Stephen Hemminger
2013-Dec-05 16:55 UTC
[Bridge] [patch net/stable v2] br: fix use of ->rx_handler_data in code executed on non-rx_handler path
On Thu, 5 Dec 2013 16:27:37 +0100 Jiri Pirko <jiri at resnulli.us> 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. >I think this patch is simpler/better, it restores the old logic. Ps. submitting patches to bugzilla is a good way to have them ignored. From Stephen Hemminger <stephen at networkplumber.org> Check that incoming STP packet is received on a port assigned to bridge before processing. It is possible to receive packet on non-bridge port because they are multicast. See: https://bugzilla.kernel.org/show_bug.cgi?id=64911 Regression introduced by: commit 716ec052d2280d511e10e90ad54a86f5b5d4dcc2 Author: Hong Zhiguo <zhiguohong at tencent.com> Date: Sat Sep 14 22:42:28 2013 +0800 bridge: fix NULL pointer deref of br_port_get_rcu Reported-by: Alexander Y. Fomichev Signed-off-by: Stephen Hemminger <stephen at networkplumber.org> --- a/net/bridge/br_stp_bpdu.c 2013-06-11 09:50:21.522919061 -0700 +++ b/net/bridge/br_stp_bpdu.c 2013-12-05 08:46:56.090463702 -0800 @@ -153,6 +153,9 @@ void br_stp_rcv(const struct stp_proto * if (buf[0] != 0 || buf[1] != 0 || buf[2] != 0) goto err; + if (!br_port_exists(dev)) + goto err; + p = br_port_get_rcu(dev); if (!p) goto err;
Gao feng
2013-Dec-06 02:26 UTC
Re: [patch net/stable v2] br: fix use of ->rx_handler_data in code executed on non-rx_handler path
On 12/06/2013 12:55 AM, Stephen Hemminger wrote:> On Thu, 5 Dec 2013 16:27:37 +0100 > Jiri Pirko <jiri@resnulli.us> 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. >> > > > I think this patch is simpler/better, it restores the old logic. > > Ps. submitting patches to bugzilla is a good way to have them ignored. > >>From Stephen Hemminger <stephen@networkplumber.org> > > Check that incoming STP packet is received on a port assigned to bridge > before processing. It is possible to receive packet on non-bridge port > because they are multicast. > > See: > https://bugzilla.kernel.org/show_bug.cgi?id=64911 > > > Regression introduced by: > commit 716ec052d2280d511e10e90ad54a86f5b5d4dcc2 > Author: Hong Zhiguo <zhiguohong@tencent.com> > Date: Sat Sep 14 22:42:28 2013 +0800 > > bridge: fix NULL pointer deref of br_port_get_rcu > > > Reported-by: Alexander Y. Fomichev > Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> > > > --- a/net/bridge/br_stp_bpdu.c 2013-06-11 09:50:21.522919061 -0700 > +++ b/net/bridge/br_stp_bpdu.c 2013-12-05 08:46:56.090463702 -0800 > @@ -153,6 +153,9 @@ void br_stp_rcv(const struct stp_proto * > if (buf[0] != 0 || buf[1] != 0 || buf[2] != 0) > goto err; > > + if (!br_port_exists(dev)) > + goto err; > + > p = br_port_get_rcu(dev); > if (!p) > goto err;We alreay did some cleanup jobs before mark this dev is not a port of bridge (dev->priv_flags &= ~IFF_BRIDGE_PORT), such as remove the fdb related to this port(br_fdb_delete_by_port). and seems like after these cleanup jobs, before unregister this device, if new skb is received, br_handle_local_finish will call br_fdb_update to create a new fdb whose dst points to the will-be-destroied-port. I don''t know if this will cause some problems. seems we should also make sure port is unavailable before we do cleanup.
