docs(iol): record the IOL runner container kernel datapath

The feature doc covers the architecture (one irreducible userspace hop
on the unix-socket leg; the link segment kernelized through the port
bridge's swappable TAP leg), the capability gate, the anchor lifecycle,
the swap flow, the inherited capture/markers/filters, the relay
fallback and the live verification — including the ensure-then-add
lesson (tap create is strictly create-only).

The other datapath docs' roadmaps catch up: the IOL runner container
has landed (and the ethernet-switch doc's stale 'cascade deferred'
exclusion from d2f8aba58 is corrected while there); the docker doc's
adapter table no longer rejects IOLDockerVM.
This commit is contained in:
YueGuobin committed 2026-10-06 00:21:09 +08:00
1 parent 72aa379e87
commit 02340f2ed3
6 files changed
+199 -17

No files matched your search

@@ -28,15 +28,17 @@ end / TAP becomes the switch port; frames cross
## Datapath selection
A link with an `ethernet_switch` endpoint rides the fast path when the
**peer** can anchor (same compute; Docker always, QEMU/IOU/Dynamips per
**peer** can anchor (same compute; Docker always — including IOL runner
containers, per their own capability gate — QEMU/IOU/Dynamips per
their capabilities) — the switch side needs no anchoring capability of its
own, its bridge and brctl exist by construction. Exclusions, all staying
on the relay:
* **switch ↔ switch**: a kernel interface can belong to exactly one
bridge, so cascading two switch bridges needs a veth pair between them —
a separate mechanism, deferred;
* anchor-less peers (VPCS, cloud, the Dynamips-hosted hub);
* **switch ↔ switch** cascades ride the kernel too — through a veth pair
joining the two bridges (one end per switch, each side applying its own
port mode to its own end); see the wire format below;
* anchor-less peers (VPCS, cloud, the Dynamips-hosted hub, generic
`GNS3_UNIX_SOCKET_NIO` containers);
* cross-compute links, non-Ethernet ports: the common rules.
## The wire format: one NIO names the other end's interface
@@ -153,6 +155,8 @@ enable_kernel_datapath = True
## Roadmap
Switch-to-switch links (a veth pair joining the two kernel bridges, with
each side's port mode applied at its own end) and the IOL Docker node
(`bridge add_nio_tap`); cross-compute kernel links need VXLAN/GENEVE
encapsulation, deferred until every node type is kernelized.
each side's port mode applied at its own end) have landed (see above), and
so has the IOL runner container (`iol-docker-kernel-datapath.md` — its
`iol_docker` anchors absorb like any peer's). Cross-compute kernel links
need VXLAN/GENEVE encapsulation, deferred until every node type is
kernelized.