GHSA-C23C-RP3M-VPG3

Vulnerability from github – Published: 2026-03-10 18:31 – Updated: 2026-03-10 22:54
VLAI
Summary
Envoy's global rate limit may crash when the response phase limit is enabled and the response phase request is failed directly
Details

Summary

At the rate limit filter, if we enabled the response phase limit with apply_on_stream_done in the rate limit configuration and the response phase limit request fails directly, it may crash Envoy.

Details

When both the request phase limit and response phase limit are enabled, the safe gRPC client instance will be re-used for both the request phase request and response phase request.

But after the request phase request is done, the inner state of the request phase limit request in gRPC client is not cleaned up. When we send the second limit request at response phase, and the second limit request fails directly, we may access the previous request's inner state and result in crash.

PoC

This need to mock the network failure. But we have reproduced by unit test locally.

Impact

This only happens when both the request phase limit and response phase limit are enabled in the rate limit filter, and requires the request to rate limit service fails directly (For example, if from Envoy's perspective, no healthy endpoint for rate limit service may result the request fails directly). That's say, not easy to trigger this.

To workaround

This could be worked around by splitting the rate limit filter. That is, if there is a rate limit filter that contains normal rate limit configuration (request phase limit, without apply_on_stream_done) and also rate limit configuration with apply_on_stream_done (response phase limit). Splitting them into two rate limit filters and ensure one filter only contains normal rate limit configuration (without apply_on_stream_done), and one only contains rate limit configuration with apply_on_stream_done could avoid this problem.

Credit

Mandar Jog (mandarjog@gmail.com)

Show details on source website

{
  "affected": [
    {
      "package": {
        "ecosystem": "Go",
        "name": "github.com/envoyproxy/envoy"
      },
      "versions": [
        "1.37.0"
      ]
    },
    {
      "package": {
        "ecosystem": "Go",
        "name": "github.com/envoyproxy/envoy"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "1.36.0"
            },
            {
              "last_affected": "1.36.4"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-26330"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-416"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-03-10T18:31:57Z",
    "nvd_published_at": "2026-03-10T20:16:36Z",
    "severity": "MODERATE"
  },
  "details": "### Summary\n\nAt the rate limit filter, if we enabled the response phase limit with `apply_on_stream_done` in the rate limit configuration and the response phase limit request fails directly, it may crash Envoy.\n\n### Details\n\nWhen both the request phase limit and response phase limit are enabled, the safe gRPC client instance will be re-used for both the request phase request and response phase request.\n\nBut after the request phase request is done, the inner state of the request phase limit request in gRPC client is not cleaned up. When we send the second limit request at response phase, and the second limit request fails directly, we may access the previous request\u0027s inner state and result in crash.\n\n\n### PoC\n\nThis need to mock the network failure. But we have reproduced by unit test locally.\n\n### Impact\n\nThis only happens when both the request phase limit and response phase limit are enabled in the rate limit filter, and requires the request to rate limit service fails directly (For example, if from Envoy\u0027s perspective, no healthy endpoint for rate limit service may result the request fails directly). That\u0027s say, not easy to trigger this.\n\n### To workaround\n\nThis could be worked around by splitting the rate limit filter. That is, if there is a rate limit filter that contains normal rate limit configuration (request phase limit, without `apply_on_stream_done`) and also rate limit configuration with `apply_on_stream_done` (response phase limit). Splitting them into two rate limit filters and ensure one filter only contains normal rate limit configuration (without `apply_on_stream_done`), and one only contains rate limit configuration with `apply_on_stream_done` could avoid this problem. \n\n### Credit\n\nMandar Jog (mandarjog@gmail.com)",
  "id": "GHSA-c23c-rp3m-vpg3",
  "modified": "2026-03-10T22:54:51Z",
  "published": "2026-03-10T18:31:57Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/envoyproxy/envoy/security/advisories/GHSA-c23c-rp3m-vpg3"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-26330"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/envoyproxy/envoy"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:N/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Envoy\u0027s global rate limit may crash when the response phase limit is enabled and the response phase request is failed directly"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.


Loading…

Loading…

Loading…

Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.

Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

Sightings

Author Source Type Date Other

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline

Sightings and recorded dates, grouped by day · UTC

JSON
How durations are calculated

Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.

Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.

Events included in the timeline

Loading timeline…


    Loading…

    Loading…

    Loading…

    Related by attack behaviour

    Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.