OK, Thank you for your patience. the last question: > I forgot the details, but when I tried moving HPET to 64bit it did not > work on one of my machines due to an erratum and other people reported > similar issues on different CPUs/chipsets. > > TBH, I'm not interested at all to chase down these buggy implementations > and have yet another pile of quirks. Can you tell me the erranum or issue link at that time? This is very important to us. Thank you very much. > 2021年7月12日 下午3:25，Thomas Gleixner <tglx@xxxxxxxxxxxxx> 写道： > > Liao, > > On Mon, Jul 12 2021 at 12:52, Linux wrote: >>> Sorry, keeping the softirq from running for 3 minutes is simply out of >>> spec. If the sysadmin decides to do so, then he can keep the pieces. >> >> It is because the kernel thread is busy that the clocksource_watchdog >> thread is not scheduled, not softirq. > > Which thread? > > The clocksource watchdog runs from a timer_list timer callback in > softirq context. Even if the softirq is switched to the softirq thread > then still my argument of starving that for 3 minutes still stands. > > This is _not_ a kernel problem. Overcommitment is a admin problem. > > Thanks, > > tglx
