[Windows] Use uplink MAC as source if packet is output to uplink

[wenyingd]

With noEncap mode the Pod packet to remote Pod/Node is output to the
uplink interface directly. This change modifies the source MAC with the
uplink interface's MAC, so that it doesn't require hybrid configurations
on the host interface.

Fixes #3505

Signed-off-by: wenyingd wenyingd@vmware.com

[wenyingd]

/test-windows-all
/test-all
/skip-ipv6-all
/skip-ipv6-only-all

[XinShuYang]

/test-windows-e2e

[codecov-commenter]

Codecov Report

Merging #3516 (483c3f3) into main (2c7d486) will decrease coverage by 9.78%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##             main    #3516      +/-   ##
==========================================
- Coverage   65.58%   55.79%   -9.79%     
==========================================
  Files         277      391     +114     
  Lines       27264    52463   +25199     
==========================================
+ Hits        17881    29272   +11391     
- Misses       7483    20756   +13273     
- Partials     1900     2435     +535     

Flag	Coverage Δ
e2e-tests	54.23% <0.00%> (?)
integration-tests	38.18% <ø> (?)
kind-e2e-tests	55.74% <0.00%> (-0.20%)	⬇️
unit-tests	43.08% <0.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
pkg/agent/openflow/pipeline.go	75.30% <0.00%> (-1.64%)	⬇️
pkg/agent/cniserver/pod_configuration_linux.go	26.31% <0.00%> (-40.36%)	⬇️
pkg/controller/ipam/antrea_ipam_controller.go	48.71% <0.00%> (-31.57%)	⬇️
pkg/controller/networkpolicy/endpoint_querier.go	61.46% <0.00%> (-29.97%)	⬇️
pkg/controller/egress/controller.go	62.19% <0.00%> (-26.26%)	⬇️
pkg/controller/ipam/validate.go	57.95% <0.00%> (-24.31%)	⬇️
pkg/agent/cniserver/ipam/antrea_ipam_controller.go	56.66% <0.00%> (-23.34%)	⬇️
pkg/agent/cniserver/ipam/antrea_ipam.go	57.24% <0.00%> (-22.55%)	⬇️
pkg/agent/util/iptables/lock.go	60.00% <0.00%> (-21.82%)	⬇️
pkg/controller/egress/store/egressgroup.go	37.93% <0.00%> (-21.80%)	⬇️
... and 354 more

[hongliangl]

LGTM

[XinShuYang]

/test-windows-all

[wenyingd]

/test-windows-conformance
/test-e2e

[XinShuYang]

LGTM, CI tests passed on windows testbed without promiscuous mode support. I will update all windows testbeds setup once this pr is merged. @tnqn @wenyingd

[tnqn]

LGTM, CI tests passed on windows testbed without promiscuous mode support. I will update all windows testbeds setup once this pr is merged. @tnqn @wenyingd

@XinShuYang not only promiscuous mode, could you check whether it can work without enabling "MAC Address Changes" and "Forged transmits"? I think this patch is irrelavant to promiscuous mode, which is not required at all from the begining.

[tnqn]

@wenyingd without this patch, vsphere and cloud platforms need to disable MAC check to work around, could you backport it to previous releases starting from 1.2 as well?

[XinShuYang]

LGTM, CI tests passed on windows testbed without promiscuous mode support. I will update all windows testbeds setup once this pr is merged. @tnqn @wenyingd

@XinShuYang not only promiscuous mode, could you check whether it can work without enabling "MAC Address Changes" and "Forged transmits"? I think this patch is irrelavant to promiscuous mode, which is not required at all from the begining.

@tnqn I understand we need to disable "Forged transmits" to drop packets with noneffective MAC address, could you explain why it is necessary to disable "MAC Address Changes" for this PR?

[tnqn]

@XinShuYang Not only for this PR, I think in theory we don't require any of these options to be configured but all of them were required to be set to specific values due to either misunderstanding or implementation bug. My point is to ensure everything works as expected, then we can remove these restrictions from document (if there is any) to clarify.

[XinShuYang]

@tnqn Thanks for the explanation, I will retest it with new setup.

[XinShuYang]

/test-windows-all

[XinShuYang]

/test-windows-e2e