption levels than supported by
the kernel. Furthermore, this behavior was hidden by checking if lockdep is
enabled, and if it was - use only 8 buckets(!).

Fix this by using a global lock and synchronize all buckets on it when we
need to lock them all. This is pretty heavyweight, but is only done when we
need to resize the hashtable, and that doesn't happen often enough (or at all).

Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Reviewed-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
include/net/netfilter/nf_conntrack_core.h
net/netfilter/nf_conntrack_core.c
net/netfilter/nf_conntrack_helper.c
net/netfilter/nf_conntrack_netlink.c
net/netfilter/nfnetlink_cttimeout.c

index 788ef58a66b9b78807d164bce85060cc760ee336..62e17d1319ff7423dbcf176815f04cecfcdf3371 100644 (file)
@@ -79,12 +79,10 @@ print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
             const struct nf_conntrack_l3proto *l3proto,
             const struct nf_conntrack_l4proto *proto);
 
-#ifdef CONFIG_LOCKDEP
-# define CONNTRACK_LOCKS 8
-#else
-# define CONNTRACK_LOCKS 1024
-#endif
+#define CONNTRACK_LOCKS 1024
+
 extern spinlock_t nf_conntrack_locks[CONNTRACK_LOCKS];
+void nf_conntrack_lock(spinlock_t *lock);
 
 extern spinlock_t nf_conntrack_expect_lock;
 
index 3cb3cb831591ef79515b4bde66d5db825a732afb..58882de06bd74f254033e0f06521f264c18c8d44 100644 (file)
@@ -66,6 +66,21 @@ EXPORT_SYMBOL_GPL(nf_conntrack_locks);
 __cacheline_aligned_in_smp DEFINE_SPINLOCK(nf_conntrack_expect_lock);
 EXPORT_SYMBOL_GPL(nf_conntrack_expect_lock);
 
+static __read_mostly spinlock_t nf_conntrack_locks_all_lock;
+static __read_mostly bool nf_conntrack_locks_all;
+
+void nf_conntrack_lock(spinlock_t *lock) __acquires(lock)
+{
+       spin_lock(lock);
+       while (unlikely(nf_conntrack_locks_all)) {
+               spin_unlock(lock);
+               spin_lock(&nf_conntrack_locks_all_lock);
+               spin_unlock(&nf_conntrack_locks_all_lock);
+               spin_lock(lock);
+       }
+}
+EXPORT_SYMBOL_GPL(nf_conntrack_lock);
+
 static void nf_conntrack_double_unlock(unsigned int h1, unsigned int h2)
 {
        h1 %= CONNTRACK_LOCKS;
@@ -82,12 +97,12 @@ static bool nf_conntrack_double_lock(struct net *net, unsigned int h1,
        h1 %= CONNTRACK_LOCKS;
        h2 %= CONNTRACK_LOCKS;
        if (h1 <= h2) {
-               spin_lock(&nf_conntrack_locks[h1]);
+               nf_conntrack_lock(&nf_conntrack_locks[h1]);
                if (h1 != h2)
                        spin_lock_nested(&nf_conntrack_locks[h2],
                                         SINGLE_DEPTH_NESTING);
        } else {
-               spin_lock(&nf_conntrack_locks[h2]);
+               nf_conntrack_lock(&nf_conntrack_locks[h2]);
                spin_lock_nested(&nf_conntrack_locks[h1],
                                 SINGLE_DEPTH_NESTING);
        }
@@ -102,16 +117,19 @@ static void nf_conntrack_all_lock(void)
 {
        int i;
 
-       for (i = 0; i < CONNTRACK_LOCKS; i++)
-               spin_lock_nested(&nf_conntrack_locks[i], i);
+       spin_lock(&nf_conntrack_locks_all_lock);
+       nf_conntrack_locks_all = true;
+
+       for (i = 0; i < CONNTRACK_LOCKS; i++) {
+               spin_lock(&nf_conntrack_locks[i]);
+               spin_unlock(&nf_conntrack_locks[i]);
+       }
 }
 
 static void nf_conntrack_all_unlock(void)
 {
-       int i;
-
-       for (i = 0; i < CONNTRACK_LOCKS; i++)
-               spin_unlock(&nf_conntrack_locks[i]);
+       nf_conntrack_locks_all = false;
+       spin_unlock(&nf_conntrack_locks_all_lock);
 }
 
 unsigned int nf_conntrack_htable_size __read_mostly;
@@ -757,7 +775,7 @@ restart:
        hash = hash_bucket(_hash, net);
        for (; i < net->ct.htable_size; i++) {
                lockp = &nf_conntrack_locks[hash % CONNTRACK_LOCKS];
-               spin_lock(lockp);
+               nf_conntrack_lock(lockp);
                if (read_seqcount_retry(&net->ct.generation, sequence)) {
                        spin_unlock(lockp);
                        goto restart;
@@ -1382,7 +1400,7 @@ get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data),
        for (; *bucket < net->ct.htable_size; (*bucket)++) {
                lockp = &nf_conntrack_locks[*bucket % CONNTRACK_LOCKS];
                local_bh_disable();
-               spin_lock(lockp);
+               nf_conntrack_lock(lockp);
                if (*bucket < net->ct.htable_size) {
                        hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) {
                                if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
index bd9d31537905e4e372da427ac29ce5948f2af34c..3b40ec575cd56a7b73c0b25d9c8decdda979d842 100644 (file)
@@ -425,7 +425,7 @@ static void __nf_conntrack_helper_unregister(struct nf_conntrack_helper *me,
        }
        local_bh_disable();
        for (i = 0; i < net->ct.htable_size; i++) {
-               spin_lock(&nf_conntrack_locks[i % CONNTRACK_LOCKS]);
+               nf_conntrack_lock(&nf_conntrack_locks[i % CONNTRACK_LOCKS]);
                if (i < net->ct.htable_size) {
                        hlist_nulls_for_each_entry(h, nn, &net->ct.hash[i], hnnode)
                                unhelp(h, me);
index dbb1bb3edb456594184bb211332d1ee799af70ef..355e8552fd5b77682295493a22dc0a03d9338255 100644 (file)
@@ -840,7 +840,7 @@ ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
        for (; cb->args[0] < net->ct.htable_size; cb->args[0]++) {
 restart:
                lockp = &nf_conntrack_locks[cb->args[0] % CONNTRACK_LOCKS];
-               spin_lock(lockp);
+               nf_conntrack_lock(lockp);
                if (cb->args[0] >= net->ct.htable_size) {
                        spin_unlock(lockp);
                        goto out;
index 5d010f27ac018d079058000c11acc66d7086b57a..94837d236ab0e9a2fc3baa5d8fe7688a1b399afc 100644 (file)
@@ -307,12 +307,12 @@ static void ctnl_untimeout(struct net *net, struct ctnl_timeout *timeout)
 
        local_bh_disable();
        for (i = 0; i < net->ct.htable_size; i++) {
-               spin_lock(&nf_conntrack_locks[i % CONNTRACK_LOCKS]);
+               nf_conntrack_lock(&nf_conntrack_locks[i % CONNTRACK_LOCKS]);
                if (i < net->ct.htable_size) {
                        hlist_nulls_for_each_entry(h, nn, &net->ct.hash[i], hnnode)
                                untimeout(h, timeout);
                }
-               spin_unlock(&nf_conntrack_locks[i % CONNTRACK_LOCKS]);
+               nf_conntrack_lock(&nf_conntrack_locks[i % CONNTRACK_LOCKS]);
        }
        local_bh_enable();
 }