Serious bug in Red Hat Linux will affect servers based on Haswell architecture

Recently, Gil Tene, CTO and co-founder of Azul Systems, reported a very important but little-known Linux kernel patch in GoogleGroups. Users and administrators of Linux systems using Intel Haswell architecture should especially pay attention to this issue. Especially users based on Red Hat distributions (including CentOS 6.6 and Scientific Linux 6.6) should update this patch immediately. Even if Linux is running in a virtual machine, if the virtual machine is on a popular cloud platform (such as Azure, Amazon, etc.), it may also be running on a Haswell machine, and patching should be beneficial.

Tene describes the flaw as follows:

The impact of this kernel vulnerability is very simple: in some seemingly impossible situations, user processes will deadlock and hang. Any futex call waiting (even if correctly awakened) may be blocked forever. Just like Thread.park() in Java may block forever, etc. If you are lucky, you will find soft lockup messages in the dmesg log; if you are not so lucky (like us), you will have to spend several months of labor costs to troubleshoot the problem in the code, and you may not find anything. ”

Tene went on to explain how the defective code was executed (which ultimately boiled down to a switch block that missed a default case). The biggest problem now is that although the problematic code was fixed in January 2014, Around October, the defect was moved back to the Red Hat 6.6 family system. Other systems including SLES, Ubuntu, Debian, etc. may also be affected.

The repair status of these systems is now inconsistent and may be ignored. .RedHat users should use RHEL 6.6.z or newer. Tene also pointed out that another key point is that different distributions will have different choices about what to put into the kernel, which also leads to different fixes. Inconsistent.

For example, for RHEL 7.1, "In fact, the upstream 3.10 kernel does not have this bug, but the RHEL7 kernel is not a pure upstream version. Unfortunately, RHEL 7.1 (just like RHEL 6.6) included this bug in the port (based on the RHEL 7 version)... I think other distributions may do the same. ”

For RHEL-based distributions, Tene provides a quick reference list:

l RHEL 5 (including CentOS5 and Scientific Linux 5): All versions (including version 5.11) have no problems.

l RHEL 6 (including CentOS6 and Scientific Linux 6): There is no problem from version 6.0 to 6.5. But version 6.6 has a defect, and version 6.6.z has no problem

l RHEL 7 (including CentOS7 and Scientific Linux 7): 7.1 is defective. And there isn't a fix for 7.x as of May 13, 2015

Although there is some debate on Hacker News about the number of affected systems, it provides some context to check if your system needs a fix

Get LAMP Brothers’ original Linux operation and maintenance engineer video/detailed Linux tutorial for free. For details, please contact the official customer service: http://www.lampbrother.net/linux/