What's Changed
- Ship the disk templates zstd-compressed and expand them to sparse files on first use by @BinSquare in #771
- Copy the storage template by seeking between its data extents instead of scanning its whole logical size by @BinSquare in #770
- Harden fused rollout lifecycle by @BinSquare in #801
- Avoid copying CUDA module state for clone channels by @BinSquare in #804
- Load device-resident LoRA policies through managed executors by @BinSquare in #806
- Reject a registry-image machine that has no network at create instead of failing every start with a raw DNS error by @BinSquare in #807
- Perform machine file reads and writes inside the running workload container so uploads are visible to exec by @BinSquare in #810
- Reuse CUDA module handoffs across clone workers by @BinSquare in #808
- Cache pulled OCI images on the host so repeat ephemeral machine runs skip the registry pull by @BinSquare in #805
- Signal guest boot-readiness with an event-driven vsock doorbell as the primary ready signal, keeping the marker file and control-channel ping as fallbacks by @BinSquare in #811
- Map CUDA module handoffs across clone workers by @BinSquare in #814
- Preserve per-device GPU admission headroom by @BinSquare in #826
- Promote CUDA pool readiness and refill improvements by @BinSquare in #828
- Bump the workspace to 1.7.3 by @BinSquare in #829
- Install the zstd-compressed disk templates in the Arch package so the build no longer fails on the removed uncompressed storage template by @BinSquare in #830
- Set the library search path on the boot subprocess before launch so libkrun can load libkrunfw when embedded without a wrapper script by @BinSquare in #832
- Make fork-pool lease activation retry-safe by @BinSquare in #831
- Bump the workspace to 1.7.4 by @BinSquare in #833
- Avoid blocking restored container state probes by @BinSquare in #834
- Harden privileged CUDA clone startup by @BinSquare in #839
- Harden CUDA admission calibration by @BinSquare in #840
- Stabilize restored forkpoint activation by @BinSquare in #835
- Coordinate clone boots across fork pools by @BinSquare in #836
- Preserve CUDA pool refill checkpoints by @BinSquare in #841
- Wait for explicit CUDA daemon reconstruction by @BinSquare in #844
- Adapt CUDA admission to available GPU headroom by @BinSquare in #846
- Persist CUDA pool refill checkpoints by @BinSquare in #845
- Bump libkrunfw to the guest kernel with the Landlock LSM enabled by @BinSquare in #847
- Serialize concurrent VM updates by @BinSquare in #850
- Retry transient batch fork boots by @BinSquare in #853
- Expose identity for direct batch forks by @BinSquare in #856
- Bump libkrun for the mount-socket concurrent-hang fix by @BinSquare in #837
- Wake netns TAP bridge during shutdown by @BinSquare in #860
- Harden device adapter server startup by @BinSquare in #861
- Run an image machine's commands inside its image on the embedded path, so a locally created machine behaves like a cloud one by @BinSquare in #862
- Gate leases on worker readiness by @BinSquare in #858
- Preload prepared CUDA clone modules by @BinSquare in #854
- Note which guest paths are memory-backed and reset on restart by @BinSquare in #869
- Let an artifact-sourced machine run without network instead of demanding a registry pull by @BinSquare in #877
Full Changelog: v1.7.2...v1.7.5
Contributors
BinSquare