)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":7448,"name":"Carl Baldwin","email":"carl@ecbaldwin.net","username":"carl-baldwin"},"change_message_id":"1216cad514ed8e72802c9f42a871e47112bd9024","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Allow MAC address update for direct-physical ports"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Currently, neutron doesn\u0027t allow updating MAC address for direct-physical"},{"line_number":10,"context_line":"ports.  However, after a resize or a migration, a new PCI device is"},{"line_number":11,"context_line":"allocated and attached to the guest.  In this case the MAC address of the"},{"line_number":12,"context_line":"interface in the guest changes will correspond to the new physical"},{"line_number":13,"context_line":"device.  This causes DHCP request on this interface to fail because the"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"7a8ec9b2_138e6d27","line":10,"range":{"start_line":10,"start_character":25,"end_line":10,"end_character":46},"updated":"2016-09-19 18:20:48.000000000","message":"I understand that a migration might trigger this. But, why does a resize need to allocate a new PCI device?\n\nMy initial reaction is that, for resize, we should understand why a resize will do this and potentially treat this as a bug there. For migration, we might want to consider this in the spec we\u0027re looking at for providing port binding information for live migration [1]. Thoughts?\n\n [1] https://review.openstack.org/#/c/309416/","commit_id":"82c8aea8f557972bd9c1b5d4b3ee99a212ef5f6b"},{"author":{"_account_id":7448,"name":"Carl Baldwin","email":"carl@ecbaldwin.net","username":"carl-baldwin"},"change_message_id":"4948f2aad4919d3a3a628c0a3f26b698e864790e","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Allow MAC address update for direct-physical ports"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Currently, neutron doesn\u0027t allow updating MAC address for direct-physical"},{"line_number":10,"context_line":"ports.  However, after a resize or a migration, a new PCI device is"},{"line_number":11,"context_line":"allocated and attached to the guest.  In this case the MAC address of the"},{"line_number":12,"context_line":"interface in the guest changes will correspond to the new physical"},{"line_number":13,"context_line":"device.  This causes DHCP request on this interface to fail because the"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":1,"id":"7a8ec9b2_c42ec441","line":10,"range":{"start_line":10,"start_character":25,"end_line":10,"end_character":46},"in_reply_to":"7a8ec9b2_138e6d27","updated":"2016-09-20 15:04:48.000000000","message":"I now understand that resize is just a special case of migration. So, ignore what I said about resize. However, I still want to reiterate my point about the port binding for migration blueprint that I linked.","commit_id":"82c8aea8f557972bd9c1b5d4b3ee99a212ef5f6b"}],"neutron/plugins/ml2/plugin.py":[{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"9c318535127077a78ba9fb94503dcfb075b1a17b","unresolved":false,"context_lines":[{"line_number":276,"context_line":"        mac_change \u003d (new_mac is not None and"},{"line_number":277,"context_line":"                      orig_port[\u0027mac_address\u0027] !\u003d new_mac)"},{"line_number":278,"context_line":"        if (mac_change and binding.vif_type not in unplugged_types and"},{"line_number":279,"context_line":"                binding.vnic_type not in (portbindings.VNIC_DIRECT_PHYSICAL)):"},{"line_number":280,"context_line":"            raise exc.PortBound(port_id\u003dorig_port[\u0027id\u0027],"},{"line_number":281,"context_line":"                                vif_type\u003dbinding.vif_type,"},{"line_number":282,"context_line":"                                old_mac\u003dorig_port[\u0027mac_address\u0027],"}],"source_content_type":"text/x-python","patch_set":1,"id":"fa7ab95a_926dd39a","line":279,"range":{"start_line":279,"start_character":16,"end_line":279,"end_character":76},"updated":"2016-08-27 07:37:19.000000000","message":"is not this change will allow also for users to update the mac address when the port is bounded? \nmaybe there is a better change we can do.","commit_id":"82c8aea8f557972bd9c1b5d4b3ee99a212ef5f6b"},{"author":{"_account_id":9656,"name":"Ihar Hrachyshka","email":"ihrachys@redhat.com","username":"ihrachys","status":"Red Hat Networking Systems Engineer"},"change_message_id":"bdbd5c68c185577b734cb9d68790e41a29d45bf6","unresolved":false,"context_lines":[{"line_number":276,"context_line":"        mac_change \u003d (new_mac is not None and"},{"line_number":277,"context_line":"                      orig_port[\u0027mac_address\u0027] !