)]}'
{"specs/stein/approved/generic-os-vif-offloads.rst":[{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":11,"context_line":"https://blueprints.launchpad.net/nova/+spec/generic-os-vif-offloads"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"The existing method in os-vif is to pass offload metadata is via a"},{"line_number":14,"context_line":"``VIFPortProfileOVSRepresentor`` port profile object. This is currently used by"},{"line_number":15,"context_line":"the ``ovs`` reference plugin and the external ``agilio_ovs`` plugin. This spec"},{"line_number":16,"context_line":"proposes a refactor of the interface to support more VIF types and offload"},{"line_number":17,"context_line":"modes."}],"source_content_type":"text/x-rst","patch_set":1,"id":"3f79a3b5_e7edd324","line":14,"range":{"start_line":14,"start_character":2,"end_line":14,"end_character":30},"updated":"2018-11-21 14:49:27.000000000","message":"so just a note on terminology VIFPortProfileOVSRepresentor does not contain any metadata today about offloads.\nit contains metadata in the form of a pci addres and a \na netdev representor port names.\n\nthis does not convay any info about offloads it just carry\nthe metadata required to wire up the dataplane. this is similar to the vhost-user socket path which is metadata for how to wire up the contol plane.\n\nso currently VIFPortProfileOVSRepresentor does not directly support what i would consider to be offload metadata such as\nenabling TSO or Other hardware offloads.","commit_id":"811845a0ac2c245118ab88f69e0fb80c7ca40728"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":10,"context_line":""},{"line_number":11,"context_line":"https://blueprints.launchpad.net/nova/+spec/generic-os-vif-offloads"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"The existing method in os-vif is to pass offload metadata is via a"},{"line_number":14,"context_line":"``VIFPortProfileOVSRepresentor`` port profile object. This is currently used by"},{"line_number":15,"context_line":"the ``ovs`` reference plugin and the external ``agilio_ovs`` plugin. This spec"},{"line_number":16,"context_line":"proposes a refactor of the interface to support more VIF types and offload"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_f53b1d33","line":13,"range":{"start_line":13,"start_character":58,"end_line":13,"end_character":60},"updated":"2018-11-19 17:22:12.000000000","message":"s/is//","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":43,"context_line":"the hypervisor is generally able to also perform the datapath control, but this"},{"line_number":44,"context_line":"is not the case for every VIF type (hence the existence of os-vif)."},{"line_number":45,"context_line":""},{"line_number":46,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":47,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":48,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":49,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":50,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_75e5cda2","line":47,"range":{"start_line":46,"start_character":38,"end_line":47,"end_character":49},"updated":"2018-11-19 17:22:12.000000000","message":"this is probably the source of most of the confusion.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Currently, os-vif and Nova implement `switchdev SR-IOV offloads`_ for Open"},{"line_number":60,"context_line":"vSwitch. In this model, a representor netdev on the host is associated with"},{"line_number":61,"context_line":"each PCI Virtual Function. This representor functions like a handle for the"},{"line_number":62,"context_line":"datapath."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_0e1a4a18","line":61,"range":{"start_line":61,"start_character":5,"end_line":61,"end_character":25},"updated":"2018-11-19 17:22:12.000000000","message":"PCI VF? or SR-IOV VF?","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":58,"context_line":""},{"line_number":59,"context_line":"Currently, os-vif and Nova implement `switchdev SR-IOV offloads`_ for Open"},{"line_number":60,"context_line":"vSwitch. In this model, a representor netdev on the host is associated with"},{"line_number":61,"context_line":"each PCI Virtual Function. This representor functions like a handle for the"},{"line_number":62,"context_line":"datapath."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_e0334b11","line":61,"range":{"start_line":61,"start_character":5,"end_line":61,"end_character":25},"in_reply_to":"3f79a3b5_0e1a4a18","updated":"2018-11-20 12:34:20.000000000","message":"Is there a difference? But I take your point. s/PCI//\n\nSR-IOV is the PCI spec portion that describes how Virtual Functions operate. Virtual Functions are PCI devices and only operate in the context of the SR-IOV portion of the spec.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":59,"context_line":"Currently, os-vif and Nova implement `switchdev SR-IOV offloads`_ for Open"},{"line_number":60,"context_line":"vSwitch. In this model, a representor netdev on the host is associated with"},{"line_number":61,"context_line":"each PCI Virtual Function. This representor functions like a handle for the"},{"line_number":62,"context_line":"datapath."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"},{"line_number":65,"context_line":"plugin. Optionally, a netdev name can also be passed to allow for friendly"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_2e15c6e7","line":62,"range":{"start_line":62,"start_character":0,"end_line":62,"end_character":8},"updated":"2018-11-19 17:22:12.000000000","message":"datapath or e-switch vport? :P\n\n(see, /me learning how to communicate as a network engineer...)","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":59,"context_line":"Currently, os-vif and Nova implement `switchdev SR-IOV offloads`_ for Open"},{"line_number":60,"context_line":"vSwitch. In this model, a representor netdev on the host is associated with"},{"line_number":61,"context_line":"each PCI Virtual Function. This representor functions like a handle for the"},{"line_number":62,"context_line":"datapath."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"},{"line_number":65,"context_line":"plugin. Optionally, a netdev name can also be passed to allow for friendly"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_49a35cb1","line":62,"range":{"start_line":62,"start_character":0,"end_line":62,"end_character":8},"in_reply_to":"3f79a3b5_2e15c6e7","updated":"2018-11-21 14:49:27.000000000","message":"the representor netdevs are","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":59,"context_line":"Currently, os-vif and Nova implement `switchdev SR-IOV offloads`_ for Open"},{"line_number":60,"context_line":"vSwitch. In this model, a representor netdev on the host is associated with"},{"line_number":61,"context_line":"each PCI Virtual Function. This representor functions like a handle for the"},{"line_number":62,"context_line":"datapath."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"},{"line_number":65,"context_line":"plugin. Optionally, a netdev name can also be passed to allow for friendly"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_fb1a1c69","line":62,"range":{"start_line":62,"start_character":0,"end_line":62,"end_character":8},"in_reply_to":"3f79a3b5_2e15c6e7","updated":"2018-11-20 12:34:20.000000000","message":"yeah, the terminology is unclear here. Technically this translates to \"an ingress/egress endpoint on the packet processing pipeline on the NIC\". And when I meant \"for\" I actually meant \"for a virtual port on the datapath\"... Familiarity brings jargon.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":68,"context_line":"The ``ovs`` and ``agilio_ovs`` os-vif plugins then look up the associated"},{"line_number":69,"context_line":"representor for the VF and perform the datapath plugging. From Nova\u0027s"},{"line_number":70,"context_line":"perspective the hypervisor then either passes through a VF using the data from"},{"line_number":71,"context_line":"the ``VIFHostDevice`` os-vif object (in ``direct`` plugging), or plugs the"},{"line_number":72,"context_line":"Instance into a vhost-user handle with data from a ``VIFVHostUser`` os-vif"},{"line_number":73,"context_line":"object (in ``virtio-forwarder`` plugging)."},{"line_number":74,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_eee34e02","line":71,"range":{"start_line":71,"start_character":51,"end_line":71,"end_character":59},"updated":"2018-11-19 17:22:12.000000000","message":"s/plugging/vNIC types/\n\nMaybe? I know you kinda equated the two things above when you said that recently the vNIC type has been morphed into representing a plugging behaviour, but still...","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":68,"context_line":"The ``ovs`` and ``agilio_ovs`` os-vif plugins then look up the associated"},{"line_number":69,"context_line":"representor for the VF and perform the datapath plugging. From Nova\u0027s"},{"line_number":70,"context_line":"perspective the hypervisor then either passes through a VF using the data from"},{"line_number":71,"context_line":"the ``VIFHostDevice`` os-vif object (in ``direct`` plugging), or plugs the"},{"line_number":72,"context_line":"Instance into a vhost-user handle with data from a ``VIFVHostUser`` os-vif"},{"line_number":73,"context_line":"object (in ``virtio-forwarder`` plugging)."},{"line_number":74,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_3b941498","line":71,"range":{"start_line":71,"start_character":51,"end_line":71,"end_character":59},"in_reply_to":"3f79a3b5_eee34e02","updated":"2018-11-20 12:34:20.000000000","message":"Yeah, that makes more sense, thanks!","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":88,"context_line":"Proposed change"},{"line_number":89,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":90,"context_line":""},{"line_number":91,"context_line":"Option 1: Use composition in stead of inheritance"},{"line_number":92,"context_line":"-------------------------------------------------"},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"In stead of using an inheritance based pattern to model the offload"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_4e7cc221","line":91,"range":{"start_line":91,"start_character":26,"end_line":91,"end_character":34},"updated":"2018-11-19 17:22:12.000000000","message":"instead","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":91,"context_line":"Option 1: Use composition in stead of inheritance"},{"line_number":92,"context_line":"-------------------------------------------------"},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"In stead of using an inheritance based pattern to model the offload"},{"line_number":95,"context_line":"capabilities and metadata, use a composition pattern:"},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"    * Implement a ``DatapathOffloadBase`` class."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_6e7f3e16","line":94,"range":{"start_line":94,"start_character":0,"end_line":94,"end_character":8},"updated":"2018-11-19 17:22:12.000000000","message":"Instead","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":94,"context_line":"In stead of using an inheritance based pattern to model the offload"},{"line_number":95,"context_line":"capabilities and metadata, use a composition pattern:"},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"    * Implement a ``DatapathOffloadBase`` class."},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"    * Subclass this to ``DatapathOffloadRepresentor`` with the following"},{"line_number":100,"context_line":"      members:"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_499d9cb8","line":97,"range":{"start_line":97,"start_character":20,"end_line":97,"end_character":39},"updated":"2018-11-19 17:22:12.000000000","message":"I\u0027d just call this OffloadBase. Is there something non-datapath that can/should be offloaded?","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":94,"context_line":"In stead of using an inheritance based pattern to model the offload"},{"line_number":95,"context_line":"capabilities and metadata, use a composition pattern:"},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"    * Implement a ``DatapathOffloadBase`` class."},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"    * Subclass this to ``DatapathOffloadRepresentor`` with the following"},{"line_number":100,"context_line":"      members:"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_1b8038f6","line":97,"range":{"start_line":97,"start_character":20,"end_line":97,"end_character":39},"in_reply_to":"3f79a3b5_499d9cb8","updated":"2018-11-20 12:34:20.000000000","message":"Yes, unfortunately, there\u0027s confusion here too. Check out the stuff from RHEL6:\n\nhttps://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/6/html/performance_tuning_guide/network-nic-offloads\n\nIn short - one type of offload you can query the netdev (Hey, do you support generating TCP frames for me in some weird way???)...\n\nThe particular type of offloads addressed in this spec (TC/FLOWER) advertise themselves on _other_ netdevs than the ones they offload traffic _from_.\n\nI think adding an explanatory passage that describes this confusion is probably apt.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":96,"context_line":""},{"line_number":97,"context_line":"    * Implement a ``DatapathOffloadBase`` class."},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"    * Subclass this to ``DatapathOffloadRepresentor`` with the following"},{"line_number":100,"context_line":"      members:"},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"        * ``representor_name: StringField()``"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_8993b4b0","line":99,"range":{"start_line":99,"start_character":25,"end_line":99,"end_character":51},"updated":"2018-11-19 17:22:12.000000000","message":"I\u0027d call this VFRepresentorOffload","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":96,"context_line":""},{"line_number":97,"context_line":"    * Implement a ``DatapathOffloadBase`` class."