JLSEC-2026-620 json
WebSocket reader data race in auto-PONG/CLOSE-echo handling in HTTP.jl
published 4mo ago modified 3mo ago
- cvss
- —
- not scored
- cisa kev
- Not listed
- not in CISA's catalog
- epss
- —
- no score, EPSS covers CVEs only
- public exploits
- 0
- none known
Summary
### Description
The WebSocket reader task processed incoming frames by calling `ws_on_incoming_data!` without holding `ws.sendlock`. That function is not a pure parser: its auto-PONG and CLOSE-echo paths `push!` onto the shared `ws.codec.outgoing_frames` vector, while application send/ping/pong/close paths mutate the same vector while holding `ws.sendlock`. Because the reader did not take the lock, a remote peer flooding `PING` frames against a multithreaded server (`julia -t N`, N>1) could drive concurrent `push!`/`empty!` of a Julia `Vector` from two OS threads — undefined behavior that can corrupt the array metadata or segfault the process.
### Impact
A remote peer could crash, or potentially corrupt memory in, a multithreaded WebSocket server through concurrent frame processing.
### Patches
Fixed in HTTP.jl v2.4.0. A new `_process_incoming_data!` helper acquires `ws.sendlock` and runs `ws_on_incoming_data!` followed by `_flush_ws_output_locked!` under the lock, making each decode atomic with its flush; all production decode call sites route through it. The blocking read stays outside the lock so concurrent senders cannot deadlock the reader, and lock ordering is unchanged.
Reported to the JuliaLang security team through Anthropic's Coordinated Vulnerability Disclosure program.
Patch / fix
- julia HTTP 2.4.0
- github.com/JuliaWeb/HTTP.jl@f4f4027 commit
- github.com/JuliaWeb/HTTP.jl#1316 pull request
Is JLSEC-2026-620 in your dependencies? Fencer shows which of your repositories are affected as soon as an advisory lands, and opens the fix as a pull request.
Check with FencerAffected packages
| ecosystem | package | affected | fixed in |
|---|---|---|---|
| julia | HTTP | <2.4.0 |
2.4.0 |
References
Sources
- OSV JLSEC-2026-620
Fencer Intel keeps one entry per vulnerability: the CVE id wins, then GHSA, and records that alias each other are merged, except that two CVEs always keep separate entries. The same data is available as JSON.