Reverse Engineering Variability in Source Code Using Clone Detection: A Case Study for Linux Variants of Consumer Electronic Devices

The Consumer Electronics Working Group (CEWG) in the Linux Foundation has identified several problems in the re-use process of embedded Linux software for consumer electronic devices. Among these is the increasing fragmentation of Linux derivatives. Vendors of electronic devices copy the Linux sources and make their modifications to adapt it to their own devices, but fail to back port their modifications to the mainstream Linux sources. Likewise, later improvements of the Linux sources are not integrated into the vendors' variants. CEWG launched the Long Term Support Initiative (LTSI) for an industry-managed tree of the Linux sources, maintained by CEWG, that is based on the long-term stable kernel tree annually updated with the latest mainstream kernel version to address their needs. In order to justify this initiative, CEWG asked us to investigate whether and if so how much non-upstream code can be found in industry products and to which extent to and for what part of the kernel. We used large-scale clone detection techniques to compare various Linux versions to their vendor-specific variants. We found many changes that were not back ported. Some of these changes were even found in Linux subsystems where neither we nor people from the Linux Foundation would expect them. We also found instances of defects fixed in the mainstream kernel that were not integrated into the vendors' code. Overall, our investigation provides enough evidence to support the need for an LTSI and better collaboration among Linux developers both of the mainstream and the vendor variants.