},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"    * Subclass this to ``DatapathOffloadRepresentor`` with the following"},{"line_number":100,"context_line":"      members:"},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"        * ``representor_name: StringField()``"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_7b704c42","line":99,"range":{"start_line":99,"start_character":25,"end_line":99,"end_character":51},"in_reply_to":"3f79a3b5_8993b4b0","updated":"2018-11-20 12:34:20.000000000","message":"I\u0027m happy with that naming too.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":100,"context_line":"      members:"},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"        * ``representor_name: StringField()``"},{"line_number":103,"context_line":"        * ``representor_address: PCIAddressField()``"},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"    * Implement ``VIFPortProfileDatapathOffload`` by subclassing"},{"line_number":106,"context_line":"      ``VIFPortProfileBase`` and adding a ``datapath_offload`` member:"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_8e23ba12","line":103,"range":{"start_line":103,"start_character":33,"end_line":103,"end_character":52},"updated":"2018-11-19 17:22:12.000000000","message":"I\u0027d just leave this as a StringField, since we\u0027re not entirely positive that a PCI address will always be the only address type allowed for a VF representor. A safe bet in anything network-related is that everything will change one year from now, before any stable software or hardware is ever released.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":100,"context_line":"      members:"},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"        * ``representor_name: StringField()``"},{"line_number":103,"context_line":"        * ``representor_address: PCIAddressField()``"},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"    * Implement ``VIFPortProfileDatapathOffload`` by subclassing"},{"line_number":106,"context_line":"      ``VIFPortProfileBase`` and adding a ``datapath_offload`` member:"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_fbdf7c08","line":103,"range":{"start_line":103,"start_character":33,"end_line":103,"end_character":52},"in_reply_to":"3f79a3b5_8e23ba12","updated":"2018-11-20 12:34:20.000000000","message":"A generic bag of strings? Whoo boy!\n\nI\u0027m broadly in agreement here from the protocol point of view: strict typing can be pushed out to the plugin. At this point, however the \"Representor\" model strictly understands only one method of operation: a PCI device that\u0027s used to look up a netdev.\n\nThis is in contrast to the \"mdev\" model, which accepts a path.\n\nI\u0027d go as far as to say: calling the object \"Representor\" implies that this should be a PCI address. Calling the object \"GenericOffload\" means that it can work with a string.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":102,"context_line":"        * ``representor_name: StringField()``"},{"line_number":103,"context_line":"        * ``representor_address: PCIAddressField()``"},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"    * Implement ``VIFPortProfileDatapathOffload`` by subclassing"},{"line_number":106,"context_line":"      ``VIFPortProfileBase`` and adding a ``datapath_offload`` member:"},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"        * ``datapath_offload: ObjectField(\u0027DatapathOffloadBase\u0027, nullable\u003dTrue,"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_4976fc94","line":105,"range":{"start_line":105,"start_character":18,"end_line":105,"end_character":47},"updated":"2018-11-19 17:22:12.000000000","message":"Maybe just change the VIFPortProfile base class to have:\n\n offload: ObjectField(\u0027OffloadBase`, nullable\u003dTrue)\n\nand leave it at that. A nullable object field is extensible and perfectly reasonable to have in the base port profile class IMHO.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":102,"context_line":"        * ``representor_name: StringField()``"},{"line_number":103,"context_line":"        * ``representor_address: PCIAddressField()``"},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"    * Implement ``VIFPortProfileDatapathOffload`` by subclassing"},{"line_number":106,"context_line":"      ``VIFPortProfileBase`` and adding a ``datapath_offload`` member:"},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"        * ``datapath_offload: ObjectField(\u0027DatapathOffloadBase\u0027, nullable\u003dTrue,"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_fb2d3cfc","line":105,"range":{"start_line":105,"start_character":18,"end_line":105,"end_character":47},"in_reply_to":"3f79a3b5_4976fc94","updated":"2018-11-20 12:34:20.000000000","message":"If you\u0027re willing to live with the version bumping, I\u0027m happy to revert to that model:\n\nhttps://review.openstack.org/#/c/572081/5/os_vif/objects/vif.py@182","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":108,"context_line":"        * ``datapath_offload: ObjectField(\u0027DatapathOffloadBase\u0027, nullable\u003dTrue,"},{"line_number":109,"context_line":"          default\u003dNone)``"},{"line_number":110,"context_line":""},{"line_number":111,"context_line":"    * Add a ``datapath_offload`` member to ``VIFPortProfileOpenVSwitch``."},{"line_number":112,"context_line":""},{"line_number":113,"context_line":"    * Update the object versions to reflect the changes made."},{"line_number":114,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_c989ac70","line":111,"range":{"start_line":111,"start_character":4,"end_line":111,"end_character":73},"updated":"2018-11-19 17:22:12.000000000","message":"If you added an \"offload: Objectfield(\u0027OffloadBase\u0027, nullable\u003dTrue)\" field to the VIFPortProfile base class, you wouldn\u0027t need this change.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":108,"context_line":"        * ``datapath_offload: ObjectField(\u0027DatapathOffloadBase\u0027, nullable\u003dTrue,"},{"line_number":109,"context_line":"          default\u003dNone)``"},{"line_number":110,"context_line":""},{"line_number":111,"context_line":"    * Add a ``datapath_offload`` member to ``VIFPortProfileOpenVSwitch``."},{"line_number":112,"context_line":""},{"line_number":113,"context_line":"    * Update the object versions to reflect the changes made."},{"line_number":114,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_7b412ca7","line":111,"range":{"start_line":111,"start_character":4,"end_line":111,"end_character":73},"in_reply_to":"3f79a3b5_c989ac70","updated":"2018-11-20 12:34:20.000000000","message":"yep, but all port profiles would be version bumped.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":110,"context_line":""},{"line_number":111,"context_line":"    * Add a ``datapath_offload`` member to ``VIFPortProfileOpenVSwitch``."},{"line_number":112,"context_line":""},{"line_number":113,"context_line":"    * Update the object versions to reflect the changes made."},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"    * Implement a ``DatapathOffloadInfo`` class."},{"line_number":116,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_ee858eac","line":113,"range":{"start_line":113,"start_character":6,"end_line":113,"end_character":61},"updated":"2018-11-19 17:22:12.000000000","message":"As I\u0027ve mentioned before, we\u0027re not yet sending any ovo objects over the wire anywhere. I\u0027d prefer to just stop incrementing versions entirely in the ovo objects for os-vif until such time as something actually sends one of the objects over the wire, at which point we\u0027ll set all the versions to 1.0 and actually start using the versioning functionality.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":110,"context_line":""},{"line_number":111,"context_line":"    * Add a ``datapath_offload`` member to ``VIFPortProfileOpenVSwitch``."},{"line_number":112,"context_line":""},{"line_number":113,"context_line":"    * Update the object versions to reflect the changes made."},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"    * Implement a ``DatapathOffloadInfo`` class."},{"line_number":116,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_bb2024fb","line":113,"range":{"start_line":113,"start_character":6,"end_line":113,"end_character":61},"in_reply_to":"3f79a3b5_ee858eac","updated":"2018-11-20 12:34:20.000000000","message":"(Ah, helps to read further into the comments before responding to comments!) Leaving the above replies for comedic effect.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":112,"context_line":""},{"line_number":113,"context_line":"    * Update the object versions to reflect the changes made."},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"    * Implement a ``DatapathOffloadInfo`` class."},{"line_number":116,"context_line":""},{"line_number":117,"context_line":"    * Add a ``supported_datapath_offloads`` member to ``HostPortProfileInfo``:"},{"line_number":118,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_2ef7065e","line":115,"range":{"start_line":115,"start_character":6,"end_line":115,"end_character":48},"updated":"2018-11-19 17:22:12.000000000","message":"Hmmm, can you expand on what this would look like? Would it basically be a vNIC type and supported offload classes?","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":112,"context_line":""},{"line_number":113,"context_line":"    * Update the object versions to reflect the changes made."},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"    * Implement a ``DatapathOffloadInfo`` class."},{"line_number":116,"context_line":""},{"line_number":117,"context_line":"    * Add a ``supported_datapath_offloads`` member to ``HostPortProfileInfo``:"},{"line_number":118,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_fbc9bc23","line":115,"range":{"start_line":115,"start_character":6,"end_line":115,"end_character":48},"in_reply_to":"3f79a3b5_2ef7065e","updated":"2018-11-20 12:34:20.000000000","message":"Ah, this mirrors the HostPortProfileInfo in:\n\nhttps://github.com/openstack/os-vif/blob/1e4cf6160efc7fb4fd8279bd2a666e4d72f4fb6c/os_vif/objects/host_info.py#L46\n\nBasically \"stubby\" code that\u0027s intended as a skeleton for future over-the-wire negotiation. If this isn\u0027t added (even if it\u0027s unused), the tech debt would creep up.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":114,"context_line":""},{"line_number":115,"context_line":"    * Implement a ``DatapathOffloadInfo`` class."},{"line_number":116,"context_line":""},{"line_number":117,"context_line":"    * Add a ``supported_datapath_offloads`` member to ``HostPortProfileInfo``:"},{"line_number":118,"context_line":""},{"line_number":119,"context_line":"      * ``supported_datapath_offloads:"},{"line_number":120,"context_line":"        ListOfObjectsField(\u0027DatapathOffloadInfo\u0027)``"},{"line_number":121,"context_line":""},{"line_number":122,"context_line":"    * Update the os-vif OVS reference plugin to accept and use the new versions"},{"line_number":123,"context_line":"      and fields."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_4eb0428e","line":120,"range":{"start_line":117,"start_character":0,"end_line":120,"end_character":51},"updated":"2018-11-19 17:22:12.000000000","message":"will this need to indicate a set of supported datapath offloads *for specific vnic types*? i.e. is this going to need to be a hashmap and not a list of DatapathOffloadInfo objects?","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":114,"context_line":""},{"line_number":115,"context_line":"    * Implement a ``DatapathOffloadInfo`` class."},{"line_number":116,"context_line":""},{"line_number":117,"context_line":"    * Add a ``supported_datapath_offloads`` member to ``HostPortProfileInfo``:"},{"line_number":118,"context_line":""},{"line_number":119,"context_line":"      * ``supported_datapath_offloads:"},{"line_number":120,"context_line":"        ListOfObjectsField(\u0027DatapathOffloadInfo\u0027)``"},{"line_number":121,"context_line":""},{"line_number":122,"context_line":"    * Update the os-vif OVS reference plugin to accept and use the new versions"},{"line_number":123,"context_line":"      and fields."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_dba5e04a","line":120,"range":{"start_line":117,"start_character":0,"end_line":120,"end_character":51},"in_reply_to":"3f79a3b5_4eb0428e","updated":"2018-11-20 12:34:20.000000000","message":"At this point, the idea is that the os-vif plugin states with objects + version ranges:\n\n\"I understand VIF type A, B, C\"\n\"I understand port profile D, E, F\"\n\"I understand dp-offload G, H, I\"\n\nI think it\u0027s fair to ask the Neutron mechanism driver to pre-filter the valid combinations at port creation, as far as possible. Note, currently Neutron only knows VNIC/VIF combinations which Nova translates to os-vif VIF+PP+(dp-offload (future)). While it might be possible to pre-filter this at plugin registration time, it might be much simpler just to handle unexpected combinations sent from Nova as a late failure with the port going into ERROR state. It\u0027s _not_ currently expected to explode inside Nova, since VIF+VNIC gets translated 1:1 into VIF+PP+DPO. Future work with over-the-wire would shift that responsibility to Neutron. Not sure if it\u0027s good or bad...","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":126,"context_line":"offload) with a new VIF type will not require modifications to os-vif. Future"},{"line_number":127,"context_line":"offload methods will require subclassing ``DatapathOffloadBase``."