[Patch net] ipt_CLUSTERIP: fix a race condition of proc file creation

There is a race condition between clusterip_config_entry_put()
and clusterip_config_init(), after we release the spinlock in
clusterip_config_entry_put(), a new proc file with a same IP could
be created immediately since it is already removed from the configs
list, therefore it triggers this warning:

------------[ cut here ]------------
proc_dir_entry 'ipt_CLUSTERIP/172.20.0.170' already registered
WARNING: CPU: 1 PID: 4152 at fs/proc/generic.c:330 proc_register+0x2a4/0x370 fs/proc/generic.c:329
Kernel panic - not syncing: panic_on_warn set ...

As a quick fix, just move the proc_remove() inside the spinlock.

Reported-by: <redacted>
Fixes: 6c5d5cfbe3c5 ("netfilter: ipt_CLUSTERIP: check duplicate config when initializing")
Tested-by: Paolo Abeni <pabeni@redhat.com>
Cc: Xin Long <lucien.xin@gmail.com>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Cong Wang <redacted>
---
 net/ipv4/netfilter/ipt_CLUSTERIP.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/net/ipv4/netfilter/ipt_CLUSTERIP.c b/net/ipv4/netfilter/ipt_CLUSTERIP.c
index 3a84a60f6b39..1ff72b87a066 100644
--- a/net/ipv4/netfilter/ipt_CLUSTERIP.c
+++ b/net/ipv4/netfilter/ipt_CLUSTERIP.c

@@ -107,12 +107,6 @@ clusterip_config_entry_put(struct net *net, struct clusterip_config *c)
 
 	local_bh_disable();
 	if (refcount_dec_and_lock(&c->entries, &cn->lock)) {
-		list_del_rcu(&c->list);
-		spin_unlock(&cn->lock);
-		local_bh_enable();
-
-		unregister_netdevice_notifier(&c->notifier);
-
 		/* In case anyone still accesses the file, the open/close
 		 * functions are also incrementing the refcount on their own,
 		 * so it's safe to remove the entry even if it's in use. */

@@ -120,6 +114,12 @@ clusterip_config_entry_put(struct net *net, struct clusterip_config *c)
 		if (cn->procdir)
 			proc_remove(c->pde);
 #endif
+		list_del_rcu(&c->list);
+		spin_unlock(&cn->lock);
+		local_bh_enable();
+
+		unregister_netdevice_notifier(&c->notifier);
+
 		return;
 	}
 	local_bh_enable();

-- 
2.13.0

In clusterip_config_find_get() we hold RCU read lock so it could
run concurrently with clusterip_config_entry_put(), as a result,
the refcnt could go back to 1 from 0, which leads to a double
list_del()... Just replace refcount_inc() with
refcount_inc_not_zero(), as for c->refcount.

Fixes: d73f33b16883 ("netfilter: CLUSTERIP: RCU conversion")
Cc: Eric Dumazet <redacted>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Cong Wang <redacted>
---
 net/ipv4/netfilter/ipt_CLUSTERIP.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/net/ipv4/netfilter/ipt_CLUSTERIP.c b/net/ipv4/netfilter/ipt_CLUSTERIP.c
index 1ff72b87a066..4537b1686c7c 100644
--- a/net/ipv4/netfilter/ipt_CLUSTERIP.c
+++ b/net/ipv4/netfilter/ipt_CLUSTERIP.c

@@ -154,8 +154,10 @@ clusterip_config_find_get(struct net *net, __be32 clusterip, int entry)
 #endif
 		if (unlikely(!refcount_inc_not_zero(&c->refcount)))
 			c = NULL;
-		else if (entry)
-			refcount_inc(&c->entries);
+		else if (entry) {
+			if (unlikely(!refcount_inc_not_zero(&c->entries)))
+				c = NULL;
+		}
 	}
 	rcu_read_unlock_bh();
 

-- 
2.13.0

Cong Wang [off-list ref] wrote:

this needs to call clusterip_config_put(c); too, else we leak one
reference.

Other than that this looks good.

On Thu, Feb 8, 2018 at 1:59 PM, Cong Wang [off-list ref] wrote:

Reviewed-by: Xin Long <lucien.xin@gmail.com>

On Wed, Feb 07, 2018 at 09:59:17PM -0800, Cong Wang wrote:

There is a race condition between clusterip_config_entry_put()
and clusterip_config_init(), after we release the spinlock in
clusterip_config_entry_put(), a new proc file with a same IP could
be created immediately since it is already removed from the configs
list, therefore it triggers this warning:

------------[ cut here ]------------
proc_dir_entry 'ipt_CLUSTERIP/172.20.0.170' already registered
WARNING: CPU: 1 PID: 4152 at fs/proc/generic.c:330 proc_register+0x2a4/0x370 fs/proc/generic.c:329
Kernel panic - not syncing: panic_on_warn set ...

As a quick fix, just move the proc_remove() inside the spinlock.

Applied, thanks.

On Thu, Feb 8, 2018 at 12:01 AM, Florian Westphal [off-list ref] wrote:

Cong Wang [off-list ref] wrote:

quoted

In clusterip_config_find_get() we hold RCU read lock so it could
run concurrently with clusterip_config_entry_put(), as a result,
the refcnt could go back to 1 from 0, which leads to a double
list_del()... Just replace refcount_inc() with
refcount_inc_not_zero(), as for c->refcount.

Fixes: d73f33b16883 ("netfilter: CLUSTERIP: RCU conversion")
Cc: Eric Dumazet <redacted>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Cong Wang <redacted>
---
 net/ipv4/netfilter/ipt_CLUSTERIP.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/net/ipv4/netfilter/ipt_CLUSTERIP.c b/net/ipv4/netfilter/ipt_CLUSTERIP.c
index 1ff72b87a066..4537b1686c7c 100644
--- a/net/ipv4/netfilter/ipt_CLUSTERIP.c
+++ b/net/ipv4/netfilter/ipt_CLUSTERIP.c

@@ -154,8 +154,10 @@ clusterip_config_find_get(struct net *net, __be32 clusterip, int entry)
 #endif
              if (unlikely(!refcount_inc_not_zero(&c->refcount)))
                      c = NULL;
-             else if (entry)
-                     refcount_inc(&c->entries);
+             else if (entry) {
+                     if (unlikely(!refcount_inc_not_zero(&c->entries)))

this needs to call clusterip_config_put(c); too, else we leak one
reference.

Other than that this looks good.

Right, good catch! I will send v2.