1. Kernel

1.1. Interrupts

If a large number (e.g. 200,000 per second) of interrupts are seen under idle load with bridging is use, see below.

2. Network

2.1. Bridging causing large numbers of interrupts

The symptom is a high interrupt rate, e.g. 200,000 per second, when a bridge is created using a 'bridge hello time' (bridge_hello) of 0. It is believed to be a bug in the bridge code to add a timer for the current jiffies value.

The simplest work around is set the bridge hello time to be greater than zero.

On native Linux you get away with it as the timer won't fire until the next jiffy. On Xen, you'll enter Xen and then bounce straight back out as the time has already passed.

A possible hack to arch xen is to round to the next jiffy to match the native behaviour.

However, the bridge's behaviour is still pretty evil - you'll still end up executing the code HZ (100/1000) times a second, and the intention of the user was probably to disable execution of the code altogether. It won't slay the machine (like executing int 200k times a second), but its not ideal.