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"In stead of implementing potentially brittle backlevelling code, this option"},{"line_number":130,"context_line":"proposes to keep two parallel interfaces alive in Nova for at least one"},{"line_number":131,"context_line":"overlapping release cycle, before the Open vSwitch plugin is updated in os-vif."},{"line_number":132,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_cea3f243","line":129,"range":{"start_line":129,"start_character":0,"end_line":129,"end_character":8},"updated":"2018-11-19 17:22:12.000000000","message":"Instead","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":127,"context_line":"offload methods will require subclassing ``DatapathOffloadBase``."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"In stead of implementing potentially brittle backlevelling code, this option"},{"line_number":130,"context_line":"proposes to keep two parallel interfaces alive in Nova for at least one"},{"line_number":131,"context_line":"overlapping release cycle, before the Open vSwitch plugin is updated in os-vif."},{"line_number":132,"context_line":""},{"line_number":133,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":134,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_8e5bba5f","line":131,"range":{"start_line":130,"start_character":0,"end_line":131,"end_character":79},"updated":"2018-11-19 17:22:12.000000000","message":"or just forget about the versioning since nothing sends os-vif objects over the wire yet.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":174,"context_line":"    * VNIC type: ``virtio-forwarder``"},{"line_number":175,"context_line":"    * os-vif object: ``VIFVHostUser``"},{"line_number":176,"context_line":"    * ``port_profile: VIFPortProfileOpenVSwitch``"},{"line_number":177,"context_line":"    * ``port_profile.datapath_offload: DatapathOffloadRepresentor``"},{"line_number":178,"context_line":""},{"line_number":179,"context_line":"Option 2: Use renaming to create a generic representor profile"},{"line_number":180,"context_line":"--------------------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_8920949f","line":177,"updated":"2018-11-19 17:22:12.000000000","message":"I really like the above before and after explanations. Makes it super clear how the changes will affect things. ++","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"631644e123b6387e19abb081e49385a3e9a8db38","unresolved":false,"context_lines":[{"line_number":176,"context_line":"    * ``port_profile: VIFPortProfileOpenVSwitch``"},{"line_number":177,"context_line":"    * ``port_profile.datapath_offload: DatapathOffloadRepresentor``"},{"line_number":178,"context_line":""},{"line_number":179,"context_line":"Option 2: Use renaming to create a generic representor profile"},{"line_number":180,"context_line":"--------------------------------------------------------------"},{"line_number":181,"context_line":""},{"line_number":182,"context_line":"Keep using an inheritance based pattern to model the offload capabilities and"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_41fba146","line":179,"range":{"start_line":179,"start_character":14,"end_line":179,"end_character":22},"updated":"2018-10-29 15:18:02.000000000","message":"inheritance","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"371a0fa480663e4162783b9d36452ba3560dc157","unresolved":false,"context_lines":[{"line_number":203,"context_line":""},{"line_number":204,"context_line":"    * VNIC type: ``virtio-forwarder``"},{"line_number":205,"context_line":"    * os-vif object: ``VIFVHostUser``"},{"line_number":206,"context_line":"    * ``port_profile: VIFPortProfileRepresentor``"},{"line_number":207,"context_line":""},{"line_number":208,"context_line":"Additional Impact for Both Options"},{"line_number":209,"context_line":"----------------------------------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_a971d0ac","line":206,"updated":"2018-11-19 17:22:12.000000000","message":"Clearly, I\u0027m partial to Option 1. I think containing all the variants for datapath offloads in a single datapath_offload field of the VIFPortProfileOpenVSwitch object is more appropriate and gives us the flexibility to handle datapath offloads for non-OVS switch infrastructure and handle additional vendors as needed in the future without needing to change the VIFPortProfile classes at all.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4c1932c1470650285c300a2dbce3a5e426c8b454","unresolved":false,"context_lines":[{"line_number":203,"context_line":""},{"line_number":204,"context_line":"    * VNIC type: ``virtio-forwarder``"},{"line_number":205,"context_line":"    * os-vif object: ``VIFVHostUser``"},{"line_number":206,"context_line":"    * ``port_profile: VIFPortProfileRepresentor``"},{"line_number":207,"context_line":""},{"line_number":208,"context_line":"Additional Impact for Both Options"},{"line_number":209,"context_line":"----------------------------------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_1b427868","line":206,"in_reply_to":"3f79a3b5_a971d0ac","updated":"2018-11-20 12:34:20.000000000","message":"It is a more general fix, also more invasive. If versioning can be suspended and upgrades can be smoothed out, it might be technically neater.","commit_id":"4645b679959e0ebc9431e20b202602572e73d908"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"db2ac2cddd445c8bcb998fa4ebd0bb4c171a4c38","unresolved":false,"context_lines":[{"line_number":21,"context_line":""},{"line_number":22,"context_line":"Background on Offloads"},{"line_number":23,"context_line":"----------------------"},{"line_number":24,"context_line":""},{"line_number":25,"context_line":"Protocol Offloads"},{"line_number":26,"context_line":"~~~~~~~~~~~~~~~~~"},{"line_number":27,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_78ddfd3e","line":24,"updated":"2018-11-30 10:20:11.000000000","message":"I\u0027m contemplating adding the following paragraph here:\n\n    While composing this spec, it became clear that the \"offloads\" term\n    had historical meaning that caused confusion about the scope of this\n    spec. This subsection was added in order to clarify the distinctions\n    between different classes of offloads.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4bf7d7a64f33e519834a142231ef8c4a4725151b","unresolved":false,"context_lines":[{"line_number":21,"context_line":""},{"line_number":22,"context_line":"Background on Offloads"},{"line_number":23,"context_line":"----------------------"},{"line_number":24,"context_line":""},{"line_number":25,"context_line":"Protocol Offloads"},{"line_number":26,"context_line":"~~~~~~~~~~~~~~~~~"},{"line_number":27,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_70179443","line":24,"in_reply_to":"3f79a3b5_2a63eead","updated":"2018-12-10 16:42:56.000000000","message":"Done","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"3ba0fa55f7008ea1626f583966fc66e1816d1af7","unresolved":false,"context_lines":[{"line_number":21,"context_line":""},{"line_number":22,"context_line":"Background on Offloads"},{"line_number":23,"context_line":"----------------------"},{"line_number":24,"context_line":""},{"line_number":25,"context_line":"Protocol Offloads"},{"line_number":26,"context_line":"~~~~~~~~~~~~~~~~~"},{"line_number":27,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_be185d72","line":24,"in_reply_to":"3f79a3b5_78ddfd3e","updated":"2018-11-30 12:38:04.000000000","message":"+1\ni think that makes sense","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"dce454a79032239e074a73d375d49a050bbfc9db","unresolved":false,"context_lines":[{"line_number":21,"context_line":""},{"line_number":22,"context_line":"Background on Offloads"},{"line_number":23,"context_line":"----------------------"},{"line_number":24,"context_line":""},{"line_number":25,"context_line":"Protocol Offloads"},{"line_number":26,"context_line":"~~~~~~~~~~~~~~~~~"},{"line_number":27,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_2a63eead","line":24,"in_reply_to":"3f79a3b5_be185d72","updated":"2018-12-10 15:46:57.000000000","message":"Agreed","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"dce454a79032239e074a73d375d49a050bbfc9db","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"`ethtool` type offloads typically:"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"    #. are available to guests (and hosts),"},{"line_number":36,"context_line":"    #. have a strong relationship with a network endpoint,"},{"line_number":37,"context_line":"    #. have a role with generating and consuming packets,"},{"line_number":38,"context_line":"    #. can be modeled as capabilities of the virtual NIC on the instance."},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"Currently, Nova has little modelling for these types of offload capabilities."},{"line_number":41,"context_line":"Ensuring that instances can live migrate to a compute node capable of"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_2af8cebf","line":38,"range":{"start_line":35,"start_character":0,"end_line":38,"end_character":73},"updated":"2018-12-10 15:46:57.000000000","message":"style: this needs to be dedented to render correctly. Also, s/`ethtool`/``ethtool``/ Ditto for the rest of the document.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4bf7d7a64f33e519834a142231ef8c4a4725151b","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"`ethtool` type offloads typically:"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"    #. are available to guests (and hosts),"},{"line_number":36,"context_line":"    #. have a strong relationship with a network endpoint,"},{"line_number":37,"context_line":"    #. have a role with generating and consuming packets,"},{"line_number":38,"context_line":"    #. can be modeled as capabilities of the virtual NIC on the instance."},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"Currently, Nova has little modelling for these types of offload capabilities."},{"line_number":41,"context_line":"Ensuring that instances can live migrate to a compute node capable of"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_f02ac476","line":38,"range":{"start_line":35,"start_character":0,"end_line":38,"end_character":73},"in_reply_to":"3f79a3b5_2af8cebf","updated":"2018-12-10 16:42:56.000000000","message":"Done","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"174de94b6806b653e43a6714f39802fd3e34821b","unresolved":false,"context_lines":[{"line_number":25,"context_line":"Protocol Offloads"},{"line_number":26,"context_line":"~~~~~~~~~~~~~~~~~"},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"Network-specific computation being handled by dedicated peripherals is well"},{"line_number":29,"context_line":"established on many platforms. For Linux, the `ethtool man page`_ details a"},{"line_number":30,"context_line":"number of settings for the ``--offload`` option that are available on many"},{"line_number":31,"context_line":"NICs, for specific protocols."},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"`ethtool` type offloads typically:"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"    #. are available to guests (and hosts),"},{"line_number":36,"context_line":"    #. have a strong relationship with a network endpoint,"},{"line_number":37,"context_line":"    #. have a role with generating and consuming packets,"},{"line_number":38,"context_line":"    #. can be modeled as capabilities of the virtual NIC on the instance."},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"Currently, Nova has little modelling for these types of offload capabilities."},{"line_number":41,"context_line":"Ensuring that instances can live migrate to a compute node capable of"},{"line_number":42,"context_line":"providing the required features is not something Nova can currently determine"},{"line_number":43,"context_line":"ahead of time."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"This spec only touches lightly on this class of offloads."},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_8d2ab9e2","line":43,"range":{"start_line":28,"start_character":0,"end_line":43,"end_character":14},"updated":"2018-11-21 13:05:03.000000000","message":"this doesn\u0027t seem related to problem or the proposed solution in this spec. by the way there was a spec addressing the offload capabilities but it was partial implemented see [1] \n\n[1] - https://review.openstack.org/#/c/435954/","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":25,"context_line":"Protocol Offloads"},{"line_number":26,"context_line":"~~~~~~~~~~~~~~~~~"},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"Network-specific computation being handled by dedicated peripherals is well"},{"line_number":29,"context_line":"established on many platforms. For Linux, the `ethtool man page`_ details a"},{"line_number":30,"context_line":"number of settings for the ``--offload`` option that are available on many"},{"line_number":31,"context_line":"NICs, for specific protocols."},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"`ethtool` type offloads typically:"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"    #. are available to guests (and hosts),"},{"line_number":36,"context_line":"    #. have a strong relationship with a network endpoint,"},{"line_number":37,"context_line":"    #. have a role with generating and consuming packets,"},{"line_number":38,"context_line":"    #. can be modeled as capabilities of the virtual NIC on the instance."},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"Currently, Nova has little modelling for these types of offload capabilities."},{"line_number":41,"context_line":"Ensuring that instances can live migrate to a compute node capable of"},{"line_number":42,"context_line":"providing the required features is not something Nova can currently determine"},{"line_number":43,"context_line":"ahead of time."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"This spec only touches lightly on this class of offloads."