Kuniyuki Iwashima
2023-Jul-11 23:54 UTC
[Bridge] [PATCH v1 net] bridge: Return an error when enabling STP in netns.
When we create an L2 loop on a bridge in netns, we will see packets storm
even if STP is enabled.
# unshare -n
# ip link add br0 type bridge
# ip link add veth0 type veth peer name veth1
# ip link set veth0 master br0 up
# ip link set veth1 master br0 up
# ip link set br0 type bridge stp_state 1
# ip link set br0 up
# sleep 30
# ip -s link show br0
2: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state
UP mode DEFAULT group default qlen 1000
link/ether b6:61:98:1c:1c:b5 brd ff:ff:ff:ff:ff:ff
RX: bytes packets errors dropped missed mcast
956553768 12861249 0 0 0 12861249 <-. Keep
TX: bytes packets errors dropped carrier collsns | increasing
1027834 11951 0 0 0 0 <-' rapidly
This is because llc_rcv() drops all packets in non-root netns and BPDU
is dropped.
Let's show an error when enabling STP in netns.
# unshare -n
# ip link add br0 type bridge
# ip link set br0 type bridge stp_state 1
Error: bridge: STP can't be enabled in non-root netns.
Note this commit will be reverted later when we namespacify the whole LLC
infra.
Fixes: e730c15519d0 ("[NET]: Make packet reception network namespace
safe")
Suggested-by: Harry Coin <hcoin at quietfountain.com>
Link: https://lore.kernel.org/netdev/0f531295-e289-022d-5add-5ceffa0df9bc at
quietfountain.com/
Signed-off-by: Kuniyuki Iwashima <kuniyu at amazon.com>
---
net/bridge/br_stp_if.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/net/bridge/br_stp_if.c b/net/bridge/br_stp_if.c
index 75204d36d7f9..a807996ac56b 100644
--- a/net/bridge/br_stp_if.c
+++ b/net/bridge/br_stp_if.c
@@ -201,6 +201,11 @@ int br_stp_set_enabled(struct net_bridge *br, unsigned long
val,
{
ASSERT_RTNL();
+ if (!net_eq(dev_net(br->dev), &init_net)) {
+ NL_SET_ERR_MSG_MOD(extack, "STP can't be enabled in non-root
netns");
+ return -EINVAL;
+ }
+
if (br_mrp_enabled(br)) {
NL_SET_ERR_MSG_MOD(extack,
"STP can't be enabled if MRP is already enabled");
--
2.30.2
Nikolay Aleksandrov
2023-Jul-12 06:38 UTC
[Bridge] [PATCH v1 net] bridge: Return an error when enabling STP in netns.
On 12/07/2023 02:54, Kuniyuki Iwashima wrote:> When we create an L2 loop on a bridge in netns, we will see packets storm > even if STP is enabled. > > # unshare -n > # ip link add br0 type bridge > # ip link add veth0 type veth peer name veth1 > # ip link set veth0 master br0 up > # ip link set veth1 master br0 up > # ip link set br0 type bridge stp_state 1 > # ip link set br0 up > # sleep 30 > # ip -s link show br0 > 2: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default qlen 1000 > link/ether b6:61:98:1c:1c:b5 brd ff:ff:ff:ff:ff:ff > RX: bytes packets errors dropped missed mcast > 956553768 12861249 0 0 0 12861249 <-. Keep > TX: bytes packets errors dropped carrier collsns | increasing > 1027834 11951 0 0 0 0 <-' rapidly > > This is because llc_rcv() drops all packets in non-root netns and BPDU > is dropped. > > Let's show an error when enabling STP in netns. > > # unshare -n > # ip link add br0 type bridge > # ip link set br0 type bridge stp_state 1 > Error: bridge: STP can't be enabled in non-root netns. > > Note this commit will be reverted later when we namespacify the whole LLC > infra. > > Fixes: e730c15519d0 ("[NET]: Make packet reception network namespace safe") > Suggested-by: Harry Coin <hcoin at quietfountain.com> > Link: https://lore.kernel.org/netdev/0f531295-e289-022d-5add-5ceffa0df9bc at quietfountain.com/ > Signed-off-by: Kuniyuki Iwashima <kuniyu at amazon.com> > --- > net/bridge/br_stp_if.c | 5 +++++ > 1 file changed, 5 insertions(+) > > diff --git a/net/bridge/br_stp_if.c b/net/bridge/br_stp_if.c > index 75204d36d7f9..a807996ac56b 100644 > --- a/net/bridge/br_stp_if.c > +++ b/net/bridge/br_stp_if.c > @@ -201,6 +201,11 @@ int br_stp_set_enabled(struct net_bridge *br, unsigned long val, > { > ASSERT_RTNL(); > > + if (!net_eq(dev_net(br->dev), &init_net)) { > + NL_SET_ERR_MSG_MOD(extack, "STP can't be enabled in non-root netns"); > + return -EINVAL; > + } > + > if (br_mrp_enabled(br)) { > NL_SET_ERR_MSG_MOD(extack, > "STP can't be enabled if MRP is already enabled");Oh well, I guess this should do for now. Hopefully we don't break scripts that rely on configuring the bridge without any errors. Thanks for the patch, Acked-by: Nikolay Aleksandrov <razor at blackwall.org>
Ido Schimmel
2023-Jul-12 14:48 UTC
[Bridge] [PATCH v1 net] bridge: Return an error when enabling STP in netns.
