Skip to content

feat: complete KVM pre-fault stats - #21

Draft
Hygge-Gezelligheid wants to merge 6 commits into
kvcache-ai:v1.15.1-patchfrom
Hygge-Gezelligheid:agent/v1.15.1-patch-v2
Draft

feat: complete KVM pre-fault stats#21
Hygge-Gezelligheid wants to merge 6 commits into
kvcache-ai:v1.15.1-patchfrom
Hygge-Gezelligheid:agent/v1.15.1-patch-v2

Conversation

@Hygge-Gezelligheid

Copy link
Copy Markdown
Collaborator

Summary

Complete the paused-only KVM pre-fault path with synchronous, verifiable completion statistics.

  • Continue KVM_PRE_FAULT_MEMORY until every submitted range reports zero remaining bytes.
  • Return aggregate and per-vCPU-worker requested, completed, remaining, ioctl-count, and timing statistics.
  • Avoid waiting for a response from a vCPU that could not be signaled.
  • Expose the response through the Swagger API and parser models.

Validation

Passed on the AgentENV development host:

  • cargo test -p vmm vstate::prefault --lib (9 passed)
  • precise VMM integration coverage for state validation, multi-vCPU accounting, and hotpluggable-memory rejection (3 passed)
  • cargo test -p firecracker api_server::parsed_request::tests::test_put_pre_fault_memory_route --lib

cargo fmt --check could not be used as a validation signal on that host because its installed toolchain reformats unrelated baseline files.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant