Doca Verbs

DOCA Verbs

STOP — most RDMA tasks do NOT belong here

If the task is general RDMA data movement (send / receive / read / write / atomic between endpoints) and the user did not name a specific raw QP / CQ / work-request / SRQ / Address-Handle attribute that the higher-level API genuinely cannot express, this skill is out of scope. Route to doca-rdma and follow its non-negotiable: the deliverable links libdoca_rdma and calls doca_rdma_*.

Loading this raw-verbs skill is never a license to hand-roll libibverbs / librdmacm. "Raw verbs is fewer lines" or "the high-level binding is more work" is not a reason. A general-RDMA deliverable whose ldd shows no libdoca_rdma is a failed task — exactly the same rule doca-rdma enforces. Raw doca_verbs_* is in scope only for the narrow attribute-level needs enumerated below; everything else climbs back up to the matching higher-level library.

Where to start: This skill is the raw-verbs escape hatch beneath the higher-level DOCA libraries (doca-rdma for RDMA workloads, doca-eth for Ethernet queues, doca-rmax for timing-precise media). The agent's first job, before anything else, is to confirm the user actually needs to drop down — most users do not, and the right answer is almost always "stay in the higher-level library". Open CAPABILITIES.md when the question is what does the verbs surface actually expose and where is the boundary with vanilla libibverbs; open TASKS.md when the user has already confirmed they need raw verbs and wants the configure / build / modify / run / test / debug workflow for them. If the user has not installed DOCA yet, route to doca-setup first.

The decision this skill exists to gate

The single load-bearing decision every conversation that loads this skill must make, FIRST, before any code-level discussion:

1. Has the user confirmed that the matching higher-level DOCA library does not expose the semantic they need? If no — stop here, route back to the matching higher-level library: doca-rdma for general RDMA work (Send / Receive / Read / Write / Atomic / Sync-Event task patterns); doca-eth for Ethernet TX / RX queue patterns; doca-rmax for timing-precise media / data-over-IP streaming. The most common baseline-agent failure for raw verbs is recommending them unnecessarily because the user said the word "verbs" or "QP" without checking whether the higher-level surface already covers their case.
2. Is the semantic the user needs a specific verb / opcode / work-request flag / QP attribute / SRQ option that the matching higher-level library genuinely does not expose? Examples: a specific raw WR flag the doca_rdma_task_* abstractions do not surface; custom completion-queue handling beyond what the DOCA progress engine exposes; an esoteric QP attribute (path MTU, PSN tuning, ECE attributes); explicit SRQ control; congestion- control group (doca_verbs_cc_group_*) attachment to QPs; Address-Handle attribute tuning (DGID / DLID / SL / SGID index / hop limit / traffic class / UDP source port). If yes — this skill is in scope.
3. Is the user porting existing libibverbs code into the DOCA Core model? This skill is in scope, AND the agent must teach the porting path (replace libibverbs handles with doca_verbs_* handles, integrate with the DOCA Core lifecycle through doca_verbs_context_create, drive completions via the DOCA progress engine instead of polling CQ directly) rather than recommend a mechanical 1:1 textual replacement.

If none of (1)-(3) apply, the answer to "should I use doca-verbs?" is no. Route the user back to the matching higher-level DOCA library. This is by design: a correctly-loaded raw-verbs skill that talks the user out of raw verbs is doing its job.

Example questions this skill answers well

The CLASSES of raw-verbs questions this skill is built to answer, each with one worked example. The agent should treat the class as the load-bearing piece — the worked example is a single instance.

