feat: support ascend and nvidia accelerators with runc - #39
Open
mhsong1998-dot wants to merge 1 commit into
Open
feat: support ascend and nvidia accelerators with runc#39mhsong1998-dot wants to merge 1 commit into
mhsong1998-dot wants to merge 1 commit into
Conversation
mhsong1998-dot
force-pushed
the
feat/ascend-npu-runc
branch
from
September 2, 2026 02:36
d866a6e to
7244fa8
Compare
mhsong1998-dot
changed the title
mhsong1998-dot
force-pushed
the
feat/ascend-npu-runc
branch
from
September 3, 2026 02:24
ce4e78a to
e364c40
Compare
mhsong1998-dot
marked this pull request as ready for review
September 4, 2026 03:53
mhsong1998-dot
force-pushed
the
feat/ascend-npu-runc
branch
2 times, most recently
from
September 7, 2026 02:19
c7e4ae3 to
27fbf94
Compare
WenYuLuo
reviewed
Sep 9, 2026
WenYuLuo
reviewed
Sep 9, 2026
WenYuLuo
reviewed
Sep 9, 2026
WenYuLuo
reviewed
Sep 9, 2026
WenYuLuo
reviewed
Sep 9, 2026
WenYuLuo
suggested changes
Sep 9, 2026
Add provider-owned GPU and Ascend discovery, runtime selection, exclusive leases, and validated OCI edits for native runc sandboxes. Keep NVIDIA runsc support and preserve application library paths when injecting driver directories. Consume Ascend integration through the standalone adapter protocol so sandboxd no longer contains the MindCluster source dependency or adapter build target. Signed-off-by: mhsong2 <songminhui2@huawei.com>
mhsong1998-dot
force-pushed
the
feat/ascend-npu-runc
branch
from
September 9, 2026 09:29
676d987 to
52ea5cf
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add provider-owned NVIDIA and Ascend discovery, runtime selection, exclusive leases, and validated OCI edits for native runc sandboxes. NVIDIA runsc support remains available, and driver library directories are prepended without discarding application library paths.
Ascend integration now uses the schema v1 external process protocol from openYuanrong-mirror/ascend-oci-adapter#1. sandboxd keeps the trusted mount profile and validates every returned device, cgroup rule, mount, and environment entry, while the MindCluster submodule, adapter source, and adapter build target are removed from this repository.
The final PR history is one signed commit. Duplicate Ascend happy-path and library-path tests were consolidated into one 310P3 provider lifecycle test while security, runtime, lease, and device-stat coverage remains.
Validation:
go test ./...in a cn-north-4 CCE Linux Pod at the final rebased headnpu-smi, ACL runtime initialization and memory transfers, exclusive allocation, release, and device reuse[11, 22, 33, 44]with the preinstalled CANN 6.0 native Python 3.7.5 toolchain; sandboxed Add is a separate environment follow-upReliability boundary: a live runc workload does not survive a sandboxd restart in the current baseline. The lease is cleared and the device remains reallocatable; this PR does not claim workload-preserving restart recovery.
Depends on openYuanrong-mirror/ascend-oci-adapter#1 for the separately reviewed release source. Paired packaging change: inclusionAI/AKernel#45.