* [PATCH] bridge: Fix possible races between assigning rx_handler_data and setting IFF_BRIDGE_PORT bit
@ 2021-04-28 7:29 zhangzhengming
2021-04-28 8:00 ` Nikolay Aleksandrov
0 siblings, 1 reply; 2+ messages in thread
From: zhangzhengming @ 2021-04-28 7:29 UTC (permalink / raw)
To: roopa, nikolay, davem, kuba
Cc: bridge, netdev, zhangzhengming, wangxiaogang3, xuhanbing
From: Zhang Zhengming <zhangzhengming@huawei.com>
There is a crash in the function br_get_link_af_size_filtered,
as the port_exists(dev) is true and the rx_handler_data of dev is NULL.
But the rx_handler_data of dev is correct saved in vmcore.
The oops looks something like:
...
pc : br_get_link_af_size_filtered+0x28/0x1c8 [bridge]
...
Call trace:
br_get_link_af_size_filtered+0x28/0x1c8 [bridge]
if_nlmsg_size+0x180/0x1b0
rtnl_calcit.isra.12+0xf8/0x148
rtnetlink_rcv_msg+0x334/0x370
netlink_rcv_skb+0x64/0x130
rtnetlink_rcv+0x28/0x38
netlink_unicast+0x1f0/0x250
netlink_sendmsg+0x310/0x378
sock_sendmsg+0x4c/0x70
__sys_sendto+0x120/0x150
__arm64_sys_sendto+0x30/0x40
el0_svc_common+0x78/0x130
el0_svc_handler+0x38/0x78
el0_svc+0x8/0xc
In br_add_if(), we found there is no guarantee that
assigning rx_handler_data to dev->rx_handler_data
will before setting the IFF_BRIDGE_PORT bit of priv_flags.
So there is a possible data competition:
CPU 0: CPU 1:
(RCU read lock) (RTNL lock)
rtnl_calcit() br_add_slave()
if_nlmsg_size() br_add_if()
br_get_link_af_size_filtered() -> netdev_rx_handler_register
...
// The order is not guaranteed
... -> dev->priv_flags |= IFF_BRIDGE_PORT;
// The IFF_BRIDGE_PORT bit of priv_flags has been set
-> if (br_port_exists(dev)) {
// The dev->rx_handler_data has NOT been assigned
-> p = br_port_get_rcu(dev);
....
-> rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
...
Fix this by adding memory barrier instruction to ensure the order.
Signed-off-by: Zhang Zhengming <zhangzhengming@huawei.com>
Reviewed-by: Zhao Lei <zhaolei69@huawei.com>
Reviewed-by: Wang Xiaogang <wangxiaogang3@huawei.com>
---
net/bridge/br_if.c | 6 ++++++
net/bridge/br_netlink.c | 1 +
2 files changed, 7 insertions(+)
diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c
index f7d2f47..42895be 100644
--- a/net/bridge/br_if.c
+++ b/net/bridge/br_if.c
@@ -636,6 +636,12 @@ int br_add_if(struct net_bridge *br, struct net_device *dev,
err = netdev_rx_handler_register(dev, br_get_rx_handler(dev), p);
if (err)
goto err4;
+
+ /* Make sure dev->rx_handler_data is written in netdev_rx_handler_register
+ * before the bit IFF_BRIDGE_PORT of dev->priv_flags is set.
+ * coupled with smp_rmb() in br_get_link_af_size_filtered.
+ */
+ smp_wmb();
dev->priv_flags |= IFF_BRIDGE_PORT;
diff --git a/net/bridge/br_netlink.c b/net/bridge/br_netlink.c
index f2b1343..ccc1fd7 100644
--- a/net/bridge/br_netlink.c
+++ b/net/bridge/br_netlink.c
@@ -103,6 +103,7 @@ static size_t br_get_link_af_size_filtered(const struct net_device *dev,
rcu_read_lock();
if (netif_is_bridge_port(dev)) {
+ smp_rmb(); /* coupled with smp_wmb() in br_add_if() */
p = br_port_get_rcu(dev);
vg = nbp_vlan_group_rcu(p);
} else if (dev->priv_flags & IFF_EBRIDGE) {
--
2.7.4
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH] bridge: Fix possible races between assigning rx_handler_data and setting IFF_BRIDGE_PORT bit
2021-04-28 7:29 [PATCH] bridge: Fix possible races between assigning rx_handler_data and setting IFF_BRIDGE_PORT bit zhangzhengming
@ 2021-04-28 8:00 ` Nikolay Aleksandrov
0 siblings, 0 replies; 2+ messages in thread
From: Nikolay Aleksandrov @ 2021-04-28 8:00 UTC (permalink / raw)
To: zhangzhengming, roopa, davem, kuba
Cc: bridge, netdev, wangxiaogang3, xuhanbing
On 28/04/2021 10:29, zhangzhengming wrote:
> From: Zhang Zhengming <zhangzhengming@huawei.com>
>
> There is a crash in the function br_get_link_af_size_filtered,
> as the port_exists(dev) is true and the rx_handler_data of dev is NULL.
