While we are not declaring a critical vulnerability, this is information that you should know. There is a zero day exploit in the wild that can work against SLF 6. This *does* require local access, but privilege escalation is pretty trivial. There is a workaround available.
- Full vulnerability information will be tracked as CVE-2013-2094 (just a candidate currently).
- This is a bug that needs PERF_EVENTS be compiled into the kernel, which is by default on most distros (including SL/SLF).
- This bug affects kernels 2.6.37 to 3.8.8. While this is a "new" bug, it affects 2.6.37 due to being backported.
- The bug also got backported into kernel 2.6.32 on CentOS, RHEL, and SL/SLF.
- Since SL/SLF 6 uses kernel 2.6.32-71, it is vulnerable.
(As an FYI, version 5 uses kernel 2.6.18-8 and is not vulnerable)
The workaround is setting kernel.perf_event_paranoid to a value of 2, e.g.,
# /bin/sysctl kernel.perf_event_paranoid=2
(values are as follows: -1 = not paranoid, 0 = disallow raw tracepoint access for unpriv, 1 = disallow cpu events for unpriv, and 2 = disallow kernel profiling for unpriv)
By setting this parameter to 2, unprivileged users won't be able to get kernel profiling data (e.g., via the perf command). This shouldn't affect a lot of people.
As such, CST recommends that if you are using SLF 6, please set the perf_event_paranoid kernel parameter to 2.
Thanks,
- Art Lee
Comments
Post a Comment
https://gengwg.blogspot.com/