GPU VulnDB

Database/Kernel, userspace & hypervisor

OpenSSL: SSL_select_next_proto() over-reads and can send up to 255 bytes of process memory to the peer

CVSS 9.1CVE-2024-5535Kernel, userspace & hypervisorcurated

Impact

When SSL_select_next_proto() is called with a zero-length client protocol list it fails to detect that condition and returns the memory immediately after the list pointer, so up to 255 bytes of adjacent process memory can be sent to the TLS peer or the process can crash. For an operator the practical question is narrow: only an application that directly calls this function with an empty list is affected, and for ALPN on the server side libssl guarantees the client list is never empty. The realistic exposure is a service on the fleet that was accidentally configured with an empty protocol list and also mishandles the "no overlap" result - there the leak is of that daemon's heap, which on a TLS-terminating control-plane or registry service can include key material or session data. OpenSSL is shared by most networked daemons on a node, so remediation is a library update plus restarts rather than a single-service patch.

Who can reach it

A TLS peer that completes or attempts a handshake against an affected application - unauthenticated at the TLS layer. The misconfiguration is in the calling application, not under the peer's control, so the peer only harvests the leak where the application is already built or configured this way.

What to do

Update OpenSSL to a release carrying commits 4ada436a1946, 99fb78541306, cf6f91f6121f or e86ac436f0bd as shipped by your distribution, then restart every daemon linked against libssl - on a GPU fleet that is a rolling daemon-restart across ingress proxies, registries, API servers and monitoring agents. Statically linked or vendored OpenSSL copies in container images need rebuilt images. The advisory does not require a reboot.

References

Related entries

All Kernel, userspace & hypervisor entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.