On Tue, Jul 11, 2023 at 04:54:15PM -0700, Kuniyuki Iwashima wrote:> When we create an L2 loop on a bridge in netns, we will see packets storm > even if STP is enabled. > > # unshare -n > # ip link add br0 type bridge > # ip link add veth0 type veth peer name veth1 > # ip link set veth0 master br0 up > # ip link set veth1 master br0 up > # ip link set br0 type bridge stp_state 1 > # ip link set br0 up > # sleep 30 > # ip -s link show br0 > 2: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default qlen 1000 > link/ether b6:61:98:1c:1c:b5 brd ff:ff:ff:ff:ff:ff > RX: bytes packets errors dropped missed mcast > 956553768 12861249 0 0 0 12861249 <-. Keep > TX: bytes packets errors dropped carrier collsns | increasing > 1027834 11951 0 0 0 0 <-' rapidly > > This is because llc_rcv() drops all packets in non-root netns and BPDU > is dropped. > > Let's show an error when enabling STP in netns. > > # unshare -n > # ip link add br0 type bridge > # ip link set br0 type bridge stp_state 1 > Error: bridge: STP can't be enabled in non-root netns. > > Note this commit will be reverted later when we namespacify the whole LLC > infra. > > Fixes: e730c15519d0 ("[NET]: Make packet reception network namespace safe") > Suggested-by: Harry Coin <hcoin at quietfountain.com>I'm not sure that's accurate. I read his response in the link below and he says "I'd rather be warned than blocked" and "But better warned and awaiting a fix than blocked", which I agree with. The patch has the potential to cause a lot of regressions, but without actually fixing the problem. How about simply removing the error [1]? Since iproute2 commit 844c37b42373 ("libnetlink: Handle extack messages for non-error case"), it can print extack warnings and not only errors. With the diff below: # unshare -n # ip link add name br0 type bridge # ip link set dev br0 type bridge stp_state 1 Warning: bridge: STP can't be enabled in non-root netns. # echo $? 0 [1] diff --git a/net/bridge/br_stp_if.c b/net/bridge/br_stp_if.c index a807996ac56b..b5143de37938 100644 --- a/net/bridge/br_stp_if.c +++ b/net/bridge/br_stp_if.c @@ -201,10 +201,8 @@ int br_stp_set_enabled(struct net_bridge *br, unsigned long val, { ASSERT_RTNL(); - if (!net_eq(dev_net(br->dev), &init_net)) { + if (!net_eq(dev_net(br->dev), &init_net)) NL_SET_ERR_MSG_MOD(extack, "STP can't be enabled in non-root netns"); - return -EINVAL; - } if (br_mrp_enabled(br)) { NL_SET_ERR_MSG_MOD(extack,> Link: https://lore.kernel.org/netdev/0f531295-e289-022d-5add-5ceffa0df9bc at quietfountain.com/ > Signed-off-by: Kuniyuki Iwashima <kuniyu at amazon.com> > --- > net/bridge/br_stp_if.c | 5 +++++ > 1 file changed, 5 insertions(+) > > diff --git a/net/bridge/br_stp_if.c b/net/bridge/br_stp_if.c > index 75204d36d7f9..a807996ac56b 100644 > --- a/net/bridge/br_stp_if.c > +++ b/net/bridge/br_stp_if.c > @@ -201,6 +201,11 @@ int br_stp_set_enabled(struct net_bridge *br, unsigned long val, > { > ASSERT_RTNL(); > > + if (!net_eq(dev_net(br->dev), &init_net)) { > + NL_SET_ERR_MSG_MOD(extack, "STP can't be enabled in non-root netns"); > + return -EINVAL; > + } > + > if (br_mrp_enabled(br)) { > NL_SET_ERR_MSG_MOD(extack, > "STP can't be enabled if MRP is already enabled"); > -- > 2.30.2 > >
Nikolay Aleksandrov
2023-Jul-12 14:52 UTC
[Bridge] [PATCH v1 net] bridge: Return an error when enabling STP in netns.
