Musl libc Bug: Could Alpine SSH Accept Wrong Passwords?
A password check quirk in musl libc surfaces publicly
A detailed technical writeup published under the title "sshole" and shared on Hacker News on November 5, 2026, lays out a chain of small design decisions in musl libc, the C standard library behind Alpine Linux, that can under specific conditions let SSH accept a wrong password as valid. The report walks through how Unix password checking works at a low level and pinpoints where things break down, specifically when OpenSSH runs on musl without PAM, the pluggable authentication modules framework most Linux distributions rely on to handle logins.
The author presents this as a research exercise rather than an active exploit campaign. They describe testing multiple builds of OpenSSH, combining glibc and musl, with and without PAM, with and without shadow files, against roughly 17,600 unusual login attempts. Every build behaved consistently except one: musl without PAM, which in some cases accepted logins it should have rejected.
How the underlying mechanism works
On traditional Unix systems, password hashes live in /etc/shadow, and a function called crypt() hashes a typed password with a stored salt and compares the result against the saved hash. A match grants access. A special character such as an asterisk in the hash field is a long-standing Unix convention meaning the account cannot be logged into with a password at all, since no computed hash can ever equal a single asterisk.
The writeup explains that crypt() is supposed to return NULL when it cannot compute a hash, for instance if the requested algorithm is unavailable. Many programs never check for that NULL result, which causes crashes. Musl's developers addressed this by having crypt() return a fixed placeholder character instead of NULL on failure, so calling programs would not crash. According to the source, that placeholder character happens to be the same asterisk that conventionally marks a locked account.
Why the asterisk choice matters now
Musl also enforces a 256-byte cap on passwords processed through SHA-512 crypt, a hashing scheme that gets slower as input length grows, to stop someone from tying up a server by submitting an extremely long password. When a submitted password exceeds that cap, crypt() fails and again returns the placeholder asterisk, per the writeup.
The concern is that under certain conditions, this asterisk fallback can intersect with how sshd compares results, producing situations where a login that should fail instead succeeds. The source is explicit that this specific chain was found through the author's own testing, not confirmed as something exploited in the wild. It is not yet clear from the source how widely this combination of conditions appears in production systems.
Who this affects and how
Two groups matter here. The first is anyone running Alpine Linux containers or servers, since Alpine is built on musl rather than glibc, and container images are frequently built without a full PAM stack to keep them small. The second is infrastructure and platform teams maintaining cloud images and container base layers.
The writeup separately references a related 2019 issue, CVE-2019-5021, where Alpine Docker images shipped with no root password set at all, and a 2022 change in Alpine 3.16 that moved root's shadow entry to the asterisk character specifically to keep it locked while still allowing SSH key based logins to work. It also references CVE-2016-6210, a separate but related timing issue where sshd could respond faster to login attempts for usernames that do not exist at all, since there is no hash to compute, letting an attacker time responses to enumerate valid usernames. The documented fix has sshd compute a dummy hash against a decoy salt even for invalid users, deliberately wasting time so attacker-observable response times stay consistent regardless of whether the account is real.
What is sensible to do now versus what is not
For teams running Alpine-based containers, the source's technical narrative points toward checking whether your OpenSSH and musl builds are current and whether PAM is in use, since the anomaly described was specific to the no-PAM configuration. This is not the same as saying every Alpine deployment is actively vulnerable. The writeup describes a condition discovered during testing, not a confirmed breach pattern seen in production, so treating it as a reason to panic-migrate off Alpine goes further than the evidence supports.
What is reasonable is routine patching hygiene: confirm OpenSSH package versions are current, review whether SSH access to Alpine-based hosts relies on password authentication at all rather than key-based login, and check container base image update logs for related fixes. Teams using SUNS Tech's IT outsourcing services for infrastructure management often have this kind of dependency review built into routine maintenance cycles, which holds up better over time than a one-off emergency check triggered by a headline.
Signals worth watching going forward
A Debian developer has reportedly asked the glibc project to adopt musl's asterisk convention for failed crypt() calls. Debian itself moved to a library called libxcrypt six years after a related bug report, using a slightly different placeholder, "*0", specifically to avoid colliding with the locked-account convention. That one-character difference is presented in the writeup as the practical fix.
Watch for official security advisories from Alpine, musl, and OpenSSH maintainers that may formalize guidance beyond this independent writeup, since as of publication the report reads as original research rather than a coordinated vendor disclosure. Also worth tracking is whether distributions that rely on musl without PAM issue their own statements, since the source notes that most mainstream distributions route authentication through PAM, which never calls crypt() directly on a locked asterisk account in the first place.
If your organization runs infrastructure on Alpine-based images and wants a second set of eyes on how authentication is configured across your container fleet, our software consulting team can walk through your setup together with you. You can also reach out directly through our contact page to discuss specifics.