},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_5e63e5de","line":43,"range":{"start_line":28,"start_character":0,"end_line":43,"end_character":14},"in_reply_to":"3f79a3b5_78c7d93e","updated":"2018-11-21 14:49:27.000000000","message":"yes i had a very strong objection to the term offload and datapath offload in partacalar to refer to things like hardware acllaerated ovs unless we had a dedicated term to describe the tradtional offloads such as TSO or vxlan checksum offload.\n\nthis is also related as we intend to also add a protocol_offloads field to track what offloads were configured in the future.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"517051d1c5f2e89a3ecb8f7a85b03dc51cc46e45","unresolved":false,"context_lines":[{"line_number":25,"context_line":"Protocol Offloads"},{"line_number":26,"context_line":"~~~~~~~~~~~~~~~~~"},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"Network-specific computation being handled by dedicated peripherals is well"},{"line_number":29,"context_line":"established on many platforms. For Linux, the `ethtool man page`_ details a"},{"line_number":30,"context_line":"number of settings for the ``--offload`` option that are available on many"},{"line_number":31,"context_line":"NICs, for specific protocols."},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"`ethtool` type offloads typically:"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"    #. are available to guests (and hosts),"},{"line_number":36,"context_line":"    #. have a strong relationship with a network endpoint,"},{"line_number":37,"context_line":"    #. have a role with generating and consuming packets,"},{"line_number":38,"context_line":"    #. can be modeled as capabilities of the virtual NIC on the instance."},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"Currently, Nova has little modelling for these types of offload capabilities."},{"line_number":41,"context_line":"Ensuring that instances can live migrate to a compute node capable of"},{"line_number":42,"context_line":"providing the required features is not something Nova can currently determine"},{"line_number":43,"context_line":"ahead of time."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"This spec only touches lightly on this class of offloads."},{"line_number":46,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_78c7d93e","line":43,"range":{"start_line":28,"start_character":0,"end_line":43,"end_character":14},"in_reply_to":"3f79a3b5_8d2ab9e2","updated":"2018-11-21 14:09:54.000000000","message":"This spec indeed only touches lightly on this class of offloads. (in fact the only other real mention is at https://review.openstack.org/#/c/607610/4/specs/stein/approved/generic-os-vif-offloads.rst@198 )\n\nMostly, I inserted this section because people kept getting confused by what the \u0027offloads\u0027 term refers to: explaining that \u0027datapath\u0027 and \u0027protocol\u0027 offloads are separate things, appear to be something that\u0027s needed in order to understand the rest of the spec.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"174de94b6806b653e43a6714f39802fd3e34821b","unresolved":false,"context_lines":[{"line_number":71,"context_line":"       bypass the host hypervisor entirely."},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"The simplest case of this is an SR-IOV NIC in Virtual Ethernet Bridge (VEB)"},{"line_number":74,"context_line":"mode, as used by the ``mech_sriov`` Neutron driver. A special plugging mode is"},{"line_number":75,"context_line":"necessary, (namely IOMMU PCI passthrough), and the hypervisor configures the"},{"line_number":76,"context_line":"VEB with the required MAC ACL filters."},{"line_number":77,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_cd72f118","line":74,"range":{"start_line":74,"start_character":21,"end_line":74,"end_character":50},"updated":"2018-11-21 13:05:03.000000000","message":"I think it better to call it ``sriovnicswitch`` Neutron mechanism driver","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":71,"context_line":"       bypass the host hypervisor entirely."},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"The simplest case of this is an SR-IOV NIC in Virtual Ethernet Bridge (VEB)"},{"line_number":74,"context_line":"mode, as used by the ``mech_sriov`` Neutron driver. A special plugging mode is"},{"line_number":75,"context_line":"necessary, (namely IOMMU PCI passthrough), and the hypervisor configures the"},{"line_number":76,"context_line":"VEB with the required MAC ACL filters."},{"line_number":77,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_7ea58100","line":74,"range":{"start_line":74,"start_character":21,"end_line":74,"end_character":50},"in_reply_to":"3f79a3b5_78fa7944","updated":"2018-11-21 14:49:27.000000000","message":"python lib name? this is part of the neutron package normally. by i agree with moshe here we shoudl use sriovnicswitch as that is the agent, ml2 and stevadore entrypoint name.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4bf7d7a64f33e519834a142231ef8c4a4725151b","unresolved":false,"context_lines":[{"line_number":71,"context_line":"       bypass the host hypervisor entirely."},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"The simplest case of this is an SR-IOV NIC in Virtual Ethernet Bridge (VEB)"},{"line_number":74,"context_line":"mode, as used by the ``mech_sriov`` Neutron driver. A special plugging mode is"},{"line_number":75,"context_line":"necessary, (namely IOMMU PCI passthrough), and the hypervisor configures the"},{"line_number":76,"context_line":"VEB with the required MAC ACL filters."},{"line_number":77,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_10a980d9","line":74,"range":{"start_line":74,"start_character":21,"end_line":74,"end_character":50},"in_reply_to":"3f79a3b5_7ea58100","updated":"2018-12-10 16:42:56.000000000","message":"Done","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"517051d1c5f2e89a3ecb8f7a85b03dc51cc46e45","unresolved":false,"context_lines":[{"line_number":71,"context_line":"       bypass the host hypervisor entirely."},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"The simplest case of this is an SR-IOV NIC in Virtual Ethernet Bridge (VEB)"},{"line_number":74,"context_line":"mode, as used by the ``mech_sriov`` Neutron driver. A special plugging mode is"},{"line_number":75,"context_line":"necessary, (namely IOMMU PCI passthrough), and the hypervisor configures the"},{"line_number":76,"context_line":"VEB with the required MAC ACL filters."},{"line_number":77,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_78fa7944","line":74,"range":{"start_line":74,"start_character":21,"end_line":74,"end_character":50},"in_reply_to":"3f79a3b5_cd72f118","updated":"2018-11-21 14:09:54.000000000","message":"Thanks, I was looking at the python library, not the driver name...","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":111,"context_line":"plugging, and also info for the datapath control software. For the ``ovs`` VIF,"},{"line_number":112,"context_line":"the hypervisor is generally able to also perform the datapath control, but this"},{"line_number":113,"context_line":"is not the case for every VIF type (hence the existence of os-vif)."},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":116,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":117,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":118,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":119,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_9e4f3d30","line":116,"range":{"start_line":114,"start_character":0,"end_line":116,"end_character":50},"updated":"2018-11-21 14:49:27.000000000","message":"not really it has influneced both the xml generation and how the interface is plugged ever since neutron sriov was introduced and it was used to dfferenciate between macvtap and direct mode. so it not a recent change.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4bf7d7a64f33e519834a142231ef8c4a4725151b","unresolved":false,"context_lines":[{"line_number":111,"context_line":"plugging, and also info for the datapath control software. For the ``ovs`` VIF,"},{"line_number":112,"context_line":"the hypervisor is generally able to also perform the datapath control, but this"},{"line_number":113,"context_line":"is not the case for every VIF type (hence the existence of os-vif)."},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":116,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":117,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":118,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":119,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_1082e04b","line":116,"range":{"start_line":114,"start_character":0,"end_line":116,"end_character":50},"in_reply_to":"3f79a3b5_324c83c9","updated":"2018-12-10 16:42:56.000000000","message":"Done","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":111,"context_line":"plugging, and also info for the datapath control software. For the ``ovs`` VIF,"},{"line_number":112,"context_line":"the hypervisor is generally able to also perform the datapath control, but this"},{"line_number":113,"context_line":"is not the case for every VIF type (hence the existence of os-vif)."},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":116,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":117,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":118,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":119,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_e39087f5","line":116,"range":{"start_line":114,"start_character":0,"end_line":116,"end_character":50},"in_reply_to":"3f79a3b5_9e4f3d30","updated":"2018-11-30 03:15:06.000000000","message":"Crystal clear :(","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"db2ac2cddd445c8bcb998fa4ebd0bb4c171a4c38","unresolved":false,"context_lines":[{"line_number":111,"context_line":"plugging, and also info for the datapath control software. For the ``ovs`` VIF,"},{"line_number":112,"context_line":"the hypervisor is generally able to also perform the datapath control, but this"},{"line_number":113,"context_line":"is not the case for every VIF type (hence the existence of os-vif)."},{"line_number":114,"context_line":""},{"line_number":115,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":116,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":117,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":118,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":119,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_324c83c9","line":116,"range":{"start_line":114,"start_character":0,"end_line":116,"end_character":50},"in_reply_to":"3f79a3b5_e39087f5","updated":"2018-11-30 10:20:11.000000000","message":"I think I\u0027ll remove \"recently\", with no real loss.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"174de94b6806b653e43a6714f39802fd3e34821b","unresolved":false,"context_lines":[{"line_number":115,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":116,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":117,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":118,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":119,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"* The ``normal`` (or default) VNIC type indicates that the Instance is plugged"},{"line_number":122,"context_line":"  into the software bridge."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_cdb75188","line":119,"range":{"start_line":118,"start_character":55,"end_line":119,"end_character":29},"updated":"2018-11-21 13:05:03.000000000","message":"openvswith is using the following vnic_type: noraml, direct with switchdev capability or virtio-forwarder.\n  \nThe normal can be ovs-kernel or ovs-dpdk (offload is just in case of ovs-dpdk.)","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":115,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":116,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":117,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":118,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":119,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"* The ``normal`` (or default) VNIC type indicates that the Instance is plugged"},{"line_number":122,"context_line":"  into the software bridge."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_637c77c7","line":119,"range":{"start_line":118,"start_character":55,"end_line":119,"end_character":29},"in_reply_to":"3f79a3b5_1e632dae","updated":"2018-11-30 03:15:06.000000000","message":"ugh. why do networking vendors need to make all of this stuff so incomprehensible :(","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"517051d1c5f2e89a3ecb8f7a85b03dc51cc46e45","unresolved":false,"context_lines":[{"line_number":115,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":116,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":117,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":118,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":119,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"* The ``normal`` (or default) VNIC type indicates that the Instance is plugged"},{"line_number":122,"context_line":"  into the software bridge."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_d8b70d88","line":119,"range":{"start_line":118,"start_character":55,"end_line":119,"end_character":29},"in_reply_to":"3f79a3b5_cdb75188","updated":"2018-11-21 14:09:54.000000000","message":"You\u0027re right, OVS uses normal and direct, AOVS added virtio-forwarder.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4bf7d7a64f33e519834a142231ef8c4a4725151b","unresolved":false,"context_lines":[{"line_number":115,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":116,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":117,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":118,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":119,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"* The ``normal`` (or default) VNIC type indicates that the Instance is plugged"},{"line_number":122,"context_line":"  into the software bridge."