id: T011
name: PlasmaVMC Feature Deepening
status: complete
goal: Make KvmBackend functional - actual VM lifecycle, not stubs
priority: P0
owner: peerA (strategy) + peerB (implementation)
created: 2025-12-08

context: |
  Scaffold complete (5 crates) but KvmBackend methods are stubs returning errors.
  Spec defines 10 crates, but depth > breadth at this stage.
  Focus: Make one hypervisor backend (KVM) actually work.

acceptance:
  - KvmBackend.create() spawns QEMU process
  - KvmBackend.status() returns actual VM state
  - KvmBackend.start()/stop() work via QMP
  - At least one integration test with real QEMU
  - plasmavmc-server can manage a VM lifecycle end-to-end

## Gap Analysis (current vs spec)
# Existing: plasmavmc-types, hypervisor, kvm, api, server
# Missing: client, core, firecracker, mvisor, agent, storage (defer)
# Strategy: Deepen existing before expanding

steps:
  - step: S1
    action: Add QMP client library to plasmavmc-kvm
    priority: P0
    status: complete
    owner: peerB
    notes: |
      QMP = QEMU Machine Protocol (JSON over Unix socket)
      Use qapi-rs or custom implementation
      Essential for VM control commands
    deliverables:
      - QmpClient struct with connect(), command(), query_status()
      - Unit tests with mock socket

  - step: S2
    action: Implement KvmBackend.create() with QEMU spawning
    priority: P0
    status: complete
    owner: peerB
    notes: |
      Generate QEMU command line from VmSpec
      Create runtime directory (/var/run/plasmavmc/kvm/{vm_id}/)
      Spawn QEMU process with QMP socket
      Return VmHandle with PID and socket path
    deliverables:
      - Working create() returning VmHandle
      - QEMU command line builder
      - Runtime directory management

  - step: S3
    action: Implement KvmBackend.status() via QMP query
    priority: P0
    status: complete
    owner: peerB
    notes: |
      query-status QMP command
      Map QEMU states to VmStatus enum
    deliverables:
      - Working status() returning VmStatus
      - State mapping (running, paused, shutdown)

  - step: S4
    action: Implement KvmBackend.start()/stop()/kill()
    priority: P0
    status: complete
    owner: peerB
    notes: |
      start: cont QMP command
      stop: system_powerdown QMP + timeout + sigkill
      kill: quit QMP command or SIGKILL
    deliverables:
      - Working start/stop/kill lifecycle
      - Graceful shutdown with timeout

  - step: S5
    action: Integration test with real QEMU
    priority: P1
    status: complete
    owner: peerB
    notes: |
      Requires QEMU installed (test skip if not available)
      Use cirros or minimal Linux image
      Full lifecycle: create → start → status → stop → delete
    deliverables:
      - Integration test (may be #[ignore] for CI)
      - Test image management

  - step: S6
    action: Wire gRPC service to functional backend
    priority: P1
    status: complete
    owner: peerB
    notes: |
      plasmavmc-api VmService implementation
      CreateVm, StartVm, StopVm, GetVm handlers
      Error mapping to gRPC status codes
    deliverables:
      - Working gRPC endpoints
      - End-to-end test via grpcurl

blockers: []

aux_tactical: []

evidence: []

notes: |
  Foreman recommended PlasmaVMC deepening as T011 focus.
  Core differentiator: Multi-hypervisor abstraction actually working.
  S1-S4 are P0 (core functionality), S5-S6 are P1 (integration).