IPv6: convert addrconf hash list to RCU
Convert from reader/writer lock to RCU and spinlock for addrconf
hash list.
Adds an additional helper macro for hlist_for_each_entry_continue_rcu
to handle the continue case.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/include/linux/rculist.h b/include/linux/rculist.h
index 2c9b46c..004908b 100644
--- a/include/linux/rculist.h
+++ b/include/linux/rculist.h
@@ -428,5 +428,18 @@
({ tpos = hlist_entry(pos, typeof(*tpos), member); 1; }); \
pos = rcu_dereference_raw(pos->next))
+/**
+ * hlist_for_each_entry_continue_rcu - iterate over a hlist continuing after current point
+ * @tpos: the type * to use as a loop cursor.
+ * @pos: the &struct hlist_node to use as a loop cursor.
+ * @member: the name of the hlist_node within the struct.
+ */
+#define hlist_for_each_entry_continue_rcu(tpos, pos, member) \
+ for (pos = rcu_dereference((pos)->next); \
+ pos && ({ prefetch(pos->next); 1; }) && \
+ ({ tpos = hlist_entry(pos, typeof(*tpos), member); 1; }); \
+ pos = rcu_dereference(pos->next))
+
+
#endif /* __KERNEL__ */
#endif
diff --git a/include/net/if_inet6.h b/include/net/if_inet6.h
index 4984015..22a00b1 100644
--- a/include/net/if_inet6.h
+++ b/include/net/if_inet6.h
@@ -64,6 +64,7 @@
#endif
int dead;
+ struct rcu_head rcu;
};
struct ip6_sf_socklist {
diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
index 0488b9f..7ffd5eea 100644
--- a/net/ipv6/addrconf.c
+++ b/net/ipv6/addrconf.c
@@ -127,7 +127,7 @@
* Configured unicast address hash table
*/
static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
-static DEFINE_RWLOCK(addrconf_hash_lock);
+static DEFINE_SPINLOCK(addrconf_hash_lock);
static void addrconf_verify(unsigned long);
@@ -523,8 +523,13 @@
}
#endif
-/* Nobody refers to this ifaddr, destroy it */
+static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
+{
+ struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
+ kfree(ifp);
+}
+/* Nobody refers to this ifaddr, destroy it */
void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
{
WARN_ON(ifp->if_next != NULL);
@@ -545,7 +550,7 @@
}
dst_release(&ifp->rt->u.dst);
- kfree(ifp);
+ call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
}
static void
@@ -616,7 +621,7 @@
goto out2;
}
- write_lock(&addrconf_hash_lock);
+ spin_lock(&addrconf_hash_lock);
/* Ignore adding duplicate addresses on an interface */
if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
@@ -670,9 +675,9 @@
/* Add to big hash table */
hash = ipv6_addr_hash(addr);
- hlist_add_head(&ifa->addr_lst, &inet6_addr_lst[hash]);
+ hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
in6_ifa_hold(ifa);
- write_unlock(&addrconf_hash_lock);
+ spin_unlock(&addrconf_hash_lock);
write_lock(&idev->lock);
/* Add to inet6_dev unicast addr list. */
@@ -699,7 +704,7 @@
return ifa;
out:
- write_unlock(&addrconf_hash_lock);
+ spin_unlock(&addrconf_hash_lock);
goto out2;
}
@@ -717,10 +722,10 @@
ifp->dead = 1;
- write_lock_bh(&addrconf_hash_lock);
- hlist_del_init(&ifp->addr_lst);
+ spin_lock_bh(&addrconf_hash_lock);
+ hlist_del_init_rcu(&ifp->addr_lst);
__in6_ifa_put(ifp);
- write_unlock_bh(&addrconf_hash_lock);
+ spin_unlock_bh(&addrconf_hash_lock);
write_lock_bh(&idev->lock);
#ifdef CONFIG_IPV6_PRIVACY
@@ -1274,8 +1279,8 @@
struct hlist_node *node;
u8 hash = ipv6_addr_hash(addr);
- read_lock_bh(&addrconf_hash_lock);
- hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
+ rcu_read_lock_bh();
+ hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
if (!net_eq(dev_net(ifp->idev->dev), net))
continue;
if (ipv6_addr_equal(&ifp->addr, addr) &&
@@ -1285,7 +1290,8 @@
break;
}
}
- read_unlock_bh(&addrconf_hash_lock);
+ rcu_read_unlock_bh();
+
return ifp != NULL;
}
EXPORT_SYMBOL(ipv6_chk_addr);
@@ -1341,8 +1347,8 @@
struct hlist_node *node;
u8 hash = ipv6_addr_hash(addr);
- read_lock_bh(&addrconf_hash_lock);
- hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
+ rcu_read_lock_bh();
+ hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
