On Fri, Mar 14, 2025 at 4:50 AM Chris Murphy <chris@xxxxxxxxxxxxxxxxx> wrote: > > > > On Thu, Mar 13, 2025, at 9:35 AM, Rafael J. Wysocki wrote: > > On Thu, Mar 13, 2025 at 5:12 AM Chris Murphy <chris@xxxxxxxxxxxxxxxxx> wrote: > >> > >> Four kacpi_notify kernel threads are continuously using CPU, makes the laptop warm, > >> fans run continuosly. This is a Fedora debug kernel, it's not intended for production. > >> But I'm wondering if this performance hit is expected and if it's worth it (for kernel developers). > > > > No and no. > > Is it a firmware or kernel bug? Any other info I should provide? It may not be a bug. Additional computational overhead related to debug options may increase power and temperature which would increase the rate at which events are generated which would increase the overhead and so on. > I updated the firmware to the latest firmware, but the behavior is unchanged. I'm not sure what > kernel config might be responsible for this, so I'm not sure what to suggest to the kernel team > to make the performance hit go away. It's hard to say, something that would add overhead to the processing of firmware events. > > kacpi_notify is only used for Notify() processing which only happens > > when there are events signaled by the platform firmware. Like battery > > or thermal events, for example. > > It seems to happen more often than not, happens with AC power connected or not. I would look at the counters in /sys/firmware/acpi/interrupts and compare the debug case with the non-debug one. If the counters generally grow faster in the debug case, more events are signaled and that would be the reason for the increased load. Then you can try to figure out which events are involved in this (but ACPI tables inspection would be necessary I think).
