Skip to content

NVIDIA_DEVELOPING.md

Latest commit

 

History

History
83 lines (66 loc) · 3.3 KB

NVIDIA_DEVELOPING.md

File metadata and controls

83 lines (66 loc) · 3.3 KB

NVIDIA bmcweb development

Motivation

Keep NVIDIA contributions to bmcweb maintainable, sharable to partners, and reusable to future products. What follows are intended to be guidelines, not rules, and good judgement should always prevail over following a guildelines to the letter. The goal of this is to ship products. this repository is one tool of many to accomplish that goal.

If there are questions on the specific applications of these guildelines, please reach out to the bmcweb Nvidia maintainers and they will clarify intent.

Creating Additional schemas

  1. Follow the Nvidia Redfish schema pack guidelines1 to create a new schema, or modification to the schema, including mockup changes. Ensure that the schema is merged ahead of primary development to reduce rework.
  2. Run scripts/updatenvidia_schemas.py. This will pull the merged changes from #1 and incorporate them into bmcweb. Push this schema change into an MR ahead of code. Note, this script _may pull in schema changes unrelated to yours, this is expected.
  3. Create a new file, following the pattern redfish-core/lib/nvidia_<schema_name>.hpp. Ensure that schema_name is synced with the schema being extended.
  4. Add a single line callback to extend the existing route handlers to call the NVIDIA callbacks.
  5. Ensure that DMTF-compliant code is kept separate from NVIDIA OEM code additions through the use of more files. This separation helps distinguish potential DMTF schema enhancements that can be upstreamed versus OEM/ODM code.

Examples of extension handlers.

redfish-core/lib/managers.hpp

void handleManagersGet(App& app, const crow::Request& req,
                        const std::shared_ptr<bmcweb::AsyncResp>& asyncResp,
                        const std::string& managerId) {

   // Ensure that nvidia extensions are a single line change to existing code
   nvidiaManagerGet(req, asyncResp, managerId);
}

redfish-core/lib/nvidia_managers.hpp

   void nvidiaManagerGet(....){
      // Ensure that this is before any Nvidia OEM extension.
      if (!BMCWEB_NVIDIA_OEM_PROPERTIES){
         return;
      }
      // Nvidia specific implementation
   }
  1. Unless immediately being sent upstream, keep new route handlers in separate files, with the convention redfish-core/lib/nvidia_<schema>.hpp. Exceptions may be made on a case by case basis.
  2. Create gerrit MR for the code enhanced. Follow up with community to get the MR merged upstream.
  3. Once Upstream MR is merged, ensure downstream code is aligned by migrating code back to original schema from nvidia_<schema>.hpp file.
  4. Downstream changes up to 3 lines can be kept in the same file. This is intentionally very small to ensure rebases can be performed in the future.

Nvidia specific Redfish implementations should be placed in the redfish::nvidia c++ namespace.

Testing

  1. Nvidia functions shall include unit tests with target 80% coverage.
  2. Ensure DMTF service validator passes on target platform.
  3. Consider running performance tests 2.

Footnote