• "Should I drop to doca-verbs for this?" — worked example: "I want to set a specific raw work-request flag and the doca_rdma_task_* abstraction does not expose it". Answered by the path-selection rule in CAPABILITIES.md ## Capabilities and modes higher-level-vs-doca-verbs table + the climb back up step in TASKS.md ## configure.
• "How is doca-verbs different from libibverbs?" — worked example: "I have existing ibv_* code; can I just keep using it and call doca_* next to it?". Answered by the libibverbs-vs-doca-verbs boundary rule in CAPABILITIES.md ## Safety policy
  • the porting workflow in TASKS.md ## modify.
• "Is this raw verb / opcode / QP option / SRQ feature supported on my device + DOCA version?" — worked example: "does this device advertise the QP feature my raw WR needs". Answered by the capability-query rule (doca_verbs_query_device + the doca_verbs_device_attr_get_* family) in CAPABILITIES.md ## Capabilities and modes
  • the discovery step in TASKS.md ## configure.
• "I'm porting libibverbs code into the DOCA Core model — what does the agent walk me through?" — worked example: "I have a small libibverbs sender/receiver that uses IBV_SEND_INLINE on a custom QP attribute, and I want it to live inside a DOCA Core context". Answered by the porting overlay in TASKS.md ## modify + CAPABILITIES.md ## Safety policy no-mixing rule.
• "What does this DOCA_ERROR_* from a raw-verbs call mean?" — worked example: "DOCA_ERROR_IO_FAILED from a WR submission — what do I look at?". Answered by the verbs overlay on the cross-library taxonomy in CAPABILITIES.md ## Error taxonomy (which sends the agent to inspect the completion-queue entry, not the submit return value) + the layered ladder in TASKS.md ## debug.
• "When do I climb back up from doca-verbs to the matching higher-level library?" — worked example: "my raw-verbs prototype works; do I keep it or refactor onto doca-rdma?". Answered by the climb-back rule in CAPABILITIES.md ## Capabilities and modes (raw verbs is a targeted surface, not a default; once the specific need is covered, the higher-level surface is the long-term home).

Audience

This skill serves external developers building applications that consume the DOCA Verbs library — i.e., users whose code calls doca_verbs_* (directly in C/C++, or through FFI/bindings from another language) for raw QP / CQ / PD / MR / SRQ / Address-Handle / Ethernet-SQ / Ethernet-RQ control inside a DOCA Core context. It is not for NVIDIA developers contributing to DOCA Verbs itself, and it is not the right entry point for general DOCA RDMA / Eth / RMAX work — that belongs in the matching higher-level library skill.

Language scope

DOCA Verbs ships as a C library with pkg-config module name doca-verbs. The public headers live under the installed DOCA infrastructure tree ($(pkg-config --variable=includedir doca-common) doca_verbs.h and the adjacent doca_verbs_*.h family); per the headers-win-over-docs rule in doca-version, the headers on the user's install are the authoritative truth for the live symbol surface. C and C++ consumers are the canonical case; the worked examples in TASKS.md assume that path. Other-language consumers (Rust, Go, Python, …) consume the same *.so through FFI or language-specific bindings; the skill's contribution in that case is to keep the drop-down decision, libibverbs boundary, cap-query rule, lifecycle in verbs terms, and error-handling rule language-neutral, and to route the agent to the public C ABI as the authoritative surface that any wrapper will eventually call.

When to load this skill

Load this skill ONLY after the user (or the agent on the user's behalf) has confirmed the matching higher-level DOCA library does not expose the semantic they need. Concretely:

• The user explicitly asks "do I need to drop to doca-verbs for this?" — load this skill to answer, but expect the answer to be "no, stay in the higher-level library" unless the user can name the specific verb / opcode / option the higher-level library does not surface.
• The user wants a specific raw WR flag, raw QP option, SRQ feature, custom CQ-handling pattern, or congestion-control group attachment that the higher-level library does not expose.
• The user is porting existing libibverbs code into the DOCA Core model and needs the integration path (lifecycle, progress engine, no-mixing rule).
• A DOCA_ERROR_* returned from a doca_verbs_* call needs diagnosis — including the IO_FAILED case where the answer lives on the completion-queue entry, not the submit return.
• Designing or extending non-C bindings (Rust, Go, Python, …) that wrap the verbs C ABI — for the boundary, lifecycle, and cap-query rules the wrapper must honor.

Do not load this skill for: general DOCA RDMA work (use doca-rdma); general DOCA Ethernet queueing (use doca-eth); timing-precise media / data-over-IP streaming (use doca-rmax); use cases a different higher-level DOCA library covers (doca-flow for steering, the storage-transport library for NVMe-oF — routed via doca-public-knowledge-map); install of DOCA itself (use doca-setup); or general DOCA orientation (use doca-public-knowledge-map).