ML Service has individual sandbox configs for each ML model and so the seccomp allowlist setup is a bit more complex.
The .policy files in this directory fall into three categories as follows:
ml_service-seccomp-$ARCH.policy: Overall syscall allowlist for the normal Mojo ML Service at the time the service is launched. Used by ../init/ml-service.conf. By necessity, this includes all syscalls used by any of the narrowed allowlists mentioned below.ml_service-$SPECIFIC_MODEL-seccomp-$ARCH.policy: Narrowed syscall allowlist for ML Service subprocess sandboxes for running a specific model. ML Service enters this seccomp policy after it spawns a subprocess. See SetSeccompPolicyPath in ../process.cc.ml_service-AdaptiveChargingModel-seccomp-$ARCH.policy: Special case top-level allowlist used by ../init/ml-service.conf when ML Service is launched in its non-Mojo service mode for use outside Chrome.