[CI][ROCm] Ship RIXL with vllm/vllm-openai-rocm (#41634)

Signed-off-by: simondanielsson <[email protected]>

docker/Dockerfile.rocm

@@ -507,6 +507,10 @@ RUN --mount=type=bind,from=export_vllm,src=/,target=/install \
     && pip uninstall -y vllm \
     && uv pip install --system *.whl
 
+# Install RIXL wheel
+RUN --mount=type=bind,from=build_rixl,src=/app/install,target=/rixl_install \
+    uv pip install --system /rixl_install/*.whl
+
 ARG COMMON_WORKDIR
 ARG BASE_IMAGE
 ARG NIC_BACKEND
@@ -517,6 +521,10 @@ COPY --from=export_vllm /benchmarks ${COMMON_WORKDIR}/vllm/benchmarks
 COPY --from=export_vllm /examples ${COMMON_WORKDIR}/vllm/examples
 COPY --from=export_vllm /docker ${COMMON_WORKDIR}/vllm/docker
 
+# Use legacy IPC mode for HSA to avoid GPU memory pinning issues with UCX rocm_ipc
+# See: https://github.com/ROCm/rocm-libraries/issues/6266
+ENV HSA_ENABLE_IPC_MODE_LEGACY=1
+
 ENV TOKENIZERS_PARALLELISM=false
 
 # ENV that can improve safe tensor loading, and end-to-end time

docs/features/nixl_connector_usage.md

@@ -13,7 +13,7 @@ Install the NIXL library: `uv pip install nixl`, as a quick start on Nvidia plat
 - Refer to [NIXL official repository](https://github.com/ai-dynamo/nixl) for more installation instructions
 - The specified required NIXL version can be found in [requirements/kv_connectors.txt](../../requirements/kv_connectors.txt) and other relevant config files
 
-For ROCm platform, the [base ROCm docker file](../../docker/Dockerfile.rocm_base) includes RIXL and ucx already.
+For ROCm platform, the [ROCm docker file](../../docker/Dockerfile.rocm) includes RIXL and ucx already.
 
 - Refer to [RIXL official repository](https://github.com/rocm/rixl) for more information
 - The supportive libraries for RIXL can be found in [requirements/kv_connectors_rocm.txt](../../requirements/kv_connectors_rocm.txt)