Ilya Maximets <[email protected]> writes: > ovs-vswitchd doesn't use OVS_VPORT_TYPE_GRE/VXLAN/GENEVE with upstream > Linux kernel module since adding support for standard tunnel devices > with COLLECT_METADATA back in 2017. The code to use them is still > present, but it is only activated as a fallback for old kernels, so > not used in practice. And it is marked for removal in the next OVS > release this summer. Modern way to use tunnels with OVS is to create > standard tunnel ports with RTM_NEWLINK + COLLECT_METADATA and add them > as OVS_VPORT_TYPE_NETDEV. > > Device reference management and the netlink options parsing for these > legacy port types is complicated and was a CVE magnet recently. Since > there are no actual users for these port types for a very long time, > let's just remove the support entirely. > > > There are 3 parts to this set: > > 1. The first patch does the tunnel port removal, which is the primary > goal here. > > 2. Patches 2 and 3 remove extra infrastructure that is no longer in > use by anything inside the openvswitch module. It may technically > be used by some out-of-tree module, but it is unlikely, so the > proposal here is to also just remove it. Or we can consider > deprecation. It's not really a user API, it's an API for modules. > Which can be considered as users, I guess. Not sure. > > 3. Patches 4-6 remove functions from gre/vxlan/geneve modules that > were added for openvswitch in the past to support the tunnel types. > And openvswitch is the only in-tree consumer of these functions, > so we could remove them. But they are also exported symbols, so > can in theory be used by some out-of-tree modules, though I doubt > that. Not sure what the process should be here. Removal seems > reasonable, but we may consider deprecation first. > > Thoughts?
I guess the only question is whether we would consider this some kind of userspacce ABI break. BUT, I'm not sure that it is. After all, users could still create some kind of tunnel ports (but they would no long get to do INADDR_ANY type ports) and add them as regular netdev ports. This only changes how they would have to do this. And as you note, there aren't any known users for a long time. I think it's nice to remove them since they create confusion for anyone trying to debug what is going on. > Ilya Maximets (6): > openvswitch: remove support for legacy tunnel types > openvswitch: vport: remove infrastructure for vport options > openvswitch: vport: remove infrastructure for separate modules > net: geneve: remove unused geneve_dev_create_fb > net: gre: remove unused gretap_fb_dev_create > net: vxlan: remove unused vxlan_dev_create > > drivers/net/geneve.c | 48 ----- > drivers/net/vxlan/vxlan_core.c | 42 +---- > include/net/geneve.h | 5 - > include/net/gre.h | 2 - > include/net/vxlan.h | 3 - > include/uapi/linux/openvswitch.h | 31 +++- > net/ipv4/ip_gre.c | 47 ----- > net/openvswitch/Kconfig | 35 ---- > net/openvswitch/Makefile | 4 - > net/openvswitch/datapath.c | 22 +-- > net/openvswitch/vport-geneve.c | 143 --------------- > net/openvswitch/vport-gre.c | 106 ----------- > net/openvswitch/vport-netdev.c | 28 +-- > net/openvswitch/vport-netdev.h | 3 +- > net/openvswitch/vport-vxlan.c | 172 ------------------ > net/openvswitch/vport.c | 76 +------- > net/openvswitch/vport.h | 23 +-- > tools/testing/selftests/net/config | 3 - > .../selftests/net/openvswitch/openvswitch.sh | 37 ---- > .../selftests/net/openvswitch/ovs-dpctl.py | 93 +++------- > 20 files changed, 59 insertions(+), 864 deletions(-) > delete mode 100644 net/openvswitch/vport-geneve.c > delete mode 100644 net/openvswitch/vport-gre.c > delete mode 100644 net/openvswitch/vport-vxlan.c
