On Wed, 2003-10-22 at 16:16, Joshua Narins wrote: > Amazing ,Ben, per usual. > > Was there an explanation offered before how something like cpufreq and a Mac > OS upgrade could be in conflict with each other? > > If not, I would be interested in knowing.
The problem was related to some subtle issues in the protocol between the kernel and the PMU chip. In some cases, the kernel could still send a 0x78 command (request "events" from the PMU) after sending a cpufreq switch request and before the cpu got actually rebooted in the new freq. It seems the MacOS update also updates the PMU firmware and the new firmware doesn't deal with any such request coming in in that timeframe. I added a security preventing any request to be sent to the PMU in those cases (cpufreq, but also sleep, which uses a similar mecanism, so might be affected as well, and reboot/shutdown) Ben.
