netfilter: ipt_CLUSTERIP: do not hold dev

It's a terrible thing to hold dev in iptables target. When the dev is
being removed, unregister_netdevice has to wait for the dev to become
free. dmesg will keep logging the err:

  kernel:unregister_netdevice: waiting for veth0_in to become free. \
  Usage count = 1

until iptables rules with this target are removed manually.

The worse thing is when deleting a netns, a virtual nic will be deleted
instead of reset to init_net in default_device_ops exit/exit_batch. As
it is earlier than to flush the iptables rules in iptable_filter_net_ops
exit, unregister_netdevice will block to wait for the nic to become free.

As unregister_netdevice is actually waiting for iptables rules flushing
while iptables rules have to be flushed after unregister_netdevice. This
'dead lock' will cause unregister_netdevice to block there forever. As
the netns is not available to operate at that moment, iptables rules can
not even be flushed manually either.

The reproducer can be:

  # ip netns add test
  # ip link add veth0_in type veth peer name veth0_out
  # ip link set veth0_in netns test
  # ip netns exec test ip link set lo up
  # ip netns exec test ip link set veth0_in up
  # ip netns exec test iptables -I INPUT -d 1.2.3.4 -i veth0_in -j \
    CLUSTERIP --new --clustermac 89:d4:47:eb:9a:fa --total-nodes 3 \
    --local-node 1 --hashmode sourceip-sourceport
  # ip netns del test

This issue can be triggered by all virtual nics with ipt_CLUSTERIP.

This patch is to fix it by not holding dev in ipt_CLUSTERIP, but saving
the dev->ifindex instead of the dev.

As Pablo Neira Ayuso's suggestion, it will refresh c->ifindex and dev's
mc by registering a netdevice notifier, just as what xt_TEE does. So it
removes the old codes updating dev's mc, and also no need to initialize
c->ifindex with dev->ifindex.

But as one config can be shared by more than one targets, and the netdev
notifier is per config, not per target. It couldn't get e->ip.iniface
in the notifier handler. So e->ip.iniface has to be saved into config.

Note that for backwards compatibility, this patch doesn't remove the
codes checking if the dev exists before creating a config.

v1->v2:
  - As Pablo Neira Ayuso's suggestion, register a netdevice notifier to
    manage c->ifindex and dev's mc.

Reported-by: Jianlin Shi <jishi@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
This commit is contained in:
Xin Long 2017-05-24 21:24:37 +08:00 committed by Pablo Neira Ayuso
parent 34158151d2
commit 202f59afd4
1 changed files with 73 additions and 28 deletions

