On 03.07.2019 15:13, awokd via coreboot wrote:
Michal Zygowski:
On 03.07.2019 04:30, awokd via coreboot wrote:
Michal Zygowski:
On 01.07.2019 14:53, Andriy Gapon wrote:
It appears that HPET MSI support is disabled on some platforms by default:
src/vendorcode/amd/agesa/f15tn/Proc/Fch/Interface/Family/Hudson2/EnvDefHudson2.c:
TRUE, // HpetMsiDis
FWIW, I hardcoded the above to FALSE, recompiled, and reflashed my f15tn. It caused Qubes to run about half as fast. I checked the logs and didn't see any errors to explain why. I then changed it back to TRUE and HPET itself to FALSE (and recompiled/reflashed). Normal speed came back, XEN Platform timer changed from HPET to ACPI, and MSI is still enabled for some of the PCI devices. For example, the video controller on 00:01.0 is IRQ 57 and capability [a0] is MSI: Enable+.
Yes, for PCI devices it can be checked by verbose lspci. However I still wonder how did You determine that HPET is not using MSI?
Apologies, omitted that part. With the original default of HPET enabled and HpetMsiDis TRUE, the speed was fine. Xen platform timer indicated HPET in this case. Devices also still seemed to be using MSI, so I'm not sure what that option is supposed to do except slow down my system.
If timer interrupts are not handled properly for some reason (in this case HPET timer), it is only natural that it will dramatically slow down the system. The MSI can be working, but may also require some IOAPIC configuration. Can't say much without any logs from firmware/OS.
When You set the HPET option to FALSE, it probably did not touch HPET in AGESA, that is why the speed came back. I would check whether HPET is operable for this processor family. For example on Intel Braswell platform HPET was not guaranteed operable in certain conditions.
BTW have You tried with family 16h (i.e. PC Engines APU2)?
Don't have one of those, but wanted to give a (possibly useless) data point of my experience at least.
Sorry, I have mistakenly taken You as the initiator of the thread, who stated he has an APU2 platform. Should have checked the email address first.
coreboot mailing list -- coreboot@coreboot.org To unsubscribe send an email to coreboot-leave@coreboot.org