fix: HiRadixCache: fix prefetch completion race

[pabloiyu]

Motivation

SGLang allows for hierarchical KV caching. Under the default storage prefetch policy best_effort, the scheduler doesn’t have to wait for prefetch to finish: it can proceed to stage the request, which terminates the prefetch early and inserts no pages from storage. To wait for prefetch, we can use wait_complete.

However, there is a race under wait_complete: the scheduler can call self.tree_cache.check_prefetch_progress() before the prefetch thread populates operation.hash_value. The completion check then treats 0 completed == 0 planned as done.

Fix: In HiRadixCache.can_terminate_prefetch(), only consider prefetch complete when operation.hash_value is non-empty and operation.completed_tokens == len(operation.hash_value) * page_size. This ensures the scheduler waits until the page list is prepared (and with wait_complete, until IO finishes).

Modifications

Accuracy Tests

Benchmarking and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.

[gemini-code-assist]

Summary of Changes

Hello @pabloiyu, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses a race condition within SGLang's hierarchical KV caching mechanism, specifically when the wait_complete prefetch policy is active. The issue arose because the scheduler could incorrectly determine that prefetch was complete if the operation.hash_value was empty, leading to early termination. The fix ensures that the system properly waits for the hash_value to be populated, thereby guaranteeing that the page list is fully prepared before prefetch is considered done.

Highlights

• Fix Prefetch Completion Race Condition: Modified the can_terminate_prefetch method in HiRadixCache to include a check that operation.hash_value is not empty before considering the prefetch complete. This addresses a race condition where the scheduler might prematurely terminate prefetch if hash_value was not yet populated.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request addresses a race condition in HiRadixCache's prefetch completion logic. The issue occurs when the wait_complete policy is used, where the scheduler might check for completion before the prefetch operation's hash_value is populated, leading to a premature completed status. The proposed fix correctly adds a check to ensure operation.hash_value is not empty before evaluating the completion condition. This change is concise, directly targets the described problem, and appears to be a solid fix for the race condition.