> But the rx_handler_data of dev is correct saved in vmcore.
>
> The oops looks something like:
> ...
> pc : br_get_link_af_size_filtered+0x28/0x1c8 [bridge]
> ...
> Call trace:
> br_get_link_af_size_filtered+0x28/0x1c8 [bridge]
> if_nlmsg_size+0x180/0x1b0
> rtnl_calcit.isra.12+0xf8/0x148
> rtnetlink_rcv_msg+0x334/0x370
> netlink_rcv_skb+0x64/0x130
> rtnetlink_rcv+0x28/0x38
> netlink_unicast+0x1f0/0x250
> netlink_sendmsg+0x310/0x378
> sock_sendmsg+0x4c/0x70
> __sys_sendto+0x120/0x150
> __arm64_sys_sendto+0x30/0x40
> el0_svc_common+0x78/0x130
> el0_svc_handler+0x38/0x78
> el0_svc+0x8/0xc
>
> In br_add_if(), we found there is no guarantee that
> assigning rx_handler_data to dev->rx_handler_data
> will before setting the IFF_BRIDGE_PORT bit of priv_flags.
> So there is a possible data competition:
>
> CPU 0: CPU 1:
> (RCU read lock) (RTNL lock)
> rtnl_calcit() br_add_slave()
> if_nlmsg_size() br_add_if()
> > -> netdev_rx_handler_register
> ...
> // The order is not guaranteed
> ... -> dev->priv_flags |= IFF_BRIDGE_PORT;
> // The IFF_BRIDGE_PORT bit of priv_flags has been set
> -> if (br_port_exists(dev)) {
> // The dev->rx_handler_data has NOT been assigned
> -> p = br_port_get_rcu(dev);
> ....
> -> rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
> ...
>
> Fix this by adding memory barrier instruction to ensure the order.
>
> Signed-off-by: Zhang Zhengming <zhangzhengming@huawei.com>
> Reviewed-by: Zhao Lei <zhaolei69@huawei.com>
> Reviewed-by: Wang Xiaogang <wangxiaogang3@huawei.com>
> ---
> net/bridge/br_if.c | 6 ++++++
> net/bridge/br_netlink.c | 1 +
> 2 files changed, 7 insertions(+)
>
No need for memory barriers, just use br_port_get_check_rcu() in br_get_link_af_size_filtered()
and check the return value. It will be much simpler and shorter.
Thanks
> diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c
> index f7d2f47..42895be 100644
> --- a/net/bridge/br_if.c
> +++ b/net/bridge/br_if.c
> @@ -636,6 +636,12 @@ int br_add_if(struct net_bridge *br, struct net_device *dev,
> err = netdev_rx_handler_register(dev, br_get_rx_handler(dev), p);
> if (err)
> goto err4;
> +
> + /* Make sure dev->rx_handler_data is written in netdev_rx_handler_register
> + * before the bit IFF_BRIDGE_PORT of dev->priv_flags is set.
> + * coupled with smp_rmb() in br_get_link_af_size_filtered.
> + */
> + smp_wmb();
>
> dev->priv_flags |= IFF_BRIDGE_PORT;
>
> diff --git a/net/bridge/br_netlink.c b/net/bridge/br_netlink.c
> index f2b1343..ccc1fd7 100644
> --- a/net/bridge/br_netlink.c
> +++ b/net/bridge/br_netlink.c
> @@ -103,6 +103,7 @@ static size_t br_get_link_af_size_filtered(const struct net_device *dev,
>
> rcu_read_lock();
> if (netif_is_bridge_port(dev)) {
> + smp_rmb(); /* coupled with smp_wmb() in br_add_if() */
> p = br_port_get_rcu(dev);
> vg = nbp_vlan_group_rcu(p);
> } else if (dev->priv_flags & IFF_EBRIDGE) {
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2021-04-28 8:00 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-04-28 7:29 [PATCH] bridge: Fix possible races between assigning rx_handler_data and setting IFF_BRIDGE_PORT bit zhangzhengming
2021-04-28 8:00 ` Nikolay Aleksandrov
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).