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_90eb500e","line":119,"range":{"start_line":118,"start_character":55,"end_line":119,"end_character":29},"in_reply_to":"3f79a3b5_d8b70d88","updated":"2018-12-10 16:42:56.000000000","message":"Done","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":115,"context_line":"The VNIC type is a property of a VIF. It has recently taken on the semantics of"},{"line_number":116,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":117,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":118,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":119,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"* The ``normal`` (or default) VNIC type indicates that the Instance is plugged"},{"line_number":122,"context_line":"  into the software bridge."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_1e632dae","line":119,"range":{"start_line":118,"start_character":55,"end_line":119,"end_character":29},"in_reply_to":"3f79a3b5_d8b70d88","updated":"2018-11-21 14:49:27.000000000","message":"intel/my former team had planned on adding vhost-vfio in the future too but not sure if that will go ahead anymore.\n\nvhost-vfio was for a vfio mdev based approch that exposes a standard virtio-net interface to the guest as in normal or virtio-forward case but enable a  fully hardware offloaded datapath like the direct mode while also supporting livemigration.\n\n\novs-dpdk may support vnictypes other then normal at some point in the future too but for now it only works with normal\nvhost-vfio had been planned to work with both kernel and dpdk ovs.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":125,"context_line":"* The ``virtio-forwarder`` VNIC type indicates that a VF is attached via a"},{"line_number":126,"context_line":"  `virtio-forwarder`_."},{"line_number":127,"context_line":""},{"line_number":128,"context_line":"Currently, os-vif and Nova implement `switchdev SR-IOV offloads`_ for Open"},{"line_number":129,"context_line":"vSwitch with ``tc-flower`` offloads. In this model, a representor netdev on the"},{"line_number":130,"context_line":"host is associated with each Virtual Function. This representor functions like"},{"line_number":131,"context_line":"a handle for the corresponding virtual port on the NIC\u0027s packet processing"},{"line_number":132,"context_line":"pipeline."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"},{"line_number":135,"context_line":"plugin. Optionally, a netdev name can also be passed to allow for friendly"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_c3b46b87","line":132,"range":{"start_line":128,"start_character":0,"end_line":132,"end_character":9},"updated":"2018-11-30 03:15:06.000000000","message":"this is well-written and clear. Hopefully also correct.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"174de94b6806b653e43a6714f39802fd3e34821b","unresolved":false,"context_lines":[{"line_number":132,"context_line":"pipeline."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"},{"line_number":135,"context_line":"plugin. Optionally, a netdev name can also be passed to allow for friendly"},{"line_number":136,"context_line":"renaming of the representor by the os-vif plugin."},{"line_number":137,"context_line":""},{"line_number":138,"context_line":"The ``ovs`` and ``agilio_ovs`` os-vif plugins then look up the associated"},{"line_number":139,"context_line":"representor for the VF and perform the datapath plugging. From Nova\u0027s"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_4d1ee175","line":136,"range":{"start_line":135,"start_character":8,"end_line":136,"end_character":49},"updated":"2018-11-21 13:05:03.000000000","message":"I don\u0027t think it is implemented with the ovs plugin maybe it work like that with agilio_ovs.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"517051d1c5f2e89a3ecb8f7a85b03dc51cc46e45","unresolved":false,"context_lines":[{"line_number":132,"context_line":"pipeline."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"},{"line_number":135,"context_line":"plugin. Optionally, a netdev name can also be passed to allow for friendly"},{"line_number":136,"context_line":"renaming of the representor by the os-vif plugin."},{"line_number":137,"context_line":""},{"line_number":138,"context_line":"The ``ovs`` and ``agilio_ovs`` os-vif plugins then look up the associated"},{"line_number":139,"context_line":"representor for the VF and perform the datapath plugging. From Nova\u0027s"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_98ad1536","line":136,"range":{"start_line":135,"start_character":8,"end_line":136,"end_character":49},"in_reply_to":"3f79a3b5_4d1ee175","updated":"2018-11-21 14:09:54.000000000","message":"Yep, it\u0027s not implemented like that in the reference OVS plugin, but it\u0027s documented like that in the data model.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"db2ac2cddd445c8bcb998fa4ebd0bb4c171a4c38","unresolved":false,"context_lines":[{"line_number":132,"context_line":"pipeline."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"},{"line_number":135,"context_line":"plugin. Optionally, a netdev name can also be passed to allow for friendly"},{"line_number":136,"context_line":"renaming of the representor by the os-vif plugin."},{"line_number":137,"context_line":""},{"line_number":138,"context_line":"The ``ovs`` and ``agilio_ovs`` os-vif plugins then look up the associated"},{"line_number":139,"context_line":"representor for the VF and perform the datapath plugging. From Nova\u0027s"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_920f9737","line":136,"range":{"start_line":135,"start_character":8,"end_line":136,"end_character":49},"in_reply_to":"3f79a3b5_5cd0c3f5","updated":"2018-11-30 10:20:11.000000000","message":"I think the only reason I had not pushed a patch for this on the standard OVS-TC case is because the representors would need to be renamed back to their original name. I have some ideas on that, but there\u0027s no clear \"winner\".","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"49309e12a98d396a9aeb1207a6a59e2dfffbdb77","unresolved":false,"context_lines":[{"line_number":132,"context_line":"pipeline."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"},{"line_number":135,"context_line":"plugin. Optionally, a netdev name can also be passed to allow for friendly"},{"line_number":136,"context_line":"renaming of the representor by the os-vif plugin."},{"line_number":137,"context_line":""},{"line_number":138,"context_line":"The ``ovs`` and ``agilio_ovs`` os-vif plugins then look up the associated"},{"line_number":139,"context_line":"representor for the VF and perform the datapath plugging. From Nova\u0027s"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_5cd0c3f5","line":136,"range":{"start_line":135,"start_character":8,"end_line":136,"end_character":49},"in_reply_to":"3f79a3b5_98ad1536","updated":"2018-11-22 07:04:44.000000000","message":"Maybe we should do it, it will be nice if we can link the VF representor to prefix of neutron port. I will do it if I have time...","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":132,"context_line":"pipeline."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"Nova passes the PCI address it received from the PCI request to the os-vif"},{"line_number":135,"context_line":"plugin. Optionally, a netdev name can also be passed to allow for friendly"},{"line_number":136,"context_line":"renaming of the representor by the os-vif plugin."},{"line_number":137,"context_line":""},{"line_number":138,"context_line":"The ``ovs`` and ``agilio_ovs`` os-vif plugins then look up the associated"},{"line_number":139,"context_line":"representor for the VF and perform the datapath plugging. From Nova\u0027s"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_9ea09da5","line":136,"range":{"start_line":135,"start_character":8,"end_line":136,"end_character":49},"in_reply_to":"3f79a3b5_98ad1536","updated":"2018-11-21 14:49:27.000000000","message":"we had planned tp  support this in the vhost-vfio case also but it was a lower priortiy.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":153,"context_line":"a developer, using a different datapath, wish to pass offload metadata to an"},{"line_number":154,"context_line":"os-vif plugin, they would have to extend the object model, or pass the metadata"},{"line_number":155,"context_line":"using a confusingly named object. This spec aims to establish a recommended"},{"line_number":156,"context_line":"mechanism to extend the object model."},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"Proposed change"},{"line_number":159,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_63a55749","line":156,"updated":"2018-11-30 03:15:06.000000000","message":"++","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"3ba0fa55f7008ea1626f583966fc66e1816d1af7","unresolved":false,"context_lines":[{"line_number":158,"context_line":"Proposed change"},{"line_number":159,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":160,"context_line":""},{"line_number":161,"context_line":"Option 1: Use composition instead of inheritance"},{"line_number":162,"context_line":"------------------------------------------------"},{"line_number":163,"context_line":""},{"line_number":164,"context_line":"Instead of using an inheritance based pattern to model the offload"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_9e880126","line":161,"range":{"start_line":161,"start_character":0,"end_line":161,"end_character":48},"updated":"2018-11-30 12:38:04.000000000","message":"by they way this does not actully say we are going with option 1 anywhere.\n\nwe proably should state that","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"dce454a79032239e074a73d375d49a050bbfc9db","unresolved":false,"context_lines":[{"line_number":158,"context_line":"Proposed change"},{"line_number":159,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":160,"context_line":""},{"line_number":161,"context_line":"Option 1: Use composition instead of inheritance"},{"line_number":162,"context_line":"------------------------------------------------"},{"line_number":163,"context_line":""},{"line_number":164,"context_line":"Instead of using an inheritance based pattern to model the offload"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_0adab21a","line":161,"range":{"start_line":161,"start_character":0,"end_line":161,"end_character":48},"in_reply_to":"3f79a3b5_9e880126","updated":"2018-12-10 15:46:57.000000000","message":"From reading the commit message, we\u0027re going to decide which option to take once we get down to implementation. At the moment, there are pros and cons to both and I guess there isn\u0027t a clear consensus.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":192,"context_line":"Instead of bumping object versions and creating composition version maps, this"},{"line_number":193,"context_line":"option proposes that versioning be deliberately ignored until the next major"},{"line_number":194,"context_line":"release of os-vif. At this point, version negotiation and backlevelling in"},{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_239f5ff7","line":195,"updated":"2018-11-30 03:15:06.000000000","message":"++","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"174de94b6806b653e43a6714f39802fd3e34821b","unresolved":false,"context_lines":[{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":201,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_78a59900","line":198,"range":{"start_line":198,"start_character":36,"end_line":198,"end_character":79},"updated":"2018-11-21 13:05:03.000000000","message":"why do you need the ``protocol_offloads`` list of capabilities. At the time we agreed that this should taken from libvirt using this and used for scheduling: \n\u003cdevice\u003e\n\u003cname\u003enet_eth4_90_e2_ba_5e_a5_45\u003c/name\u003e\n  \u003cpath\u003e/sys/devices/pci0000:00/0000:00:03.0/0000:08:00.1/net/eth4\u003c/path\u003e\n  \u003cparent\u003epci_0000_08_00_1\u003c/parent\u003e\n  \u003ccapability type\u003d\u0027net\u0027\u003e\n    \u003cinterface\u003eeth4\u003c/interface\u003e\n    \u003caddress\u003e90:e2:ba:5e:a5:45\u003c/address\u003e\n    \u003clink speed\u003d\u002710000\u0027 state\u003d\u0027up\u0027/\u003e\n    \u003cfeature name\u003d\u0027rx\u0027/\u003e\n    \u003cfeature name\u003d\u0027tx\u0027/\u003e\n    \u003cfeature name\u003d\u0027sg\u0027/\u003e\n    \u003cfeature name\u003d\u0027tso\u0027/\u003e\n    \u003cfeature name\u003d\u0027gso\u0027/\u003e\n    \u003cfeature name\u003d\u0027gro\u0027/\u003e\n    \u003cfeature name\u003d\u0027rxvlan\u0027/\u003e\n    \u003cfeature name\u003d\u0027txvlan\u0027/\u003e\n    \u003cfeature name\u003d\u0027rxhash\u0027/\u003e\n    \u003ccapability type\u003d\u002780203\u0027/\u003e\n  \u003c/capability\u003e\n\u003c/device\u003e","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"db2ac2cddd445c8bcb998fa4ebd0bb4c171a4c38","unresolved":false,"context_lines":[{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":201,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_781e5dbe","line":198,"range":{"start_line":198,"start_character":36,"end_line":198,"end_character":79},"in_reply_to":"3f79a3b5_23743fcd","updated":"2018-11-30 10:20:11.000000000","message":"I\u0027m thinking of doing this in the respin:\n\n    s/``VIFPortProfileBase``/``VIFBase`` or ``VIFPortProfileBase``/\n\nI agree with Jay: It should actually be made very clear what the entries are, where they originate and how they should be set... all of which leads to spec sprawl. I can start on the skeleton of a second spec if it\u0027s a blocker, but I don\u0027t have a use case to fix on this (yet)...","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":201,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_23743fcd","line":198,"range":{"start_line":198,"start_character":36,"end_line":198,"end_character":79},"in_reply_to":"3f79a3b5_28a76553","updated":"2018-11-30 03:15:06.000000000","message":"While the use cases around a protocol_offloads field have potential, I don\u0027t believe we should add anything more than a cursory mention of a possible protocol_offloads field to this spec (let\u0027s keep this spec as focused as possible on the datapath_offloads field and only mention protocol_offloads field as a reason we named the thing datapath_offloads and not just \"offloads\").","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":201,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_3e84e956","line":198,"range":{"start_line":198,"start_character":36,"end_line":198,"end_character":79},"in_reply_to":"3f79a3b5_382161b3","updated":"2018-11-21 14:49:27.000000000","message":"actully i would prefer to add the field now ranter then later or at least add a todo so that when we do add it in the future we dont have to debate the name again.