What problem does it solve?
Developers offloading AES-GCM authenticated encryption to BlueField DPU or ConnectX accelerators face a complex API surface: capability queries, mmap permission flags, key-size constraints, and security-critical tag-verification handling. This Skill teaches AI agents the correct DOCA AES-GCM workflows so they avoid common failures like inventing AES-192 support, skipping capability checks, or consuming plaintext after a failed tag verification.
Core Features & Use Cases
- Capability Discovery: Query
doca_aes_gcm_cap_* functions to verify task-type support, per-key-type support (only DOCA_AES_GCM_KEY_128 and _256), and maximum plaintext buffer size before configuring anything.
- Six Task Workflows: Step-by-step guidance for configure, build, modify, run, test, and debug verbs, including known-vector round-trip validation against NIST GCMVS or RFC 5288 test vectors.
- Security-Critical Error Handling: Maps
DOCA_ERROR_* codes to AES-GCM root causes and treats decrypt tag-verification failure as a tampering signal whose plaintext output must never be consumed.
- Use Case: A developer encrypting 4 KiB TLS records at line rate asks whether to offload to the accelerator; the agent applies the path-selection rules, confirms device capabilities, sets source/destination mmap permissions, and validates with a published test vector before bulk submission.
Quick Start
Ask your agent to help configure doca_aes_gcm_task_encrypt on your BlueField, including checking which key sizes the device supports and setting the correct mmap permissions.