· via Hacker News – Front Page (native)
Critical OpenSSL EVP_EncodeBlock buffer overflow fix on hold pending maintainer consensus
A patch for a CVSS 9.8 buffer overflow in OpenSSL's base64 helper EVP_EncodeBlock is drawing attention on Hacker News, but the pull request remains on hold while maintainers debate the fix approach.
Critical base64 buffer overflow fix lands in OpenSSL's repo
A patch for a buffer overflow in OpenSSL's EVP_EncodeBlock() function reached the Hacker News front page on November 20, 2026. The flaw is tracked as CVE-2026-121481, and the Hacker News listing carries a CVSS score of 9.8, placing it at the top end of the severity scale. The fix itself is pull request #33652 against the openssl/openssl repository, submitted by contributor nmouha, and — unusually for a flaw rated this severe — it was still sitting under a hold label at the time the story circulated.
What the patch does
According to the pull request, the change addresses a buffer overflow in EVP_EncodeBlock(), one of OpenSSL's low-level base64 encoding helpers, and applies an equivalent guard to its decoding counterpart EVP_DecodeBlock(). Beyond the two public functions, it corrects a faulty length check in a base64encode() helper inside http_client.c, refreshes the documentation, and adds tests, including coverage that documents how zero-length inputs already behave.
EVP_EncodeBlock() converts raw bytes into base64 text by writing into a buffer supplied by the caller. A buffer overflow in such a routine means the function can write past the end of that buffer when its assumptions about input length break — a classic memory-corruption bug, and the kind that becomes dangerous in any call site where input sizes are not fully under the caller's control.
Two remedies, one open question
The pull request page shows the description in two revisions that part ways on the central question: what should these functions do with invalid input? One revision makes the function signal failure with a -1 return, rejecting bad arguments in a way the caller can handle. The other revision instead invokes OPENSSL_assert() on inputs that would cause undefined behavior — an approach that aborts the process rather than returning an error. Which behavior ultimately ships could matter a great deal to applications that have long treated these helpers as infallible.
Maintainers have not agreed
The review trail visible on the pull request shows the disagreement playing out in real time. Maintainer nhorman had approved the changes on November 11, 2026. The next day, November 12, t8m requested changes. On November 13, bob-beck attached a hold label to the pull request, with a description indicating that community consensus still needs to form around how to proceed. In other words, the dispute appears to concern the shape of the remedy — error return versus assertion — rather than the existence of the overflow itself.
What the source does not say
The Hacker News item and the pull request page leave several operational questions unanswered. There is no list of affected OpenSSL versions, no indication that a security release containing the fix has shipped, and no claim of exploitation in the wild. OpenSSL's security advisories normally carry exactly this kind of version and availability detail, so teams should treat an official advisory, once published, as the authoritative follow-up rather than the pull request alone.
Why it matters
OpenSSL ships inside an enormous share of the world's TLS stacks, and its base64 helpers are not exotic: encoding and decoding routines of this kind appear anywhere binary data must travel as text, from certificate handling to HTTP plumbing — the http_client.c fix in the same patch hints at internal usage as well. A 9.8-rated memory-corruption bug in one of them deserves priority triage even before full details arrive. The immediate practical step for anyone maintaining software that links OpenSSL is to establish whether their code, or a vendored dependency, calls EVP_EncodeBlock() or EVP_DecodeBlock() with lengths that are not strictly validated, and to watch for the advisory and patched releases. The hold label also carries a broader lesson: a headline severity score does not mean a fix is finalized, and here the merged, released remedy was still pending consensus as of the snapshot.
- #openssl
- #security
- #cve
- #buffer-overflow
- #open-source