\n\nim not sure i agree this should be in the VIFBase object but ill have to think about it. it could.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"49309e12a98d396a9aeb1207a6a59e2dfffbdb77","unresolved":false,"context_lines":[{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":201,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_bc355745","line":198,"range":{"start_line":198,"start_character":36,"end_line":198,"end_character":79},"in_reply_to":"3f79a3b5_3e84e956","updated":"2018-11-22 07:04:44.000000000","message":"@Sean, Jan mention that you have a use case to have capabilities to travel along the VIF. What is the use-case can we add it to the spec?","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4bf7d7a64f33e519834a142231ef8c4a4725151b","unresolved":false,"context_lines":[{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":201,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_30f27ce5","line":198,"range":{"start_line":198,"start_character":36,"end_line":198,"end_character":79},"in_reply_to":"3f79a3b5_781e5dbe","updated":"2018-12-10 16:42:56.000000000","message":"Done","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"3ba0fa55f7008ea1626f583966fc66e1816d1af7","unresolved":false,"context_lines":[{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":201,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_7eef25fb","line":198,"range":{"start_line":198,"start_character":36,"end_line":198,"end_character":79},"in_reply_to":"3f79a3b5_781e5dbe","updated":"2018-11-30 12:38:04.000000000","message":"i am still leaning towards VIFPortProfileBase\n\nthe vif objects are intended to model the atripbute of the interface that are implemnetaiton independant and can be supported by all back ends. \n\ndata plane offloads definitely do not fit that category which is why the are plased in the port profile which models implementation specific info.\n\ni can see the argument being made that protocl offloads could be generic enough to include in the vifbase but its not clear to me that vmware/hyperv vif would ever use/support them so \nmy prefernce would be to group the with the backend sepcific info but i would be open to the VIF base if we had a non libvirt hyperviors that would supprot them to show they would be generic enough.\n\nanyway we can leave that to a later time to figure out since we are not going to add the field until we have code/a feature that needs it.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"517051d1c5f2e89a3ecb8f7a85b03dc51cc46e45","unresolved":false,"context_lines":[{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":201,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_382161b3","line":198,"range":{"start_line":198,"start_character":36,"end_line":198,"end_character":79},"in_reply_to":"3f79a3b5_78a59900","updated":"2018-11-21 14:09:54.000000000","message":"This spec does not propose to add the list of capabilities to the VIF object, it\u0027s here to illustrate that, in theory, the port capabilities could be exposed in this fashion. The hand-wavy usecase for this type of metadata is described in https://review.openstack.org/#/c/607610/4/specs/stein/approved/generic-os-vif-offloads.rst@40\n\nUpon reflection, it makes a _lot_ more sense that these capabilities apply to VIFBase, not VIFPortProfileBase (if the hypervisor does not take care of them). sean-k-mooney mentioned a case where it _might_ be useful to have the capabilities travel along the VIF rather than via the hypervisor capabilities.\n\nIn any case, this section serves as an example contrasting protocol offloads with datapath offloads, it\u0027s not intended to be implemented by this spec.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"a39ab071e1ae6d4535cac14d038c0ca0470e3380","unresolved":false,"context_lines":[{"line_number":195,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":196,"context_line":""},{"line_number":197,"context_line":"Should protocol offloads also need to be modeled in os-vif,"},{"line_number":198,"context_line":"``VIFPortProfileBase`` could gain a ``protocol_offloads`` list of capabilities."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"Summary of plugging methods affected for Option 1"},{"line_number":201,"context_line":"-------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_28a76553","line":198,"range":{"start_line":198,"start_character":36,"end_line":198,"end_character":79},"in_reply_to":"3f79a3b5_bc355745","updated":"2018-11-28 13:42:36.000000000","message":"there are two future looking usecase for this\n\nfirsth this would provide a vector for neutron to specify offloads that must be enabled in the future similar\nto the sriov nic feature flags that https://review.openstack.org/#/c/435954/ is now abandoned but more general. e.g. it would allow you to say i want a port that has GSO enabled.\n\nsecond it may provided to adress the gap in ovs migration that exitst today namely the fact it only works if teh virtio feature flags supported by the destination host are a weak super set of the features that were enabled in the interface.\nso if you have a cloud that you rolling upgrade and therefor may have different version of qemu/kvm/ovs deployed it is not safe to migrate form a new node to an older node today.\n\nif we were to store the vif object in the future we could use this field to encode the virtio flags that were enabled and used that to validate there support in the scheduler.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":241,"context_line":"    * VNIC type: ``virtio-forwarder``"},{"line_number":242,"context_line":"    * os-vif object: ``VIFVHostUser``"},{"line_number":243,"context_line":"    * ``port_profile: VIFPortProfileOpenVSwitch``"},{"line_number":244,"context_line":"    * ``port_profile.datapath_offload: DatapathOffloadRepresentor``"},{"line_number":245,"context_line":""},{"line_number":246,"context_line":"Option 2: Use inheritance to create a generic representor profile"},{"line_number":247,"context_line":"-----------------------------------------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_e3472779","line":244,"updated":"2018-11-30 03:15:06.000000000","message":"++ on above. as mentioned in a previous revision, seeing the before/after like this is very useful.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":270,"context_line":""},{"line_number":271,"context_line":"    * VNIC type: ``virtio-forwarder``"},{"line_number":272,"context_line":"    * os-vif object: ``VIFVHostUser``"},{"line_number":273,"context_line":"    * ``port_profile: VIFPortProfileRepresentor``"},{"line_number":274,"context_line":""},{"line_number":275,"context_line":"Additional Impact for Both Options"},{"line_number":276,"context_line":"----------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_a3412f65","line":273,"updated":"2018-11-30 03:15:06.000000000","message":"-1 for the reasons noted previously... it\u0027s an incorrect model and not future-proof.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"3ba0fa55f7008ea1626f583966fc66e1816d1af7","unresolved":false,"context_lines":[{"line_number":270,"context_line":""},{"line_number":271,"context_line":"    * VNIC type: ``virtio-forwarder``"},{"line_number":272,"context_line":"    * os-vif object: ``VIFVHostUser``"},{"line_number":273,"context_line":"    * ``port_profile: VIFPortProfileRepresentor``"},{"line_number":274,"context_line":""},{"line_number":275,"context_line":"Additional Impact for Both Options"},{"line_number":276,"context_line":"----------------------------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_beddbd13","line":273,"in_reply_to":"3f79a3b5_a3412f65","updated":"2018-11-30 12:38:04.000000000","message":"it jsut as future proof as option 1\nboth still use inheritance and composition they just do it at different places and with different scopes.\n\nsince the consensus seams to be drifting to option 1 however\nthis change will likely motivate looking at alternative to ovo in the future when serialization is important as the\nmain drawback to option 1 is the serialized form of the VIF object will get appreciably bigger.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"3ba0fa55f7008ea1626f583966fc66e1816d1af7","unresolved":false,"context_lines":[{"line_number":285,"context_line":"Additional notes"},{"line_number":286,"context_line":"----------------"},{"line_number":287,"context_line":""},{"line_number":288,"context_line":"* No corresponding changes in Neutron are expected: currently os-vif is Nova"},{"line_number":289,"context_line":"  only."},{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_deb19902","line":289,"range":{"start_line":288,"start_character":71,"end_line":289,"end_character":7},"updated":"2018-11-30 12:38:04.000000000","message":"its used in kuryer kuberneties too.\nand they are useing the libvary out of contrack by\ndefining vif objecst themselve which has never been supported and never will be.\n\ni bring this up as it may impact them and if it does the fix will be to replace there external vif types/vif profiles with ones in os-vif","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4bf7d7a64f33e519834a142231ef8c4a4725151b","unresolved":false,"context_lines":[{"line_number":285,"context_line":"Additional notes"},{"line_number":286,"context_line":"----------------"},{"line_number":287,"context_line":""},{"line_number":288,"context_line":"* No corresponding changes in Neutron are expected: currently os-vif is Nova"},{"line_number":289,"context_line":"  only."},{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_10e180fa","line":289,"range":{"start_line":288,"start_character":71,"end_line":289,"end_character":7},"in_reply_to":"3f79a3b5_deb19902","updated":"2018-12-10 16:42:56.000000000","message":"Done","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":287,"context_line":""},{"line_number":288,"context_line":"* No corresponding changes in Neutron are expected: currently os-vif is Nova"},{"line_number":289,"context_line":"  only."},{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"},{"line_number":293,"context_line":"  methods of describing representors. Leaving the address member a string was"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_43561b2f","line":290,"range":{"start_line":290,"start_character":26,"end_line":290,"end_character":34},"updated":"2018-11-30 03:15:06.000000000","message":"addresses","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"4bf7d7a64f33e519834a142231ef8c4a4725151b","unresolved":false,"context_lines":[{"line_number":287,"context_line":""},{"line_number":288,"context_line":"* No corresponding changes in Neutron are expected: currently os-vif is Nova"},{"line_number":289,"context_line":"  only."},{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"},{"line_number":293,"context_line":"  methods of describing representors. Leaving the address member a string was"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_b0b52cf2","line":290,"range":{"start_line":290,"start_character":26,"end_line":290,"end_character":34},"in_reply_to":"3f79a3b5_43561b2f","updated":"2018-12-10 16:42:56.000000000","message":"Done","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"174de94b6806b653e43a6714f39802fd3e34821b","unresolved":false,"context_lines":[{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"},{"line_number":293,"context_line":"  methods of describing representors. Leaving the address member a string was"},{"line_number":294,"context_line":"  deemed an acceptable compromise."},{"line_number":295,"context_line":""},{"line_number":296,"context_line":"Alternatives"},{"line_number":297,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_78efd9f9","line":294,"range":{"start_line":293,"start_character":38,"end_line":294,"end_character":34},"updated":"2018-11-21 13:05:03.000000000","message":"I don\u0027t think it a good idea to make it a string it will open for developer bugs.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"517051d1c5f2e89a3ecb8f7a85b03dc51cc46e45","unresolved":false,"context_lines":[{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"},{"line_number":293,"context_line":"  methods of describing representors. Leaving the address member a string was"},{"line_number":294,"context_line":"  deemed an acceptable compromise."},{"line_number":295,"context_line":""},{"line_number":296,"context_line":"Alternatives"},{"line_number":297,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_93d2d631","line":294,"range":{"start_line":293,"start_character":38,"end_line":294,"end_character":34},"in_reply_to":"3f79a3b5_78efd9f9","updated":"2018-11-21 14:09:54.000000000","message":"I am agnostic - I\u0027ll defer to Nova cores decision here: my crystal ball is completely busted.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"},{"line_number":293,"context_line":"  methods of describing representors. Leaving the address member a string was"},{"line_number":294,"context_line":"  deemed an acceptable compromise."