[7.0] [Feature] nvidia kernel CXL Type-2 VFIO support for devel-7.0: NVGrace EGM and header dependency fix - part4 · Pull Request !9622 · anolis/cloud-kernel

bugzilla: https://bugzilla.openanolis.cn/show_bug.cgi?id=38445

This is the part 4 of backporting nvidia-kernel cxl infrastructure.
Part1: https://gitee.com/anolis/cloud-kernel/pulls/9617
Part2: https://gitee.com/anolis/cloud-kernel/pulls/9619
Part3: https://gitee.com/anolis/cloud-kernel/pulls/9620

This PR adds NVGrace EGM support and related platform integration, including:

- A new `vfio/nvgrace-egm` module for managing EGM regions.
- EGM registration and unregistration infrastructure.
- ECC error page handling for EGM-backed memory.
- Retired page handling and invalid retired-page range protection.
- PFNMAP range registration with `memory_failure`.
- Sysfs exposure for EGM size, GPU list, and GPU-to-EGM region association.
- Improved EGM registration error handling and propagation.
- Cleanup and robustness fixes for allocation failures, sparse warnings, and smatch warnings.
- Blackwell-Next GPU readiness check through CXL DVSEC.
- VFIO-CXL public header dependency cleanup so VFIO-CXL users rely on public CXL headers instead of private/internal header layout.

commits:

NVIDIA: SAUCE: vfio/nvgrace-egm: Introduce module to manage EGM
NVIDIA: SAUCE: vfio/nvgrace-egm: Handle pages with ECC errors on the EGM
NVIDIA: SAUCE: vfio/nvgrace-egm: Move the egm header file to include
NVIDIA: SAUCE: vfio/nvgrace-egm: Free region memory during unregistration
NVIDIA: SAUCE: vfio/nvgrace-egm: Move region hash initialization
NVIDIA: SAUCE: vfio/nvgrace-egm: Handle and convey EGM registration errors
NVIDIA: SAUCE: vfio/nvgrace-gpu: Handle EGM registration failure
NVIDIA: SAUCE: vfio/nvgrace-egm: Address sparse errors
NVIDIA: SAUCE: vfio/nvgrace-gpu: Address smatch errors
NVIDIA: SAUCE: vfio/nvgrace-egm: Ensure ACPI value reads are successful
NVIDIA: SAUCE: vfio/nvgrace-egm: Avoid invalid retired pages base
NVIDIA: SAUCE: vfio/nvgrace-egm: Update EGM unregistration API
NVIDIA: SAUCE: vfio/nvgrace-egm: track GPUs associated with the EGM regions
NVIDIA: SAUCE: vfio/nvgrace-egm: list gpus through sysfs
NVIDIA: SAUCE: vfio/nvgrace-egm: expose the egm size through sysfs
NVIDIA: SAUCE: vfio/nvgrace-egm: register EGM PFNMAP range with memory_failure
NVIDIA: SAUCE: vfio/nvgrace-egm: Address smatch errors
NVIDIA: SAUCE: vfio/nvgrace-egm: Add null pointer checks after memory allocations
NVIDIA: SAUCE: vfio/nvgrace-egm: split zapping EGM into 1GB chunks
NVIDIA: SAUCE: vfio/nvgrace-gpu: Add Blackwell-Next GPU readiness check via CXL DVSEC
NVIDIA: VR: SAUCE: cxl: Fix VFIO-CXL public header dependencies

类型	指派人员	状态
审查	xunlei 半叶枫尘 winterddd xueshuai 冯光辉 Guixin Liu XiangZao	已完成 (0/0人)
测试	小龙	进行中 (0/1人)

GVPanolis/cloud-kernel

搜索帮助

GVP anolis/cloud-kernel