\u003d new_mac)"},{"line_number":278,"context_line":"        if (mac_change and binding.vif_type not in unplugged_types and"},{"line_number":279,"context_line":"                binding.vnic_type not in (portbindings.VNIC_DIRECT_PHYSICAL)):"},{"line_number":280,"context_line":"            raise exc.PortBound(port_id\u003dorig_port[\u0027id\u0027],"},{"line_number":281,"context_line":"                                vif_type\u003dbinding.vif_type,"},{"line_number":282,"context_line":"                                old_mac\u003dorig_port[\u0027mac_address\u0027],"}],"source_content_type":"text/x-python","patch_set":1,"id":"7a8ec9b2_78ed93b6","line":279,"range":{"start_line":279,"start_character":16,"end_line":279,"end_character":76},"in_reply_to":"7a8ec9b2_643410de","updated":"2016-09-20 16:02:32.000000000","message":"It seems like the agent does not handle this type of ports at all. In which case we should not get into a situation where the change of MAC address through API layer makes the compute node miss an update.\n\nBut I don\u0027t know whether bind_port for sriov driver needs to know about mac address change. How do we make sure that skipping unbind/bind is safe? In theory, the driver may still want to be informed about such a change.","commit_id":"82c8aea8f557972bd9c1b5d4b3ee99a212ef5f6b"},{"author":{"_account_id":7448,"name":"Carl Baldwin","email":"carl@ecbaldwin.net","username":"carl-baldwin"},"change_message_id":"4948f2aad4919d3a3a628c0a3f26b698e864790e","unresolved":false,"context_lines":[{"line_number":276,"context_line":"        mac_change \u003d (new_mac is not None and"},{"line_number":277,"context_line":"                      orig_port[\u0027mac_address\u0027] !\u003d new_mac)"},{"line_number":278,"context_line":"        if (mac_change and binding.vif_type not in unplugged_types and"},{"line_number":279,"context_line":"                binding.vnic_type not in (portbindings.VNIC_DIRECT_PHYSICAL)):"},{"line_number":280,"context_line":"            raise exc.PortBound(port_id\u003dorig_port[\u0027id\u0027],"},{"line_number":281,"context_line":"                                vif_type\u003dbinding.vif_type,"},{"line_number":282,"context_line":"                                old_mac\u003dorig_port[\u0027mac_address\u0027],"}],"source_content_type":"text/x-python","patch_set":1,"id":"7a8ec9b2_643410de","line":279,"range":{"start_line":279,"start_character":16,"end_line":279,"end_character":76},"in_reply_to":"fa7ab95a_3d8f3a76","updated":"2016-09-20 15:04:48.000000000","message":"The CLI is irrelevant. The mac address can be updated through the API. Anyone that can use a CLI can also hit the API directly. The restriction has been that you cannot update the mac address when the port is bound. I\u0027m not keen on changing this restriction without some thought.\n\nSince this use case for changing the the mac address of a bound port is limited to migration (of which resize is a special case) I think it would be very to restrict it to that use case by including this in the port binding for migration blueprint that we have in discussion for Ocata [1].\n\n [1] https://review.openstack.org/#/c/309416/","commit_id":"82c8aea8f557972bd9c1b5d4b3ee99a212ef5f6b"},{"author":{"_account_id":17973,"name":"Ludovic Beliveau","email":"lbeliveau@gmail.com","username":"lbeliveau"},"change_message_id":"afb414137d083d7c0013df0da0f7af0ccce58449","unresolved":false,"context_lines":[{"line_number":276,"context_line":"        mac_change \u003d (new_mac is not None and"},{"line_number":277,"context_line":"                      orig_port[\u0027mac_address\u0027] !\u003d new_mac)"},{"line_number":278,"context_line":"        if (mac_change and binding.vif_type not in unplugged_types and"},{"line_number":279,"context_line":"                binding.vnic_type not in (portbindings.VNIC_DIRECT_PHYSICAL)):"},{"line_number":280,"context_line":"            raise exc.PortBound(port_id\u003dorig_port[\u0027id\u0027],"},{"line_number":281,"context_line":"                                vif_type\u003dbinding.vif_type,"},{"line_number":282,"context_line":"                                old_mac\u003dorig_port[\u0027mac_address\u0027],"}],"source_content_type":"text/x-python","patch_set":1,"id":"fa7ab95a_3d8f3a76","line":279,"range":{"start_line":279,"start_character":16,"end_line":279,"end_character":76},"in_reply_to":"fa7ab95a_926dd39a","updated":"2016-08-29 16:11:18.000000000","message":"I\u0027m not a neutron expert ... So are you suggesting there exist a CLI command for a user to update a port\u0027s MAC address manually ?\n\nI submitted a new patch set for this issue.  It\u0027s putting more restriction on the condition which would allow a direct-physical port to allow the MAC address to be updated.  But following your question, maybe this is not what we want.\n\nOther options would be:\n- In nova, first unbind the port then update it\u0027s attributes (MAC address).  But to be honest, this complicates things in nova and add extra rpc calls that could otherwise be avoided.\n- Add a new parameter the the update_port RPC call to specify that MAC address is safe to be updated.\n\nWhat do you think ?","commit_id":"82c8aea8f557972bd9c1b5d4b3ee99a212ef5f6b"}]}