},{"line_number":295,"context_line":""},{"line_number":296,"context_line":"Alternatives"},{"line_number":297,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_9e8e5d23","line":294,"range":{"start_line":293,"start_character":38,"end_line":294,"end_character":34},"in_reply_to":"3f79a3b5_93d2d631","updated":"2018-11-21 14:49:27.000000000","message":"the main concern with restringing to a pci adress object is that powerpc uses uuid not tradtional pci address and we want this to work on multiple architectures.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"49309e12a98d396a9aeb1207a6a59e2dfffbdb77","unresolved":false,"context_lines":[{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"},{"line_number":293,"context_line":"  methods of describing representors. Leaving the address member a string was"},{"line_number":294,"context_line":"  deemed an acceptable compromise."},{"line_number":295,"context_line":""},{"line_number":296,"context_line":"Alternatives"},{"line_number":297,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_dc94d3e8","line":294,"range":{"start_line":293,"start_character":38,"end_line":294,"end_character":34},"in_reply_to":"3f79a3b5_9e8e5d23","updated":"2018-11-22 07:04:44.000000000","message":"Ok this seem like a general problem in nova not just os-vif, right?","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"0994748ebe0d583089e5fdfd46506553e501740f","unresolved":false,"context_lines":[{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"},{"line_number":293,"context_line":"  methods of describing representors. Leaving the address member a string was"},{"line_number":294,"context_line":"  deemed an acceptable compromise."},{"line_number":295,"context_line":""},{"line_number":296,"context_line":"Alternatives"},{"line_number":297,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_234d1f93","line":294,"range":{"start_line":293,"start_character":38,"end_line":294,"end_character":34},"in_reply_to":"3f79a3b5_abb92feb","updated":"2018-11-30 03:15:06.000000000","message":"I agree with what Sean says above.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"7c8b560c4749d181d067a0e4e5e6b2fb37a5bd8a","unresolved":false,"context_lines":[{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"},{"line_number":293,"context_line":"  methods of describing representors. Leaving the address member a string was"},{"line_number":294,"context_line":"  deemed an acceptable compromise."},{"line_number":295,"context_line":""},{"line_number":296,"context_line":"Alternatives"},{"line_number":297,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_a89295f9","line":294,"range":{"start_line":293,"start_character":38,"end_line":294,"end_character":34},"in_reply_to":"3f79a3b5_dc94d3e8","updated":"2018-11-28 13:09:03.000000000","message":"The PCIAddress ovo object could be updated:\n\nhttps://github.com/openstack/oslo.versionedobjects/blob/stable/rocky/oslo_versionedobjects/fields.py#L393\n\nThat has decent far-and-wide consequences. From the code, the only type checking it actually enforces is enforcing the string to match a particular regular expression.\n\nStrictly speaking, the only time this verification would trip up is with a glorious bug. It might as well be an opaque handle, since the only two interested parties in this is the API VIF-to-os-vif conversion routines and the os-vif plugin receiving the object. Nova doesn\u0027t send it to libvirt. And it\u0027s difficult to see a scenario where the plugin will not _know_ the format of this member before it reads it.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"a39ab071e1ae6d4535cac14d038c0ca0470e3380","unresolved":false,"context_lines":[{"line_number":290,"context_line":"* Even though representor adresses are currently modeled as PCI address"},{"line_number":291,"context_line":"  objects, it was felt that stricter type checking would be of limited"},{"line_number":292,"context_line":"  benefit. Future networking systems might require paths, UUIDs or other"},{"line_number":293,"context_line":"  methods of describing representors. Leaving the address member a string was"},{"line_number":294,"context_line":"  deemed an acceptable compromise."},{"line_number":295,"context_line":""},{"line_number":296,"context_line":"Alternatives"},{"line_number":297,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_abb92feb","line":294,"range":{"start_line":293,"start_character":38,"end_line":294,"end_character":34},"in_reply_to":"3f79a3b5_dc94d3e8","updated":"2018-11-28 13:42:36.000000000","message":"thats debatable. yes but we should not over constrian here if we dont need too. The intree ovs plugin will continue to use pci address only but other plugins may support other formats.\nwe can do addtional validation of the addres in the ovs plugin if we choose to to enfroce the format but i think we\ncan avoid that.\n\neven in the powervm case the pci address do exist at the kernel level as far as im aware the powervm hypervior just does not expose them to nova.\n\nbut in any case the VIFPortProfileRepresentor will _always_\nbe used to model kernel swtichdev representer netdev VF address\n\nif you are useing a different kind of logical port\nthat dose not implement the swtichdev representer netdev in the kernel then this profile should not be used.\n\nfor example if you are using a vfio_mediated device instead of a representer netdev then that would be a different profile so the value of representer address is not unbounded.\nit has to model a address to a switch dev port representer the format of which may depend on the plugin that consumes it.\n\nif you are really concerned about this change we could add a additional enum/string field that tracked how address filed is formatted but i think this is not needed at this time.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":12171,"name":"Moshe Levi","email":"moshele@nvidia.com","username":"moshele"},"change_message_id":"174de94b6806b653e43a6714f39802fd3e34821b","unresolved":false,"context_lines":[{"line_number":403,"context_line":""},{"line_number":404,"context_line":"* Unit tests for the os-vif changes will test the object model impact."},{"line_number":405,"context_line":""},{"line_number":406,"context_line":"* Third-party CI is already testing the accelerated plugging modes, no new"},{"line_number":407,"context_line":"  new functionality needs to be tested."},{"line_number":408,"context_line":""},{"line_number":409,"context_line":"Documentation Impact"},{"line_number":410,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_d8b5cdf1","line":407,"range":{"start_line":406,"start_character":0,"end_line":407,"end_character":38},"updated":"2018-11-21 13:05:03.000000000","message":"Mellanox Third-party CI only test the ovs with swithcdev mode. Does Netronume as CI as well for the agilio_ovs?","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"90e5da0327e240231fb024e4022423869dcab790","unresolved":false,"context_lines":[{"line_number":403,"context_line":""},{"line_number":404,"context_line":"* Unit tests for the os-vif changes will test the object model impact."},{"line_number":405,"context_line":""},{"line_number":406,"context_line":"* Third-party CI is already testing the accelerated plugging modes, no new"},{"line_number":407,"context_line":"  new functionality needs to be tested."},{"line_number":408,"context_line":""},{"line_number":409,"context_line":"Documentation Impact"},{"line_number":410,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_deb135e5","line":407,"range":{"start_line":406,"start_character":0,"end_line":407,"end_character":38},"in_reply_to":"3f79a3b5_13bb26bf","updated":"2018-11-21 14:49:27.000000000","message":"i think this is still acurate.\n\nwhen vrouter support for offload lands it would be nice to have more ci for os-vif but i would not block this spec on requiring it.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"517051d1c5f2e89a3ecb8f7a85b03dc51cc46e45","unresolved":false,"context_lines":[{"line_number":403,"context_line":""},{"line_number":404,"context_line":"* Unit tests for the os-vif changes will test the object model impact."},{"line_number":405,"context_line":""},{"line_number":406,"context_line":"* Third-party CI is already testing the accelerated plugging modes, no new"},{"line_number":407,"context_line":"  new functionality needs to be tested."},{"line_number":408,"context_line":""},{"line_number":409,"context_line":"Documentation Impact"},{"line_number":410,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_13bb26bf","line":407,"range":{"start_line":406,"start_character":0,"end_line":407,"end_character":38},"in_reply_to":"3f79a3b5_d8b5cdf1","updated":"2018-11-21 14:09:54.000000000","message":"Netronome\u0027s CI for agilio_ovs is sadly out of date. Implementing this spec will allow work to be scheduled on it. At this moment, Pike and Queens passed manual verification.\n\nFollow-on work to test Tungsten Fabric is the core reason for this spec.","commit_id":"a2b939148e02982a260b3fbe7fe21f89729fde88"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"65834e663bd5512840d1756b344d9a9d16a197f0","unresolved":false,"context_lines":[{"line_number":99,"context_line":"``VIFPortProfileOVSRepresentor`` port profile object. This is used by the"},{"line_number":100,"context_line":"``ovs`` reference plugin and the external ``agilio_ovs`` plugin."},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"With ``vrouter`` being a potential third user of such metadata (proposed in a"},{"line_number":103,"context_line":"follow-on spec), it\u0027s worthwhile to abstract the interface before the pattern"},{"line_number":104,"context_line":"solidifies further."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":"This spec is limited to refactoring the interface, with future expansion in"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_87217fb5","line":103,"range":{"start_line":102,"start_character":63,"end_line":103,"end_character":15},"updated":"2018-12-13 23:08:08.000000000","message":"This: https://review.openstack.org/#/c/567148/\n\nOK, could have linked to the blueprint too:\n\nhttps://blueprints.launchpad.net/nova/+spec/vrouter-hw-offloads\n\nsince the gerrit series will be lost later when looking at this.","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"9643cd3300e9fd43ab1d8d2105992b9be2e19e7c","unresolved":false,"context_lines":[{"line_number":99,"context_line":"``VIFPortProfileOVSRepresentor`` port profile object. This is used by the"},{"line_number":100,"context_line":"``ovs`` reference plugin and the external ``agilio_ovs`` plugin."},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"With ``vrouter`` being a potential third user of such metadata (proposed in a"},{"line_number":103,"context_line":"follow-on spec), it\u0027s worthwhile to abstract the interface before the pattern"},{"line_number":104,"context_line":"solidifies further."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":"This spec is limited to refactoring the interface, with future expansion in"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_c2ff5e66","line":103,"range":{"start_line":102,"start_character":63,"end_line":103,"end_character":15},"in_reply_to":"3f79a3b5_87217fb5","updated":"2018-12-14 11:00:14.000000000","message":"Thanks, will add the links in the next respin.","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"65834e663bd5512840d1756b344d9a9d16a197f0","unresolved":false,"context_lines":[{"line_number":120,"context_line":"The VNIC type is a property of a VIF. It has taken on the semantics of"},{"line_number":121,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":122,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":123,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":124,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"* The ``normal`` (or default) VNIC type indicates that the Instance is plugged"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_c700d707","line":123,"range":{"start_line":123,"start_character":26,"end_line":123,"end_character":53},"updated":"2018-12-13 23:08:08.000000000","message":"https://github.com/openstack/nova/blob/e3eb5f916580a9bab8f67b0fd685c6b3b23a97b7/nova/network/model.py#L111 right?","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"9643cd3300e9fd43ab1d8d2105992b9be2e19e7c","unresolved":false,"context_lines":[{"line_number":120,"context_line":"The VNIC type is a property of a VIF. It has taken on the semantics of"},{"line_number":121,"context_line":"describing a specific \"plugging mode\" for the VIF. In the Nova network API,"},{"line_number":122,"context_line":"there is a list of VNIC types that will trigger a PCI request, if Neutron has"},{"line_number":123,"context_line":"passed a VIF to Nova with one of those VNIC types set. `Open vSwitch offloads`_"},{"line_number":124,"context_line":"uses the following VNIC types to distinguish between offloaded modes:"},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"* The ``normal`` (or default) VNIC type indicates that the Instance is plugged"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_42302e53","line":123,"range":{"start_line":123,"start_character":26,"end_line":123,"end_character":53},"in_reply_to":"3f79a3b5_c700d707","updated":"2018-12-14 11:00:14.000000000","message":"Correct: and specifically this is where a PCI request is triggered:\nhttps://github.com/openstack/nova/blob/e3eb5f916580a9bab8f67b0fd685c6b3b23a97b7/nova/network/neutronv2/api.py#L1921\n\nI\u0027ll add both references in a respin.","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"65834e663bd5512840d1756b344d9a9d16a197f0","unresolved":false,"context_lines":[{"line_number":130,"context_line":""},{"line_number":131,"context_line":"In addition, the Agilio OVS VIF type implements the following offload mode:"},{"line_number":132,"context_line":""},{"line_number":133,"context_line":"* The ``virtio-forwarder`` VNIC type indicates that a VF is attached via a"},{"line_number":134,"context_line":"  `virtio-forwarder`_."