View File

@ -47,7 +47,7 @@ struct clusterip_config {
__be32 clusterip; /* the IP address */ __be32 clusterip; /* the IP address */
u_int8_t clustermac[ETH_ALEN]; /* the MAC address */ u_int8_t clustermac[ETH_ALEN]; /* the MAC address */
struct net_device *dev; /* device */ int ifindex; /* device ifindex */
u_int16_t num_total_nodes; /* total number of nodes */ u_int16_t num_total_nodes; /* total number of nodes */
unsigned long local_nodes; /* node number array */ unsigned long local_nodes; /* node number array */
@ -57,6 +57,9 @@ struct clusterip_config {
enum clusterip_hashmode hash_mode; /* which hashing mode */ enum clusterip_hashmode hash_mode; /* which hashing mode */
u_int32_t hash_initval; /* hash initialization */ u_int32_t hash_initval; /* hash initialization */
struct rcu_head rcu; struct rcu_head rcu;
char ifname[IFNAMSIZ]; /* device ifname */
struct notifier_block notifier; /* refresh c->ifindex in it */
}; };
#ifdef CONFIG_PROC_FS #ifdef CONFIG_PROC_FS
@ -98,9 +101,8 @@ clusterip_config_put(struct clusterip_config *c)
* entry(rule) is removed, remove the config from lists, but don't free it * entry(rule) is removed, remove the config from lists, but don't free it
* yet, since proc-files could still be holding references */ * yet, since proc-files could still be holding references */
static inline void static inline void
clusterip_config_entry_put(struct clusterip_config *c) clusterip_config_entry_put(struct net *net, struct clusterip_config *c)
{ {
struct net *net = dev_net(c->dev);
struct clusterip_net *cn = net_generic(net, clusterip_net_id); struct clusterip_net *cn = net_generic(net, clusterip_net_id);
local_bh_disable(); local_bh_disable();
@ -109,8 +111,7 @@ clusterip_config_entry_put(struct clusterip_config *c)
spin_unlock(&cn->lock); spin_unlock(&cn->lock);
local_bh_enable(); local_bh_enable();
dev_mc_del(c->dev, c->clustermac); unregister_netdevice_notifier(&c->notifier);
dev_put(c->dev);
/* In case anyone still accesses the file, the open/close /* In case anyone still accesses the file, the open/close
* functions are also incrementing the refcount on their own, * functions are also incrementing the refcount on their own,
@ -170,19 +171,55 @@ clusterip_config_init_nodelist(struct clusterip_config *c,
set_bit(i->local_nodes[n] - 1, &c->local_nodes); set_bit(i->local_nodes[n] - 1, &c->local_nodes);
} }
static struct clusterip_config * static int
clusterip_config_init(const struct ipt_clusterip_tgt_info *i, __be32 ip, clusterip_netdev_event(struct notifier_block *this, unsigned long event,
struct net_device *dev) void *ptr)
{ {
struct net *net = dev_net(dev); struct net_device *dev = netdev_notifier_info_to_dev(ptr);
struct clusterip_config *c; struct clusterip_config *c;
c = container_of(this, struct clusterip_config, notifier);
switch (event) {
case NETDEV_REGISTER:
if (!strcmp(dev->name, c->ifname)) {
c->ifindex = dev->ifindex;
dev_mc_add(dev, c->clustermac);
}
break;
case NETDEV_UNREGISTER:
if (dev->ifindex == c->ifindex) {
dev_mc_del(dev, c->clustermac);
c->ifindex = -1;
}
break;
case NETDEV_CHANGENAME:
if (!strcmp(dev->name, c->ifname)) {
c->ifindex = dev->ifindex;
dev_mc_add(dev, c->clustermac);
} else if (dev->ifindex == c->ifindex) {
dev_mc_del(dev, c->clustermac);
c->ifindex = -1;
}
break;
}
return NOTIFY_DONE;
}
static struct clusterip_config *
clusterip_config_init(struct net *net, const struct ipt_clusterip_tgt_info *i,
__be32 ip, const char *iniface)
{
struct clusterip_net *cn = net_generic(net, clusterip_net_id); struct clusterip_net *cn = net_generic(net, clusterip_net_id);
struct clusterip_config *c;
int err;
c = kzalloc(sizeof(*c), GFP_ATOMIC); c = kzalloc(sizeof(*c), GFP_ATOMIC);
if (!c) if (!c)
return ERR_PTR(-ENOMEM); return ERR_PTR(-ENOMEM);
c->dev = dev; strcpy(c->ifname, iniface);
c->ifindex = -1;
c->clusterip = ip; c->clusterip = ip;
memcpy(&c->clustermac, &i->clustermac, ETH_ALEN); memcpy(&c->clustermac, &i->clustermac, ETH_ALEN);
c->num_total_nodes = i->num_total_nodes; c->num_total_nodes = i->num_total_nodes;
@ -213,17 +250,27 @@ clusterip_config_init(const struct ipt_clusterip_tgt_info *i, __be32 ip,
cn->procdir, cn->procdir,
&clusterip_proc_fops, c); &clusterip_proc_fops, c);
if (!c->pde) { if (!c->pde) {
err = -ENOMEM;
goto err;
}
}
#endif
c->notifier.notifier_call = clusterip_netdev_event;
err = register_netdevice_notifier(&c->notifier);
if (!err)
return c;
#ifdef CONFIG_PROC_FS
proc_remove(c->pde);
err:
#endif
spin_lock_bh(&cn->lock); spin_lock_bh(&cn->lock);
list_del_rcu(&c->list); list_del_rcu(&c->list);
spin_unlock_bh(&cn->lock); spin_unlock_bh(&cn->lock);
kfree(c); kfree(c);
return ERR_PTR(-ENOMEM); return ERR_PTR(err);
}
}
#endif
return c;
} }
#ifdef CONFIG_PROC_FS #ifdef CONFIG_PROC_FS
@ -425,15 +472,14 @@ static int clusterip_tg_check(const struct xt_tgchk_param *par)
e->ip.iniface); e->ip.iniface);
return -ENOENT; return -ENOENT;
} }
config = clusterip_config_init(cipinfo,
e->ip.dst.s_addr, dev);
if (IS_ERR(config)) {
dev_put(dev); dev_put(dev);
config = clusterip_config_init(par->net, cipinfo,
e->ip.dst.s_addr,
e->ip.iniface);
if (IS_ERR(config))
return PTR_ERR(config); return PTR_ERR(config);
} }
dev_mc_add(config->dev, config->clustermac);
}
} }
cipinfo->config = config; cipinfo->config = config;
@ -458,7 +504,7 @@ static void clusterip_tg_destroy(const struct xt_tgdtor_param *par)
/* if no more entries are referencing the config, remove it /* if no more entries are referencing the config, remove it
* from the list and destroy the proc entry */ * from the list and destroy the proc entry */
clusterip_config_entry_put(cipinfo->config); clusterip_config_entry_put(par->net, cipinfo->config);
clusterip_config_put(cipinfo->config); clusterip_config_put(cipinfo->config);
@ -558,10 +604,9 @@ arp_mangle(void *priv,
* addresses on different interfacs. However, in the CLUSTERIP case * addresses on different interfacs. However, in the CLUSTERIP case
* this wouldn't work, since we didn't subscribe the mcast group on * this wouldn't work, since we didn't subscribe the mcast group on
* other interfaces */ * other interfaces */
if (c->dev != state->out) { if (c->ifindex != state->out->ifindex) {
pr_debug("not mangling arp reply on different " pr_debug("not mangling arp reply on different interface: cip'%d'-skb'%d'\n",
"interface: cip'%s'-skb'%s'\n", c->ifindex, state->out->ifindex);
c->dev->name, state->out->name);
clusterip_config_put(c); clusterip_config_put(c);
return NF_ACCEPT; return NF_ACCEPT;
} }