[PATCH net-next] rocker: fix a neigh entry leak issue

[PATCH net-next] rocker: fix a neigh entry leak issue

[Ying Xue]

Once we get a neighbour through looking up arp cache or creating a
new one in rocker_port_ipv4_resolve(), the neighbour's refcount is
already taken. But we don't put the refcount again after it's used,
which make the neighbour entry leaked.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
---
 drivers/net/ethernet/rocker/rocker.c |    1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/ethernet/rocker/rocker.c b/drivers/net/ethernet/rocker/rocker.c
index 1fc006b..7504540 100644
--- a/drivers/net/ethernet/rocker/rocker.c
+++ b/drivers/net/ethernet/rocker/rocker.c
@@ -3053,6 +3053,7 @@ static int rocker_port_ipv4_resolve(struct rocker_port *rocker_port,
 	else
 		neigh_event_send(n, NULL);
 
+	neigh_release(n);
 	return err;
 }
 
-- 
1.7.9.5

Re: [PATCH net-next] rocker: fix a neigh entry leak issue

[Jiri Pirko]

Thu, May 14, 2015 at 09:06:30AM CEST, ying.xue@windriver.com wrote:
>Once we get a neighbour through looking up arp cache or creating a
>new one in rocker_port_ipv4_resolve(), the neighbour's refcount is
>already taken. But we don't put the refcount again after it's used,
>which make the neighbour entry leaked.
>
>Signed-off-by: Ying Xue <ying.xue@windriver.com>

Acked-by: Jiri Pirko <jiri@resnulli.us>

Re: [PATCH net-next] rocker: fix a neigh entry leak issue

[Eric Dumazet]

On Thu, 2015-05-14 at 15:06 +0800, Ying Xue wrote:
> Once we get a neighbour through looking up arp cache or creating a
> new one in rocker_port_ipv4_resolve(), the neighbour's refcount is
> already taken. But we don't put the refcount again after it's used,
> which make the neighbour entry leaked.
> 
> Signed-off-by: Ying Xue <ying.xue@windriver.com>
> ---
>  drivers/net/ethernet/rocker/rocker.c |    1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/net/ethernet/rocker/rocker.c b/drivers/net/ethernet/rocker/rocker.c
> index 1fc006b..7504540 100644
> --- a/drivers/net/ethernet/rocker/rocker.c
> +++ b/drivers/net/ethernet/rocker/rocker.c
> @@ -3053,6 +3053,7 @@ static int rocker_port_ipv4_resolve(struct rocker_port *rocker_port,
>  	else
>  		neigh_event_send(n, NULL);
>  
> +	neigh_release(n);
>  	return err;
>  }
>  

I do not think this fix is complete.

neigh_create() can return an error, and we'll crash.

diff --git a/drivers/net/ethernet/rocker/rocker.c b/drivers/net/ethernet/rocker/rocker.c
index f0a9cb44be6bedc2171f6c9f79200cfa5d0bb0fa..ff0befbd8b4351c6f8fef93513967d5450d29e64 100644
--- a/drivers/net/ethernet/rocker/rocker.c
+++ b/drivers/net/ethernet/rocker/rocker.c
@@ -3037,11 +3037,11 @@ static int rocker_port_ipv4_resolve(struct rocker_port *rocker_port,
 	struct neighbour *n = __ipv4_neigh_lookup(dev, (__force u32)ip_addr);
 	int err = 0;
 
-	if (!n)
+	if (!n) {
 		n = neigh_create(&arp_tbl, &ip_addr, dev);
-	if (!n)
-		return -ENOMEM;
-
+		if (IS_ERR(n))
+			return PTR_ERR(n);
+	}
 	/* If the neigh is already resolved, then go ahead and
 	 * install the entry, otherwise start the ARP process to
 	 * resolve the neigh.

Re: [PATCH net-next] rocker: fix a neigh entry leak issue

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Thu, 14 May 2015 07:06:22 -0700

> On Thu, 2015-05-14 at 15:06 +0800, Ying Xue wrote:
>> Once we get a neighbour through looking up arp cache or creating a
>> new one in rocker_port_ipv4_resolve(), the neighbour's refcount is
>> already taken. But we don't put the refcount again after it's used,
>> which make the neighbour entry leaked.
>> 
>> Signed-off-by: Ying Xue <ying.xue@windriver.com>
>> ---
>>  drivers/net/ethernet/rocker/rocker.c |    1 +
>>  1 file changed, 1 insertion(+)
>> 
>> diff --git a/drivers/net/ethernet/rocker/rocker.c b/drivers/net/ethernet/rocker/rocker.c
>> index 1fc006b..7504540 100644
>> --- a/drivers/net/ethernet/rocker/rocker.c
>> +++ b/drivers/net/ethernet/rocker/rocker.c
>> @@ -3053,6 +3053,7 @@ static int rocker_port_ipv4_resolve(struct rocker_port *rocker_port,
>>  	else
>>  		neigh_event_send(n, NULL);
>>  
>> +	neigh_release(n);
>>  	return err;
>>  }
>>  
> 
> I do not think this fix is complete.
> 
> neigh_create() can return an error, and we'll crash.

Correct, neigh_create() return values must be checked with IS_ERR().