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"Currently, os-vif and Nova implement `switchdev SR-IOV offloads`_ for Open"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_eafefa01","line":133,"range":{"start_line":133,"start_character":8,"end_line":133,"end_character":24},"updated":"2018-12-13 23:08:08.000000000","message":"oooh yes added via I3da059fb62f085ff04532f8ec05857acfb628b06","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"9643cd3300e9fd43ab1d8d2105992b9be2e19e7c","unresolved":false,"context_lines":[{"line_number":130,"context_line":""},{"line_number":131,"context_line":"In addition, the Agilio OVS VIF type implements the following offload mode:"},{"line_number":132,"context_line":""},{"line_number":133,"context_line":"* The ``virtio-forwarder`` VNIC type indicates that a VF is attached via a"},{"line_number":134,"context_line":"  `virtio-forwarder`_."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"Currently, os-vif and Nova implement `switchdev SR-IOV offloads`_ for Open"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_62044a6d","line":133,"range":{"start_line":133,"start_character":8,"end_line":133,"end_character":24},"in_reply_to":"3f79a3b5_eafefa01","updated":"2018-12-14 11:00:14.000000000","message":"Memories of a happier, more innocent time...","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"65834e663bd5512840d1756b344d9a9d16a197f0","unresolved":false,"context_lines":[{"line_number":159,"context_line":""},{"line_number":160,"context_line":"Currently, switchdev VF offloads are modelled for one port profile only. Should"},{"line_number":161,"context_line":"a developer, using a different datapath, wish to pass offload metadata to an"},{"line_number":162,"context_line":"os-vif plugin, they would have to extend the object model, or pass the metadata"},{"line_number":163,"context_line":"using a confusingly named object. This spec aims to establish a recommended"},{"line_number":164,"context_line":"mechanism to extend the object model."},{"line_number":165,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_8a0f462b","line":162,"range":{"start_line":162,"start_character":0,"end_line":162,"end_character":13},"updated":"2018-12-13 23:08:08.000000000","message":"Looky here! https://github.com/Netronome/agilio-ovs-openstack-plugin","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"9643cd3300e9fd43ab1d8d2105992b9be2e19e7c","unresolved":false,"context_lines":[{"line_number":159,"context_line":""},{"line_number":160,"context_line":"Currently, switchdev VF offloads are modelled for one port profile only. Should"},{"line_number":161,"context_line":"a developer, using a different datapath, wish to pass offload metadata to an"},{"line_number":162,"context_line":"os-vif plugin, they would have to extend the object model, or pass the metadata"},{"line_number":163,"context_line":"using a confusingly named object. This spec aims to establish a recommended"},{"line_number":164,"context_line":"mechanism to extend the object model."},{"line_number":165,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_4247cea6","line":162,"range":{"start_line":162,"start_character":0,"end_line":162,"end_character":13},"in_reply_to":"3f79a3b5_8a0f462b","updated":"2018-12-14 11:00:14.000000000","message":"Yep, but that plugin will stay pretty much a shim on the ovs plugin, so as the ovs plugin changes it will change.\n\nThe \"future work\" is to finally move vrouter into os-vif... https://github.com/Juniper/contrail-nova-vif-driver/","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"65834e663bd5512840d1756b344d9a9d16a197f0","unresolved":false,"context_lines":[{"line_number":198,"context_line":""},{"line_number":199,"context_line":"Instead of bumping object versions and creating composition version maps, this"},{"line_number":200,"context_line":"option proposes that versioning be deliberately ignored until the next major"},{"line_number":201,"context_line":"release of os-vif. At this point, version negotiation and backlevelling in"},{"line_number":202,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":203,"context_line":""},{"line_number":204,"context_line":"Should protocol offloads also need to be modeled in os-vif, ``VIFBase`` or"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_aa31e2f1","line":201,"range":{"start_line":201,"start_character":19,"end_line":201,"end_character":32},"updated":"2018-12-13 23:08:08.000000000","message":"nit: \"Currently\" - I was reading this like \"at this point in the implementation of this spec\" or something like that - a work item as part of this blueprint, not \"today we don\u0027t do version negotiation with these things at all so trying to be careful about them is not really a priority\".","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"3d393fae99b1c40bfb36653a8485bfb09bad4190","unresolved":false,"context_lines":[{"line_number":197,"context_line":"overlapping release cycle, before the Open vSwitch plugin is updated in os-vif."},{"line_number":198,"context_line":""},{"line_number":199,"context_line":"Instead of bumping object versions and creating composition version maps, this"},{"line_number":200,"context_line":"option proposes that versioning be deliberately ignored until the next major"},{"line_number":201,"context_line":"release of os-vif. At this point, version negotiation and backlevelling in"},{"line_number":202,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":203,"context_line":""},{"line_number":204,"context_line":"Should protocol offloads also need to be modeled in os-vif, ``VIFBase`` or"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_232091ae","line":201,"range":{"start_line":200,"start_character":35,"end_line":201,"end_character":18},"updated":"2018-12-14 15:03:46.000000000","message":"we may not be able to ignore the versions by the way.\n\nkuryr-kubernetes is using os-vif in a way that depends on our object versioning eventhough we currently conisder that part of our internal implemnetation and not the public interfaces.\n\nthat said i do think we should minimize the backlevleing code until we release 2.0 but we will come up with some solution for kuryr \n\ni started a conversation on this topic here but i think most of the changes will need to wait untill train.\nhttp://lists.openstack.org/pipermail/openstack-discuss/2018-December/000569.html","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"9643cd3300e9fd43ab1d8d2105992b9be2e19e7c","unresolved":false,"context_lines":[{"line_number":198,"context_line":""},{"line_number":199,"context_line":"Instead of bumping object versions and creating composition version maps, this"},{"line_number":200,"context_line":"option proposes that versioning be deliberately ignored until the next major"},{"line_number":201,"context_line":"release of os-vif. At this point, version negotiation and backlevelling in"},{"line_number":202,"context_line":"os-vif is not used in Nova or os-vif plugins."},{"line_number":203,"context_line":""},{"line_number":204,"context_line":"Should protocol offloads also need to be modeled in os-vif, ``VIFBase`` or"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_42ccae30","line":201,"range":{"start_line":201,"start_character":19,"end_line":201,"end_character":32},"in_reply_to":"3f79a3b5_aa31e2f1","updated":"2018-12-14 11:00:14.000000000","message":"Will fix!","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"65834e663bd5512840d1756b344d9a9d16a197f0","unresolved":false,"context_lines":[{"line_number":227,"context_line":"    * os-vif object: ``VIFVHostUser``"},{"line_number":228,"context_line":"    * ``port_profile: VIFPortProfileOVSRepresentor``"},{"line_number":229,"context_line":""},{"line_number":230,"context_line":"* After Option 1 has been adopted in Nova:"},{"line_number":231,"context_line":""},{"line_number":232,"context_line":"  * VIF type: ``ovs`` (os-vif plugin: ``ovs``)"},{"line_number":233,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_2a439248","line":230,"range":{"start_line":230,"start_character":8,"end_line":230,"end_character":16},"updated":"2018-12-13 23:08:08.000000000","message":"Is there another option? Oh I guess inheritance is option 2? Maybe just say \"the composition model\".","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"9643cd3300e9fd43ab1d8d2105992b9be2e19e7c","unresolved":false,"context_lines":[{"line_number":227,"context_line":"    * os-vif object: ``VIFVHostUser``"},{"line_number":228,"context_line":"    * ``port_profile: VIFPortProfileOVSRepresentor``"},{"line_number":229,"context_line":""},{"line_number":230,"context_line":"* After Option 1 has been adopted in Nova:"},{"line_number":231,"context_line":""},{"line_number":232,"context_line":"  * VIF type: ``ovs`` (os-vif plugin: ``ovs``)"},{"line_number":233,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_62e0ca94","line":230,"range":{"start_line":230,"start_character":8,"end_line":230,"end_character":16},"in_reply_to":"3f79a3b5_2a439248","updated":"2018-12-14 11:00:14.000000000","message":"Yeah, apologies, this one snuck in from the previous revision. On respin it will be \u0027s/Option 1/this model/\u0027","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"65834e663bd5512840d1756b344d9a9d16a197f0","unresolved":false,"context_lines":[{"line_number":257,"context_line":"os-vif needs to issue a release before these profiles will be available to"},{"line_number":258,"context_line":"general CI testing in Nova. Once this is done, Nova can be adapted to use the"},{"line_number":259,"context_line":"new generic interfaces. The ``ovs`` and ``agilio_ovs`` plugins can make use"},{"line_number":260,"context_line":"of the new interfaces one release cycle later. For any future extensions,"},{"line_number":261,"context_line":"changes to Nova will still be required to convert the Neutron VIF object to an"},{"line_number":262,"context_line":"os-vif object."},{"line_number":263,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_ca33feee","line":260,"range":{"start_line":260,"start_character":22,"end_line":260,"end_character":45},"updated":"2018-12-13 23:08:08.000000000","message":"Do you mean in Train? If so, why? For n-1 compatibility?","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"9643cd3300e9fd43ab1d8d2105992b9be2e19e7c","unresolved":false,"context_lines":[{"line_number":257,"context_line":"os-vif needs to issue a release before these profiles will be available to"},{"line_number":258,"context_line":"general CI testing in Nova. Once this is done, Nova can be adapted to use the"},{"line_number":259,"context_line":"new generic interfaces. The ``ovs`` and ``agilio_ovs`` plugins can make use"},{"line_number":260,"context_line":"of the new interfaces one release cycle later. For any future extensions,"},{"line_number":261,"context_line":"changes to Nova will still be required to convert the Neutron VIF object to an"},{"line_number":262,"context_line":"os-vif object."},{"line_number":263,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_706650f3","line":260,"range":{"start_line":260,"start_character":22,"end_line":260,"end_character":45},"in_reply_to":"3f79a3b5_ca33feee","updated":"2018-12-14 11:00:14.000000000","message":"I agree this is confusing and badly written, I\u0027ll try to chat on IRC. It would be good to clarify the release sequence precisely.","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"65834e663bd5512840d1756b344d9a9d16a197f0","unresolved":false,"context_lines":[{"line_number":381,"context_line":"Upgrade impact"},{"line_number":382,"context_line":"--------------"},{"line_number":383,"context_line":""},{"line_number":384,"context_line":"A deployer would be encouraged to upgrade the os-vif library and plugins to"},{"line_number":385,"context_line":"avoid deprecation warnings in Nova."},{"line_number":386,"context_line":""},{"line_number":387,"context_line":"Implementation"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_4a712eaa","line":384,"updated":"2018-12-13 23:08:08.000000000","message":"I assume we\u0027ll bump the minimum required version of os-vif in nova in both requirements.txt and lower-constraints.txt anyway, and deployers should be following at least those minimums.","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"9643cd3300e9fd43ab1d8d2105992b9be2e19e7c","unresolved":false,"context_lines":[{"line_number":381,"context_line":"Upgrade impact"},{"line_number":382,"context_line":"--------------"},{"line_number":383,"context_line":""},{"line_number":384,"context_line":"A deployer would be encouraged to upgrade the os-vif library and plugins to"},{"line_number":385,"context_line":"avoid deprecation warnings in Nova."},{"line_number":386,"context_line":""},{"line_number":387,"context_line":"Implementation"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_f0aee0e3","line":384,"in_reply_to":"3f79a3b5_4a712eaa","updated":"2018-12-14 11:00:14.000000000","message":"I believe that\u0027s accurate. Thanks, I\u0027ll restate it that way to be clearer.","commit_id":"65e41ecbdb35ee135e84bf994e961284744e2bd6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"7a32cd10bd6f3dc7ed453d9e2f3ef46e190e5bb0","unresolved":false,"context_lines":[{"line_number":263,"context_line":"general CI testing in Nova. Once this is done, Nova can be adapted to use the"},{"line_number":264,"context_line":"new generic interfaces."},{"line_number":265,"context_line":""},{"line_number":266,"context_line":"* In Stein, os-vif\u0027s object model wil gain the interfaces described in this"},{"line_number":267,"context_line":"  spec. If needed, a major os-vif release will be issued."},{"line_number":268,"context_line":"* Then, Nova will depend on the new release and use the new interfaces for new"},{"line_number":269,"context_line":"  plugins."}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_6e1e6962","line":266,"range":{"start_line":266,"start_character":34,"end_line":266,"end_character":37},"updated":"2018-12-14 20:44:01.000000000","message":"will","commit_id":"14b4f73e255f0d79ae1e29a1ff2f4f3abdf86969"}]}
