Hans van Kranenburg
2020-Dec-15 11:18 UTC
[Pkg-xen-devel] [PATCH 7/9] debian/xen.init: Load xen_acpi_processor on boot
Hi, On 12/1/20 1:34 AM, Elliott Mitchell wrote:> On Mon, Nov 30, 2020 at 11:43:04PM +0100, Hans van Kranenburg wrote: >> On 9/14/20 6:23 AM, Elliott Mitchell wrote: >>> This allows more control of processor state, potentially resulting in >>> reduced power usage. Alternatively simply more information on processor >>> use. >> >> Ooh, nice. >> >> Would you mind to explain how you discovered this one? > > I have a Xen machine which isn't heavily loaded. Tried running `sensors` > and what do I find? Minimum power usage on the processor was > unexpectedly high for the processor and its usage. So I try a handy > search engine to see what turns up. > > The source of information I found mentioned the kernel module > xen_acpi_processor. Also mentioned `xenpm get-cpuidle-states` and > `xenpm get-cpufreq-states` (which depend on xen_acpi_processor) as > sources of information. > > xen_acpi_processor dropped power usage significantly by itself. > `xenpm get-cpuidle-states` though listed fewer levels of CPU-idle on > several cores. The number of cores which were missing deeper power > saving states was the delta between Domain 0's vCPUs and the number of > cores. Allowing Domain 0 to get vCPUs equal to the cores on boot, but > then hot-remove several (`xl vcpu-set 0 12`) got the rest of the cores to > low-power idle when possible. > > Other hardware on the machine still consumes plenty of power, but the > main processor now uses about one-third of what it had been.I'm testing this change, but it seems I can't get past the "modprobe: ERROR: could not insert 'xen_acpi_processor': No such device" error on any hardware I have to test with. (with linux-image-5.9.0-4-amd64 version 5.9.11-1 as dom0 kernel). https://paste.debian.net/plainh/bbd01e81 Any idea? Hans
Elliott Mitchell
2020-Dec-15 16:43 UTC
[Pkg-xen-devel] [PATCH 7/9] debian/xen.init: Load xen_acpi_processor on boot
On Tue, Dec 15, 2020 at 12:18:15PM +0100, Hans van Kranenburg wrote:> I'm testing this change, but it seems I can't get past the "modprobe: > ERROR: could not insert 'xen_acpi_processor': No such device" error on > any hardware I have to test with. (with linux-image-5.9.0-4-amd64 > version 5.9.11-1 as dom0 kernel).According to appropriate documentation, ENODEV comes from the module itself and not the standard loading process. In linux-5.9/drivers/xen/xen-acpi-processor.c there are 3 occurrences of ENODEV, so it would be one of those. One of those has a pr_warn(), so it would directly show up in `dmesg`. The other two don't look as likely to trigger and I'm unsure what messages they would produce. My main Xen machine is presently on Xen 4.11 and kernel 4.19. Seems the comment above read_acpi_id() has been broken. :-( Right now if Dom0 has fewer vCPUs than the machine has actual processors, the C-states don't get setup for the processors which lack vCPUs. Workaround I've got right now is hot unplugging some processors after boot. -- (\___(\___(\______ --=> 8-) EHM <=-- ______/)___/)___/) \BS ( | ehem+sigmsg at m5p.com PGP 87145445 | ) / \_CS\ | _____ -O #include <stddisclaimer.h> O- _____ | / _/ 8A19\___\_|_/58D2 7E3D DDF4 7BA6 <-PGP-> 41D1 B375 37D0 8714\_|_/___/5445