HomePublicationsBEng Thesis

Exploring IVSHMEM in the Jailhouse Hypervisor
Ref: CISTER-TR-191211       Publication Date: 8, Oct, 2019

Abstract:
Nowadays, modern multicore processors come with virtualization features that provide the creation of different virtual environments inside the same machine, which magnify its ability to use all the resources available. The combination of multiprocessor systems with virtualization is highly demanded by the embedded systems domain. Virtualization technologies like hypervisor software are responsible for managing virtual machines and control their access to physical resources. Several virtualization technologies and hypervisors exist for different industry domains. One of those, is Jailhouse, a static partitioning hypervisor that partitions the hardware resources and directly assigns applications to each partition, providing them with access to the actual physical resources. This hypervisor focuses on giving the applications the isolation they need; however this can be seen as a limitation as it may restrict communication between applications running in different partitions. There are inter virtual machine’s communication mechanisms (based on networking or sharedmemory) that solves this limitation. This project aims at exploring this aspect in Jailhouse and focuses on using the Jailhouse hypervisor and a shared-memory mechanism to manage to send information between two partitions. Specifically, it aims at: (1) understanding the jailhouse hypervisor with respect to its features. For that, demonstrations are executed on top of two architectures, Intel based x86-64, using QEMU; and ARM, using a Banana Pi-M1 board; and (2) understanding a shared memory-based communication protocol, denoted as IVSHMEM, and create a use case in which two partitions exchange information using this protocol. Results of this work are promising as the x86-64 use case was successfully executed on top of QEMU, however, the ARM use case is still an on-going endeavour.

Diana Ramos

Document:

View
PDF Download
PDF