A CPI support was added to Linux in 2002 and a dummy wake op was needed for AMD chipsets that weren’t idle when they were supposed to be. It was a quick ’n’ dirty workaround. Move forwards 20 years, and you guessed it, that workaround is still being applied, only now it’s not being particularly helpful, quietly causing performance issues with some workloads. Its original job has long been redundant on all but the oldest Athlon boxes. Once AMD engineers eventually spotted it, a patch was forthcoming and folded into the Linux 6.0 kernel. How many workarounds are there lurking in the bowels of the code waiting to be uncovered? How much cruft? Probably more than we like to think. Still, at least this wasn’t a remote code execution vulnerability, just a performance drag.