[WIP] nvme_driver: spawn threads for device shutdown

[mattkur]

No description provided.

[Copilot]

Pull Request Overview

This pull request modifies the NVMe device shutdown process to use separate threads for each device shutdown operation to prevent blocking the async runtime when kernel operations block.

• Replaces direct async shutdown calls with thread-spawned shutdown operations
• Maintains concurrent shutdown behavior while avoiding runtime blocking
• Preserves existing tracing and error handling patterns

Comments suppressed due to low confidence (1)

openhcl/underhill_core/src/nvme_manager.rs:292

• The variable name 'shutdown_driver' is misleading as it represents a task spawner, not a driver. Consider renaming to 'task_spawner' or 'spawner' to better reflect its purpose.

                        let shutdown_driver = self.driver_source.simple();

[Copilot]

The spawned tasks return JoinHandle results that need to be handled. The current code ignores potential task failures or panics. Consider using try_join_all or explicitly handling the Result from each JoinHandle to ensure shutdown errors are properly propagated.

Suggested change

	join_all(handles).await;
	if let Err(e) = futures::future::try_join_all(handles).await {
	tracing::error!("Error during NVMe driver shutdown: {:?}", e);
	}

[mattkur]

This doesn't work. I found that the cc.en 1 -> 0, when held by a vmexit, is still serialized.