if (!net_eq(dev_net(ifp->idev->dev), net))
continue;
if (ipv6_addr_equal(&ifp->addr, addr)) {
@@ -1353,7 +1359,7 @@
}
}
}
- read_unlock_bh(&addrconf_hash_lock);
+ rcu_read_unlock_bh();
return ifp;
}
@@ -2698,10 +2704,10 @@
write_unlock_bh(&idev->lock);
/* clear hash table */
- write_lock_bh(&addrconf_hash_lock);
- hlist_del_init(&ifa->addr_lst);
+ spin_lock_bh(&addrconf_hash_lock);
+ hlist_del_init_rcu(&ifa->addr_lst);
__in6_ifa_put(ifa);
- write_unlock_bh(&addrconf_hash_lock);
+ spin_unlock_bh(&addrconf_hash_lock);
__ipv6_ifa_notify(RTM_DELADDR, ifa);
atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
@@ -2946,11 +2952,10 @@
for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
struct hlist_node *n;
- hlist_for_each_entry(ifa, n,
- &inet6_addr_lst[state->bucket], addr_lst) {
+ hlist_for_each_entry_rcu(ifa, n, &inet6_addr_lst[state->bucket],
+ addr_lst)
if (net_eq(dev_net(ifa->idev->dev), net))
return ifa;
- }
}
return NULL;
}
@@ -2962,10 +2967,9 @@
struct net *net = seq_file_net(seq);
struct hlist_node *n = &ifa->addr_lst;
- hlist_for_each_entry_continue(ifa, n, addr_lst) {
+ hlist_for_each_entry_continue_rcu(ifa, n, addr_lst)
if (net_eq(dev_net(ifa->idev->dev), net))
return ifa;
- }
while (++state->bucket < IN6_ADDR_HSIZE) {
hlist_for_each_entry(ifa, n,
@@ -2989,9 +2993,9 @@
}
static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
- __acquires(addrconf_hash_lock)
+ __acquires(rcu)
{
- read_lock_bh(&addrconf_hash_lock);
+ rcu_read_lock_bh();
return if6_get_idx(seq, *pos);
}
@@ -3005,9 +3009,9 @@
}
static void if6_seq_stop(struct seq_file *seq, void *v)
- __releases(addrconf_hash_lock)
+ __releases(rcu)
{
- read_unlock_bh(&addrconf_hash_lock);
+ rcu_read_unlock_bh();
}
static int if6_seq_show(struct seq_file *seq, void *v)
@@ -3081,8 +3085,8 @@
struct hlist_node *n;
u8 hash = ipv6_addr_hash(addr);
- read_lock_bh(&addrconf_hash_lock);
- hlist_for_each_entry(ifp, n, &inet6_addr_lst[hash], addr_lst) {
+ rcu_read_lock_bh();
+ hlist_for_each_entry_rcu(ifp, n, &inet6_addr_lst[hash], addr_lst) {
if (!net_eq(dev_net(ifp->idev->dev), net))
continue;
if (ipv6_addr_equal(&ifp->addr, addr) &&
@@ -3091,7 +3095,7 @@
break;
}
}
- read_unlock_bh(&addrconf_hash_lock);
+ rcu_read_unlock_bh();
return ret;
}
#endif
@@ -3107,7 +3111,8 @@
unsigned long now, next;
int i;
- spin_lock_bh(&addrconf_verify_lock);
+ rcu_read_lock_bh();
+ spin_lock(&addrconf_verify_lock);
now = jiffies;
next = now + ADDR_CHECK_FREQUENCY;
@@ -3116,8 +3121,8 @@
for (i=0; i < IN6_ADDR_HSIZE; i++) {
restart:
- read_lock(&addrconf_hash_lock);
- hlist_for_each_entry(ifp, node, &inet6_addr_lst[i], addr_lst) {
+ hlist_for_each_entry_rcu(ifp, node,
+ &inet6_addr_lst[i], addr_lst) {
unsigned long age;
#ifdef CONFIG_IPV6_PRIVACY
unsigned long regen_advance;
@@ -3139,7 +3144,6 @@
age >= ifp->valid_lft) {
spin_unlock(&ifp->lock);
in6_ifa_hold(ifp);
- read_unlock(&addrconf_hash_lock);
ipv6_del_addr(ifp);
goto restart;
} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
@@ -3161,7 +3165,6 @@
if (deprecate) {
in6_ifa_hold(ifp);
- read_unlock(&addrconf_hash_lock);
ipv6_ifa_notify(0, ifp);
in6_ifa_put(ifp);
@@ -3179,7 +3182,7 @@
in6_ifa_hold(ifp);
in6_ifa_hold(ifpub);
spin_unlock(&ifp->lock);
- read_unlock(&addrconf_hash_lock);
+
spin_lock(&ifpub->lock);
ifpub->regen_count = 0;
spin_unlock(&ifpub->lock);
@@ -3199,12 +3202,12 @@
spin_unlock(&ifp->lock);
}
}
- read_unlock(&addrconf_hash_lock);
}
addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
add_timer(&addr_chk_timer);
- spin_unlock_bh(&addrconf_verify_lock);
+ spin_unlock(&addrconf_verify_lock);
+ rcu_read_unlock_bh();
}
static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
@@ -4621,10 +4624,10 @@
/*
* Check hash table.
*/
- write_lock_bh(&addrconf_hash_lock);
+ spin_lock_bh(&addrconf_hash_lock);
for (i = 0; i < IN6_ADDR_HSIZE; i++)
WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
- write_unlock_bh(&addrconf_hash_lock);
+ spin_unlock_bh(&addrconf_hash_lock);
del_timer(&addr_chk_timer);
rtnl_unlock();