fck synology

Sorry to post this in English. I also contacted Synology in 2017 and they said it's a HW issue with my NAS when all the evidence I had tells me it can't be just me. The failure is deterministic and it was really due to my unit's issue, I would think the failure symptom would have FAR worse. But it was always xz that was causing the issue. So, FUCK SYNOLOGY! With that said, the root cause of this issue is the CPU bug with Marvell PJ4B which utilizes a multi-stage branch prediction framework. The xz's decoder hits a code-placement-sensitive, prediction-dependent execution fault. The unchanged binary passes with general branch prediction disabled and crashes again when restored. The real fix (well, HW WAR) needs to be done in the kernel. See https://lists.infradead.org/pipermail/linux-arm-kernel/2012-November/133978.html This fucking Synology didn't want to fix this issue and told me to upgrade my NAS by saying it is my unit's problem. Anyhow, I worked around this problem by avoiding the failing instruction layout and changing the link placement. - LDFLAGS=-Wl,--section-start=.text=0x2384 + LDFLAGS=-Wl,--section-start=.text=0x238c