Messages in this thread | | | From | Jarod Wilson <> | Date | Thu, 19 Mar 2020 15:29:51 -0400 | Subject | Re: [PATCH net] ipv6: don't auto-add link-local address to lag ports |
| |
On Thu, Mar 19, 2020 at 1:06 PM Eric Dumazet <eric.dumazet@gmail.com> wrote: > > On 3/19/20 9:42 AM, Jarod Wilson wrote: > > > Interesting. We'll keep digging over here, but that's definitely not > > working for this particular use case with OVS for whatever reason. > > I did a quick test and confirmed that my bonding slaves do not have link-local addresses, > without anything done to prevent them to appear. > > You might add a selftest, if you ever find what is the trigger :)
Okay, have a basic reproducer, courtesy of Marcelo:
# ip link add name bond0 type bond # ip link set dev ens2f0np0 master bond0 # ip link set dev ens2f1np2 master bond0 # ip link set dev bond0 up # ip a s 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000 link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 inet 127.0.0.1/8 scope host lo valid_lft forever preferred_lft forever inet6 ::1/128 scope host valid_lft forever preferred_lft forever 2: ens2f0np0: <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP> mtu 1500 qdisc mq master bond0 state UP group default qlen 1000 link/ether 00:0f:53:2f:ea:40 brd ff:ff:ff:ff:ff:ff 5: ens2f1np2: <NO-CARRIER,BROADCAST,MULTICAST,SLAVE,UP> mtu 1500 qdisc mq master bond0 state DOWN group default qlen 1000 link/ether 00:0f:53:2f:ea:40 brd ff:ff:ff:ff:ff:ff 11: bond0: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000 link/ether 00:0f:53:2f:ea:40 brd ff:ff:ff:ff:ff:ff inet6 fe80::20f:53ff:fe2f:ea40/64 scope link valid_lft forever preferred_lft forever
(above trimmed to relevant entries, obviously)
# sysctl net.ipv6.conf.ens2f0np0.addr_gen_mode=0 net.ipv6.conf.ens2f0np0.addr_gen_mode = 0 # sysctl net.ipv6.conf.ens2f1np2.addr_gen_mode=0 net.ipv6.conf.ens2f1np2.addr_gen_mode = 0
# ip a l ens2f0np0 2: ens2f0np0: <BROADCAST,MULTICAST,SLAVE,UP,LOWER_UP> mtu 1500 qdisc mq master bond0 state UP group default qlen 1000 link/ether 00:0f:53:2f:ea:40 brd ff:ff:ff:ff:ff:ff inet6 fe80::20f:53ff:fe2f:ea40/64 scope link tentative valid_lft forever preferred_lft forever # ip a l ens2f1np2 5: ens2f1np2: <NO-CARRIER,BROADCAST,MULTICAST,SLAVE,UP> mtu 1500 qdisc mq master bond0 state DOWN group default qlen 1000 link/ether 00:0f:53:2f:ea:40 brd ff:ff:ff:ff:ff:ff inet6 fe80::20f:53ff:fe2f:ea40/64 scope link tentative valid_lft forever preferred_lft forever
Looks like addrconf_sysctl_addr_gen_mode() bypasses the original "is this a slave interface?" check, and results in an address getting added, while w/the proposed patch added, no address gets added.
Looking back through git history again, I see a bunch of 'Fixes: d35a00b8e33d ("net/ipv6: allow sysctl to change link-local address generation mode")' patches, and I guess that's where this issue was also introduced.
-- Jarod Wilson jarod@redhat.com
| |