On 12/07/2023 17:48, Ido Schimmel wrote:> On Tue, Jul 11, 2023 at 04:54:15PM -0700, Kuniyuki Iwashima wrote: >> When we create an L2 loop on a bridge in netns, we will see packets storm >> even if STP is enabled. >> >> # unshare -n >> # ip link add br0 type bridge >> # ip link add veth0 type veth peer name veth1 >> # ip link set veth0 master br0 up >> # ip link set veth1 master br0 up >> # ip link set br0 type bridge stp_state 1 >> # ip link set br0 up >> # sleep 30 >> # ip -s link show br0 >> 2: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default qlen 1000 >> link/ether b6:61:98:1c:1c:b5 brd ff:ff:ff:ff:ff:ff >> RX: bytes packets errors dropped missed mcast >> 956553768 12861249 0 0 0 12861249 <-. Keep >> TX: bytes packets errors dropped carrier collsns | increasing >> 1027834 11951 0 0 0 0 <-' rapidly >> >> This is because llc_rcv() drops all packets in non-root netns and BPDU >> is dropped. >> >> Let's show an error when enabling STP in netns. >> >> # unshare -n >> # ip link add br0 type bridge >> # ip link set br0 type bridge stp_state 1 >> Error: bridge: STP can't be enabled in non-root netns. >> >> Note this commit will be reverted later when we namespacify the whole LLC >> infra. >> >> Fixes: e730c15519d0 ("[NET]: Make packet reception network namespace safe") >> Suggested-by: Harry Coin <hcoin at quietfountain.com> > > I'm not sure that's accurate. I read his response in the link below and > he says "I'd rather be warned than blocked" and "But better warned and > awaiting a fix than blocked", which I agree with. The patch has the > potential to cause a lot of regressions, but without actually fixing the > problem. > > How about simply removing the error [1]? Since iproute2 commit > 844c37b42373 ("libnetlink: Handle extack messages for non-error case"), > it can print extack warnings and not only errors. With the diff below: > > # unshare -n > # ip link add name br0 type bridge > # ip link set dev br0 type bridge stp_state 1 > Warning: bridge: STP can't be enabled in non-root netns. > # echo $? > 0 > > [1] > diff --git a/net/bridge/br_stp_if.c b/net/bridge/br_stp_if.c > index a807996ac56b..b5143de37938 100644 > --- a/net/bridge/br_stp_if.c > +++ b/net/bridge/br_stp_if.c > @@ -201,10 +201,8 @@ int br_stp_set_enabled(struct net_bridge *br, unsigned long val, > { > ASSERT_RTNL(); > > - if (!net_eq(dev_net(br->dev), &init_net)) { > + if (!net_eq(dev_net(br->dev), &init_net)) > NL_SET_ERR_MSG_MOD(extack, "STP can't be enabled in non-root netns"); > - return -EINVAL; > - } > > if (br_mrp_enabled(br)) { > NL_SET_ERR_MSG_MOD(extack, >I'd prefer this approach to changing user-visible behaviour and potential regressions. Just change the warning message. Thanks, Nik
Apparently Analagous Threads
- [Bridge] [PATCH v1 net] bridge: Return an error when enabling STP in netns.
- [Bridge] [PATCH v1 net] bridge: Return an error when enabling STP in netns.
- [Bridge] llc needs namespace awareness asap, was Re: Patch fixing STP if bridge in non-default namespace.
- [Bridge] [PATCH] Fix typo in net/bridge/br_stp_if.c
- [RFC PATCH 0/3] vsock: support network namespace