Abstract—Some real-time kernels (such as a recent realtime version of Linux) permit to execute interrupt handlers in dedicated threads, to control their interference on realtime applications. However, from the stand-point of real-time analysis, such threads are challenging and cannot be dealt with in the traditional ways. Furthermore, the application of traditional scheduling solutions (such as fixed priorities) proves ineffective in striking a good trade-off between predictability and hardware performance. This paper shows how the problem ackled by using the resource reservation abstraction and an appropriate model for schedulability analysis.