From time to time there is a point where you encounter very annoying things in open source software or while using non-free tools. One of them was qemu.
As some of you might be aware of, I am a huge fan of Xen. I love the idea of using paravirtualization for almost native performance of the virtual guests, furthermore version 4.0 really shows a high level of maturity.
While working I have to use kvm in most cases. At the beginning this was very hard for me :P, but well, after a couple of weeks I began to arrange myself with the fact that kvm really has many advantages over Xen. Of them would be the hardware acceleration for qemu quests which leads to a very good performance in HVM systems.
One thing I encounter during the last time I installed kvm was the annoying fact that there is actually a hard-coded value for the maximum amount of network interfaces. Whatdafuck. I tried to boot a VM with 12 network devices and there it goes, kvm throws an error, it was something like “too many devices”. Awesome.
Doing some research, I found out that this “bug” is known for a while now and nobody ever changed it. There are plans to make this value more dynamic, but still MAX_NICS is set to 8 today (afaik). So i had to download the source, raise the value to 24 and rebuild it. Jesus, why…
It works now and I am very satisfied with this solution. An attempt to raise the limit permanently was declined several times (in the upstream version), since the devs want to have a dynamic solution.
Even more fun was the thing I encountered afterwards. HP offers a tool called hpacucli, a RAID configuration and monitoring tool for Linux systems. Since 2010, there is a well-known bug.. a bug I saw today.
hpacucli may cause the kernel to crash if being called like 2 or 3 times:
For me, this is something which simply may not happen. Usually HP servers are being used for critical systems, since they cost a lot of money.