[Bug 192316] Invariant TSC gets misdetected on Intel Core 2 Duo processors, resulting in sluggish system behavior

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Wed Sep 17 16:15:39 UTC 2014


https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=192316

--- Comment #3 from Jan Kokemüller <jan.kokemueller at gmail.com> ---
Created attachment 147407
  --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=147407&action=edit
acpidump of a Lenovo SL510

Thanks for the detailed analysis. I've had another look at how my laptop
(Lenovo SL510) handles C-states. By default the OS queries the _CST method of
the firmware and this results in the following C-states:

dev.cpu.0.cx_supported: C1/1/1 C2/2/57
dev.cpu.1.cx_supported: C1/1/1 C2/2/57

So this means that cpu_can_deep_sleep does not get set. When I insert a "return
(ENXIO);" at the start of the acpi_cpu_cx_cst function in
sys/dev/acpica/acpi_cpu.c the OS falls back to FADT/P_BLK parsing and I get the
following:

dev.cpu.0.cx_supported: C1/1/0 C2/2/1 C3/3/57
dev.cpu.1.cx_supported: C1/1/0 C2/2/1 C3/3/57

... which looks closer to the actual hardware to me. I've attached the
acpidumps. The _CST method is in ssdt5.dsl. My laptop seems to fall into one of
the cases at lines 115, 318, or 531. I wonder why they do this weird remapping
of C-states?

-- 
You are receiving this mail because:
You are the assignee for the bug.


More information about the freebsd-bugs mailing list