Re: [patch net/stable] br: fix use of ->rx_handler_data in code executed on non-rx_handler path

[Jiri Pirko <jiri@resnulli.us>]

Thu, Dec 05, 2013 at 03:58:42PM CET, mst@redhat.com wrote:
>On Thu, Dec 05, 2013 at 03:50:25PM +0100, Jiri Pirko wrote:
>> br_stp_rcv() is reached by non-rx_handler path. That means there is no
>> guarantee that dev is bridge port and therefore simple NULL check of
>> ->rx_handler_data is not enough. There is need to check if dev is really
>> bridge port and since only rcu read lock is held here, do it by checking
>> ->rx_handler pointer.
>> 
>> Note that synchronize_net() in netdev_rx_handler_unregister() ensures
>> this approach as valid.
>> 
>> Introduced originally by:
>> commit f350a0a87374418635689471606454abc7beaa3a
>>   "bridge: use rx_handler_data pointer to store net_bridge_port pointer"
>> 
>> Fixed but not in the best way by:
>> commit b5ed54e94d324f17c97852296d61a143f01b227a
>>   "bridge: fix RCU races with bridge port"
>> 
>> Reintroduced by:
>> commit 716ec052d2280d511e10e90ad54a86f5b5d4dcc2
>>   "bridge: fix NULL pointer deref of br_port_get_rcu"
>> 
>> Please apply to stable trees as well. Thanks.
>> 
>> Reported-by: Laine Stump <laine@redhat.com>
>> Signed-off-by: Jiri Pirko <jiri@resnulli.us>
>
>I would also add:
>https://bugzilla.redhat.com/show_bug.cgi?id=1025770
>
>> ---
>>  net/bridge/br_private.h  | 12 ++++++++++++
>>  net/bridge/br_stp_bpdu.c |  2 +-
>>  2 files changed, 13 insertions(+), 1 deletion(-)
>> 
>> diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
>> index 229d820..67a2d4b 100644
>> --- a/net/bridge/br_private.h
>> +++ b/net/bridge/br_private.h
>> @@ -204,6 +204,13 @@ static inline struct net_bridge_port *br_port_get_rcu(const struct net_device *d
>>  	return rcu_dereference(dev->rx_handler_data);
>>  }
>>  
>> +static inline bool br_rx_handler_check_rcu(const struct net_device *dev);
>
>Can't we reorder functions?
>Forward-declaring it like this is ugly.

I know. But makes more sense to have these getters here, near the
struct. I originally moved the definitions near br_handle_frame_finish
definition but I think that it is nicer to do it this way...

>
>> +
>> +static inline struct net_bridge_port *br_port_get_check_rcu(const struct net_device *dev)
>> +{
>> +	return br_rx_handler_check_rcu(dev) ? br_port_get_rcu(dev) : NULL;
>> +}
>> +
>>  static inline struct net_bridge_port *br_port_get_rtnl(const struct net_device *dev)
>>  {
>>  	return br_port_exists(dev) ?
>> @@ -426,6 +433,11 @@ netdev_features_t br_features_recompute(struct net_bridge *br,
>>  int br_handle_frame_finish(struct sk_buff *skb);
>>  rx_handler_result_t br_handle_frame(struct sk_buff **pskb);
>>  
>> +static inline bool br_rx_handler_check_rcu(const struct net_device *dev)
>> +{
>> +	return rcu_dereference(dev->rx_handler) == br_handle_frame;
>> +}
>> +
>>  /* br_ioctl.c */
>>  int br_dev_ioctl(struct net_device *dev, struct ifreq *rq, int cmd);
>>  int br_ioctl_deviceless_stub(struct net *net, unsigned int cmd,
>> diff --git a/net/bridge/br_stp_bpdu.c b/net/bridge/br_stp_bpdu.c
>> index 8660ea3..bdb459d 100644
>> --- a/net/bridge/br_stp_bpdu.c
>> +++ b/net/bridge/br_stp_bpdu.c
>> @@ -153,7 +153,7 @@ void br_stp_rcv(const struct stp_proto *proto, struct sk_buff *skb,
>>  	if (buf[0] != 0 || buf[1] != 0 || buf[2] != 0)
>>  		goto err;
>>  
>> -	p = br_port_get_rcu(dev);
>> +	p = br_port_get_check_rcu(dev);
>>  	if (!p)
>>  		goto err;
>>  
>> -- 
>> 1.8.3.1