Rack's unbounded multipart preamble buffering enables DoS (memory exhaustion)

Description

Summary

Rack::Multipart::Parser buffers the entire multipart preamble (bytes before the first boundary) in memory without any size limit. A client can send a large preamble followed by a valid boundary, causing significant memory use and potential process termination due to out-of-memory (OOM) conditions.

Details

While searching for the first boundary, the parser appends incoming data into a shared buffer (@sbuf.concat(content)) and scans for the boundary pattern:

@sbuf.scan_until(@body_regex)

If the boundary is not yet found, the parser continues buffering data indefinitely. There is no trimming or size cap on the preamble, allowing attackers to send arbitrary amounts of data before the first boundary.

Impact

Remote attackers can trigger large transient memory spikes by including a long preamble in multipart/form-data requests. The impact scales with allowed request sizes and concurrency, potentially causing worker crashes or severe slowdown due to garbage collection.

Mitigation

  • Upgrade: Use a patched version of Rack that enforces a preamble size limit (e.g., 16 KiB) or discards preamble data entirely per RFC 2046 § 5.1.1.
  • Workarounds:
  • Limit total request body size at the proxy or web server level.
  • Monitor memory and set per-process limits to prevent OOM conditions.

Basic information

Type
reviewed
Severity
high
Advisory on GitHub
Open advisory ↗
Repository advisory
Open repository advisory ↗
Source code
Browse source ↗
Published (advisory)
2025-10-07 17:26:16 UTC
Updated
2025-10-13 15:29:38 UTC
GitHub reviewed
2025-10-07 17:26:16 UTC
NVD published
2025-10-07

EPSS Score

Score Percentile
0.16% 36.53%

CVSS Scores

Base score Version Severity Vector
7.5 3.1
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H Click to expand
Attack vector (AV:N)
Could be attacked over the internet or any normal routed network—not just someone sitting at the machine.
Attack complexity (AC:L)
Once they can reach the bug, pulling it off is straightforward—no weird race conditions or rare setup.
Privileges required (PR:N)
No account or special rights needed—anonymous or random user is enough.
User interaction (UI:N)
Nobody has to click “OK” or open a trap file; it can work without a victim helping.
Scope (S:U)
Damage stays in the same “trust bubble” as the broken component—no big spill into unrelated systems.
Confidentiality (C:N)
Doesn’t really leak secrets in a meaningful way.
Integrity (I:N)
Data isn’t meaningfully altered or forged.
Availability (A:H)
Could take the service down hard or make it unusable for people who depend on it.

Identifiers

CWEs

CWE id Name
CWE-400 Uncontrolled Resource Consumption

Credits

  • kwkr (reporter)
  • ioquatix (remediation_reviewer)
  • jeremyevans (remediation_developer)

Affected packages (3)

Vulnerable version ranges and first patched releases as published by GitHub.

Ecosystem Package Vulnerable range First patched Vulnerable functions
rubygems rack < 2.2.19 2.2.19
rubygems rack >= 3.1, < 3.1.17 3.1.17
rubygems rack >= 3.2, < 3.2.2 3.2.2

References

cvelogic Threat Intelligence