Native events failed the official OCSF 1.8.0 validator: HTTP Activity emitted is_src_dst_assignment_known, sandbox Process Activity events had no actor, connection_info lacked direction_id, and SSH Activity events named no endpoint. Schema tests missed these because they checked only top-level, non-profile attributes. Add a connection direction enum, require an actor on Process Activity and an endpoint on SSH Activity at compile time, and name the supervisor's SSH server with svc_name and its socket path. Validate against OCSF 1.8.0 only. validate_required_fields and validate_known_fields share one recursive walker that rejects undefined attributes at every level and, for events, attributes from undeclared profiles; objects are validated as objects. Records labelled with another version fail. Vendor tactic and technique so ATT&CK-mapped findings validate, drop the unused v1.7.0 schemas, and make the embedded-schema drift test tolerate CRLF checkouts. Keep it fixed with a per-builder conformance test and a Podman e2e test that runs an L7 policy against a local server and validates the supervisor's real Network, HTTP and SSH Activity JSONL. Top-level container stays unchanged; #4915 tracks it. The schema downgrade is unchanged. Closes #4283 Signed-off-by: Adel Zaalouk <azaalouk@redhat.com>
Network supervisor
This crate forbids unsafe code, including its local tests. Typed socket APIs provide original-destination lookup for transparent proxying.
Socket attribution tests pass a duplicate connected socket as a spawned child's stdin. Re-executing the test binary checks multiple holders with the same binary identity; executing a different binary checks ambiguous shared-socket identity. The fixture kills and reaps the child when the parent test ends.
The manual proxy performance baseline still counts allocations. Its test-only
global allocator uses openshell-driver-vm::allocation_tracking; the unsafe
allocator implementation belongs to the VM crate. The VM dependency disables
default features, so this instrumentation does not
require the compute runtime or Unix-only sandbox library. It is a development
dependency only.
Sandbox instance inputs
run_networking takes a SandboxInstanceNetwork with the sandbox identity,
the sandbox instance's GatewayEndpoint, the SPIFFE Workload API socket for
provider token grants, the policy.local denial source, and proxy CA files.
None of these are read from the process environment, so one process can run
several sandbox instances. Each instance owns its provider token cache and
HTTP client; a token acquired for one sandbox is never served to another.
Without a Workload API socket, token grants fail closed. A process hosting
several instances serves /v1/denials from a per-instance
SandboxInstanceDenialLog instead of the shared shorthand log files.
Reserved destinations
preauthorize_transparent_open matches a staged TCP open against the reserved
OTLP relay address (openshell_core::sandbox_env::OTLP_RELAY_ADDR) before host
mapping, policy evaluation, and SSRF checks. The address is an unroutable
label, so the proxy never dials it. When run_networking receives a
ReservedDestination, the proxy hands the workload stream to its
ReservedStreamHandler; a handler that refuses, for example at its connection
cap, turns the open into a resource-exhausted denial. Without a handler, the
proxy denies the open at once. It logs the first attempt at warn and later ones
at debug, and emits no OCSF denial event: a standard OpenTelemetry SDK retries
for the life of the agent, and one event per retry would flood the OCSF log.
The supervisor installs its OTLP relay receiver as the handler while the agent
trace relay is active.