Gao feng
2013-Dec-06 02:26 UTC
[Bridge] [patch net/stable v2] br: fix use of ->rx_handler_data in code executed on non-rx_handler path
On 12/06/2013 12:55 AM, Stephen Hemminger wrote:> On Thu, 5 Dec 2013 16:27:37 +0100 > Jiri Pirko <jiri at resnulli.us> 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. >> > > > I think this patch is simpler/better, it restores the old logic. > > Ps. submitting patches to bugzilla is a good way to have them ignored. > >>From Stephen Hemminger <stephen at networkplumber.org> > > Check that incoming STP packet is received on a port assigned to bridge > before processing. It is possible to receive packet on non-bridge port > because they are multicast. > > See: > https://bugzilla.kernel.org/show_bug.cgi?id=64911 > > > Regression introduced by: > commit 716ec052d2280d511e10e90ad54a86f5b5d4dcc2 > Author: Hong Zhiguo <zhiguohong at tencent.com> > Date: Sat Sep 14 22:42:28 2013 +0800 > > bridge: fix NULL pointer deref of br_port_get_rcu > > > Reported-by: Alexander Y. Fomichev > Signed-off-by: Stephen Hemminger <stephen at networkplumber.org> > > > --- a/net/bridge/br_stp_bpdu.c 2013-06-11 09:50:21.522919061 -0700 > +++ b/net/bridge/br_stp_bpdu.c 2013-12-05 08:46:56.090463702 -0800 > @@ -153,6 +153,9 @@ void br_stp_rcv(const struct stp_proto * > if (buf[0] != 0 || buf[1] != 0 || buf[2] != 0) > goto err; > > + if (!br_port_exists(dev)) > + goto err; > + > p = br_port_get_rcu(dev); > if (!p) > goto err;We alreay did some cleanup jobs before mark this dev is not a port of bridge (dev->priv_flags &= ~IFF_BRIDGE_PORT), such as remove the fdb related to this port(br_fdb_delete_by_port). and seems like after these cleanup jobs, before unregister this device, if new skb is received, br_handle_local_finish will call br_fdb_update to create a new fdb whose dst points to the will-be-destroied-port. I don't know if this will cause some problems. seems we should also make sure port is unavailable before we do cleanup.
Vlad Yasevich
2013-Dec-07 01:44 UTC
Re: [patch net/stable v2] br: fix use of ->rx_handler_data in code executed on non-rx_handler path
On 12/05/2013 09:26 PM, Gao feng wrote:> On 12/06/2013 12:55 AM, Stephen Hemminger wrote: >> On Thu, 5 Dec 2013 16:27:37 +0100 >> Jiri Pirko <jiri@resnulli.us> 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. >>> >> >> >> I think this patch is simpler/better, it restores the old logic. >> >> Ps. submitting patches to bugzilla is a good way to have them ignored. >> >> >From Stephen Hemminger <stephen@networkplumber.org> >> >> Check that incoming STP packet is received on a port assigned to bridge >> before processing. It is possible to receive packet on non-bridge port >> because they are multicast. >> >> See: >> https://bugzilla.kernel.org/show_bug.cgi?id=64911 >> >> >> Regression introduced by: >> commit 716ec052d2280d511e10e90ad54a86f5b5d4dcc2 >> Author: Hong Zhiguo <zhiguohong@tencent.com> >> Date: Sat Sep 14 22:42:28 2013 +0800 >> >> bridge: fix NULL pointer deref of br_port_get_rcu >> >> >> Reported-by: Alexander Y. Fomichev >> Signed-off-by: Stephen Hemminger <stephen@networkplumber.org> >> >> >> --- a/net/bridge/br_stp_bpdu.c 2013-06-11 09:50:21.522919061 -0700 >> +++ b/net/bridge/br_stp_bpdu.c 2013-12-05 08:46:56.090463702 -0800 >> @@ -153,6 +153,9 @@ void br_stp_rcv(const struct stp_proto * >> if (buf[0] != 0 || buf[1] != 0 || buf[2] != 0) >> goto err; >> >> + if (!br_port_exists(dev)) >> + goto err; >> + >> p = br_port_get_rcu(dev); >> if (!p) >> goto err; > > > We alreay did some cleanup jobs before mark this dev is not a port of bridge (dev->priv_flags &= ~IFF_BRIDGE_PORT), > such as remove the fdb related to this port(br_fdb_delete_by_port). > > and seems like after these cleanup jobs, before unregister this device, if new skb is received, > br_handle_local_finish will call br_fdb_update to create a new fdb whose dst points to the will-be-destroied-port.Not really. We disable the port first before removing the fdb as a result, br_handle_local_finish() will not update the fdb because the port is in disabled state. -vlad> > I don''t know if this will cause some problems. > seems we should also make sure port is unavailable before we do cleanup. > -- > To unsubscribe from this list: send the line "unsubscribe netdev" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html >