[AprilNEA]

The guest's NFSv4 grace period ran for the full 10 s on every boot: end_grace_early() got EBUSY because nfsd had no client tracker, and the macOS client retries a refused OPEN every 2 s. The first open under ~/ArcBox after boot waited 6–12 s (e2e nfs_export 8/10/12 s).

• The agent mounts rpc_pipefs and starts nfsdcld on tmpfs before nfsd threads start, so nfsd logs "no clients to reclaim, skipping NFSv4 grace period". Without /sbin/nfsdcld in the rootfs it warns and keeps the old path.
• Boot assets bump to 0.8.12, which ships nfsdcld (boot-assets 7eff507).

Measured on HV and VZ against the published 0.8.12: first open returns at once; e2e boot_assets nfs_export 0.0015 s (HV) and 0.0025 s (VZ).

---

Note

Medium Risk
Changes guest NFS bring-up ordering and boot asset pin; mis-timed nfsdcld readiness could still leave grace stalls, but behavior degrades to the prior path with warnings.

Overview
Fixes 6–12 s stalls on the first NFS open under ~/ArcBox after boot, caused by the full NFSv4 grace period when the kernel refused early v4_end_grace without a client tracker.

The guest agent now mounts rpc_pipefs, starts nfsdcld (tmpfs state) before nfsd threads, and waits until nfsdcld holds an inotify fd so startup ordering does not race with nfsd_mutex. Rootfs without /sbin/nfsdcld still works but logs a warning and keeps the old grace behavior. spawn_supervised factors shared daemon spawning for rpc.mountd and nfsdcld; end_grace_early skips when grace already ended and warns on failure instead of debug-only.

Boot assets pin moves to 0.8.12, which ships nfsdcld. Docs add a log entry with measurements; e2e boot_assets comments reflect nfsdcld vs legacy grace timing.

^{Reviewed by Cursor Bugbot for commit 0903eb5. Bugbot is set up for automated code reviews on this repo. Configure here.}

[greptile-apps]

Confidence Score: 5/5

[Medium impact] This PR appears safe to merge; no actionable issues were found.

Summary

Starts nfsdcld before nfsd so a fresh guest can skip the NFSv4 grace period.

• Updated boot assets let NFS opens skip the grace-period wait.

AprilNEA explicitly describes the full grace period on older rootfs images as an intentional fallback.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Ensure NFS export] --> B[Mount rpc_pipefs]
  B --> C[Start nfsdcld]
  C --> D[Wait for inotify descriptor]
  D --> E[Start nfsd threads]
  E --> F[Fresh tracker database lets nfsd skip grace]
  A -->|Tracker unavailable| G[Warn and continue]
  G --> E

Loading

_{Reviews (1) · Last reviewed commit: "chore(boot): bump boot assets to 0.8.12" · Reviewed by Greptile}