CVE-2024-6387: RCE in OpenSSH's server, on glibc-based Linux systems

Open link in next tab

https://seclists.org/oss-sec/2024/q3/2

Regression in signal handler.

This vulnerability is exploitable remotely on glibc-based Linux systems, where syslog() itself calls async-signal-unsafe functions (for example, malloc() and free()): an unauthenticated remote code execution as root, because it affects sshd's privileged code, which is not sandboxed and runs with full privileges.

Sign in to add comment

8.5p1 <= OpenSSH < 9.8p1 is vulnerable

what does that mean? I don't understand multiple signs in the same sentence and what is the significance of having "OpenSSH" in the middle?

You can read them as separate statements with the middle repeated and a logical AND between them:

If (8.5p1 <= your OpenSSH version) AND (your OpenSSH version < 9.8p1) Then you are vulnerable

It’s the same as saying if your OpenSSH version is between these two versions (including 8.5p1, but not 9.8p1), then you are vulnerable

I don't get it... wouldn't everything < 9.8p1 already include <= 8.5p1? So why is it even necessary to mention?

Because this is a regression and this particular issue was introduced in 8.5p1. So it only affects versions newer than that, up until when it was fixed in 9.8p1.

For an integer, 4 < x < 6 x has to be 5. It's the only value that satisfies all sides of the equation.
You are deriving a set of values for open ssh that satisfies all sides of the equation.

I think it's more of a mathematical representation than programming representation (I mean, I don't know of a language that would accept that syntax).
Certainly psuedocode would have quick statements like this

seems to work fine in C and I can find quite a bit of examples of it being used actually

Oh, I can't find any examples. What are you searching for?
The closest I can find is an old hlsl offhand comment showing the syntax in isolation, but no example.
https://stackoverflow.com/a/29689866