LowGHSA-935w-9g4m-p28pCVE-2026-105752PYSEC-2026-4308

vLLM: Harmony tool continuations drop `cache_salt` — restoring a cross-tenant prefix-cache membership oracle

Published 6 Oct 2026 · Updated 15 Oct 2026

Is your project on an affected vllm?

Private repo? Upload its lockfile instead, or connect GitHub to check every repo.

Look after several Python apps? Indie watches up to 15 repos, private ones too, for $5 a month. See Indie

Affected package

vllmPyPI
Affected
<0.30.0
Fixed in
0.30.0
Update to the fixed version for your release line, for example:pip install 'vllm>=0.30.0'

Details

Affected

  • Ecosystem / package: pip / vllm
  • Affected versions: vLLM ≤ 0.25.1 (confirmed on 0.25.1, commit `752a3a504485`). The lower bound predates 0.25.1; maintainers can confirm how far back the tool-continuation re-submission has omitted the salt.

Summary

On the GPT-OSS "Harmony" path (POST /v1/responses), a request that uses a built-in or MCP tool runs as a multi-turn loop: after each tool call vLLM re-renders the full next-turn Harmony prompt and re-submits it to the engine. Turn 1 correctly carries request.cache_salt, but the tool-continuation re-submission rebuilds the engine input via tokens_input(token_ids) with no cache_salt. The continuation prefix is therefore cached in the global *unsalted* namespace even though the caller opted into salting. A second tenant who can guess the low-entropy post-tool history submits the reconstructed continuation (unsalted) and reads exact per-turn cached-token counts from the Responses usage — restoring the prompt-membership oracle that cache_salt is documented to prevent.

Silently dropping a preserved salt *after* the supported tool workflow is enabled is a broken isolation control: the caller enabled salting and every turn should stay isolated, but continuation turns leak into the shared cache.

This is distinct from GHSA-4qjh-9fv9-r85r (CVE-2025-46570): that advisory is the prefix-cache membership oracle for which cache_salt is the documented mitigation, and its PR-17045 fix does not close this site — the Harmony tool continuation silently drops the preserved salt, caching in the unsalted namespace and leaking exact cached_tokens_per_turn counts from a different sink (the Responses serving continuation, not general TTFT timing).

Affected code

Links pinned to the confirmed commit `752a3a504485` (v0.25.1):

The tool-continuation re-submission rebuilds the engine input with no cache_salt:

# vllm/entrypoints/openai/responses/serving.py Lines 711-715
            if isinstance(context, HarmonyContext):
                token_ids = context.render_for_completion()
                engine_input = tokens_input(token_ids)

                sampling_params.max_tokens = max_model_len - len(token_ids)

Contrast with the correct turn-1 call, which does preserve the caller's salt:

# vllm/entrypoints/openai/responses/serving.py Lines 754-755
        prompt_token_ids = render_for_completion(messages)
        engine_input = tokens_input(prompt_token_ids, cache_salt=request.cache_salt)

tokens_input stores the salt on the engine input only when it is passed, so the continuation input carries none and lands in the unsalted namespace:

# vllm/inputs/engine.py Lines 51-66
def tokens_input(
    prompt_token_ids: list[int],
    *,
    prompt: str | None = None,
    cache_salt: str | None = None,
) -> TokensInput:
    """
    Construct [`TokensInput`][vllm.inputs.engine.TokensInput]
    from optional values.
    """
    inputs = TokensInput(type="token", prompt_token_ids=prompt_token_ids)

    if prompt is not None:
        inputs["prompt"] = prompt
    if cache_salt is not None:
        inputs["cache_salt"] = cache_salt

Impact

An authenticated tenant of a shared deployment can recover whether a guessed post-tool prompt or history was processed by another tenant, with exact cached-token counts rather than noisy latency — the exact prompt-membership oracle cache_salt is documented to prevent. It defeats the multi-user prefix-cache isolation guarantee for salted Harmony tool sessions.

Preconditions: a GPT-OSS Harmony model on /v1/responses; prefix caching enabled (default); an operator-enabled built-in or MCP tool server; the victim sets cache_salt and triggers at least one tool continuation; and the attacker can reconstruct the post-tool history closely enough to match the token prefix. The AC:H metric reflects that guessable-history precondition.

Suggested Fix

Propagate request.cache_salt into every Harmony (and Parsable) tool-continuation re-submission — at the continuation call site call tokens_input(token_ids, cache_salt=request.cache_salt), mirroring the correct turn-1 call. Carry the salt on the HarmonyContext (thread the originating request into the context) so no continuation path can omit it:

# vllm/entrypoints/openai/responses/serving.py
             if isinstance(context, HarmonyContext):
                 token_ids = context.render_for_completion()
-                engine_input = tokens_input(token_ids)
+                engine_input = tokens_input(
+                    token_ids,
+                    cache_salt=(
+                        context.request.cache_salt
+                        if context.request is not None
+                        else None
+                    ),
+                )

with HarmonyContext.__init__ gaining a request: ResponsesRequest | None = None parameter (stored as self.request) that _create_responses passes when constructing the context. The continuation prefix is then cached in the victim's salted namespace, mirroring turn 1.

Suggested regression test: assert cached_tokens_per_turn == 0 for a different-salt probe against a salted victim continuation (the four-way control from the proof of concept).

Credit

Reported by: Patch the Planet (Trail of Bits + OpenAI collaboration)

This vulnerability was discovered using GPT-5.5-Cyber as part of the Patch the Planet security initiative.

---

Proposed fix: a fix for this issue is proposed in a public pull request: https://github.com/vllm-project/vllm/pull/51818

Is GHSA-935w-9g4m-p28p in any of your client repos?

Checking one package by hand is fine. Checking it across forty client projects is not. NPMscan reads the lockfiles in every GitHub repo you pick, finds each project with an affected vllm, and gives you the command that fixes it.

More advisories for vllm

  • LowCVE-2026-115917vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server · 28 Oct 2026
  • MediumCVE-2026-115913vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine · 28 Oct 2026
  • MediumCVE-2026-115914vLLM DeepSeek-OCR image processing vulnerable to denial of service · 28 Oct 2026
  • HighCVE-2026-115915vLLM: Untrusted EPS inputs can trigger denial of service or Ghostscript RCE · 28 Oct 2026
  • HighCVE-2026-115912vLLM: Public multimodal SHM handle forgery reaches worker-side `pickle.loads()` with insecure serialization disabled · 28 Oct 2026
  • MediumCVE-2026-115916vLLM: Out-of-vocabulary prompt token id passes input validation and triggers a CUDA device-side assert (engine crash / DoS) · 28 Oct 2026
All 113 advisories for vllm

References

Advisory data from OSV.dev (opens OSV.dev in a new tab), refreshed daily.