Created attachment 278489 [details] frequency_info I'm using ArchLinux distro and Gentoo on MSI Apache Pro GE 72 6QD. Until 4.15 all works correctly, the CPU frequency was 800 Mhz on idle. From the kernel 4.17 (and now 4.18.6) the CPU Frequency is 3.1 GHz and the temperature is 65-68 degree Celsius (always idle). scalling_governor = powersave and "cpupower frequency-info" shows: current CPU frequency: Unable to call hardware
Created attachment 278491 [details] scalling_governor cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
Created attachment 278493 [details] turbostat -d sleep 10
When I'm disabling on BIOS the C-states the CPU Frequency changes between 1.1GHz to 2.5GHz. But with the kernel 4.15 and C-States enabled the CPU frequency is less and the battery life is better.
Created attachment 278495 [details] ACPI Error: Method parse/execution failed on boot
The problem seems to be solved if C-States is disabled from BIOS but in my opinion this is not a good solution.
what's the latest kernel you have tried? can you please confirm if the problem still exists in the latest -rc, say, 4.19-rc4? for this error sep 13 14:57:31 SYSTEM kernel: ACPI Error: No handler for Region [EC__] ((____ptrval____)) [EmbeddedControl] (20180531/evregion-132) sep 13 14:57:31 SYSTEM kernel: ACPI Error: Region EmbeddedControl (ID=3) has no handler (20180531/exfldio-265) sep 13 14:57:31 SYSTEM kernel: ACPI Error: Method parse/execution failed \_SB.PCI0.LPCB.EC._REG, AE_NOT_EXIST (20180531/psparse-516) I recalled that it has been fixed in some time between 4.18-rc7/-rc8
I have tried (and currently running) "4.18.7-arch1-1-ARCH" The latest ArchLinux Kernel is "linux-4.18.8.arch1-1" (2018-09-19) I'm waiting for updates. When the updates are available I'll notify you (after try it). Thank you
@mvasi90, Do you have turbostat result when using 4.15? According to the logs of turbostat from 4.18 you pasted in Comment 2, the %busy field is around 70%+, which results in high cpu frequency IMO.
@Chen_Yu I don't have the kernel 4.15 ArchLinux is a full rolling-release distribution and doesn't survive without regular updates I'm very busy and I can't search for older kernel versions. Note: I never update the system messed up to avoid the accumulation of garbage. First: I'm restoring a clean backup for entire system, later I update/upgrade it Second: I'm removing all garbage like as previous kernel versions, etc Three: Full backup clean system and restore it to all my "computers/external disks, etc"
@mvasi90 Could you please run turbostat -i 5 --out turbostat_10.log and wait for 30 seconds, and provide the turbostat_10.log Do you have chance to run the latest kernel version?
Created attachment 279391 [details] turbostat_10.log The latest kernel version for ArchLinux is '4.18.16.arch1-1' and currently I'm using '4.18.14-arch1-1-ARCH'. This is no big difference; I'm trying to update every week but the version 4.19 still not available to this distribution. Note: To avoid overheating I disabled C-State on BIOS. Doing it the CPU Frequency in idle mode is 800MHz. I don't know if this have any impact on turbostat log for you. Note: turbostat installation gets 404 Error,to avoid it I updated repo list only without upgrade packages (sudo pacman -Sy and sudo pacman -S turbostat). I hope it is not a problem if the kernel version for which it has been compiled is newer. I'm working and I can't interrupt work to update entire system and reboot.
(In reply to Zhang Rui from comment #6) > what's the latest kernel you have tried? > can you please confirm if the problem still exists in the latest -rc, say, > 4.19-rc4? > > for this error > sep 13 14:57:31 SYSTEM kernel: ACPI Error: No handler for Region [EC__] > ((____ptrval____)) [EmbeddedControl] (20180531/evregion-132) > sep 13 14:57:31 SYSTEM kernel: ACPI Error: Region EmbeddedControl (ID=3) has > no handler (20180531/exfldio-265) > sep 13 14:57:31 SYSTEM kernel: ACPI Error: Method parse/execution failed > \_SB.PCI0.LPCB.EC._REG, AE_NOT_EXIST (20180531/psparse-516) > I recalled that it has been fixed in some time between 4.18-rc7/-rc8 Now I'm using the version 4.19.2-arch1-1-ARCH. The same ACPI error: [ +0,004109] ACPI Error: No handler for Region [EC__] (00000000a3d39ad4) [EmbeddedControl] (20180810/evregion-132) [ +0,000011] fbcon: Taking over console [ +0,000003] ACPI Error: Region EmbeddedControl (ID=3) has no handler (20180810/exfldio-265) [ +0,000006] ACPI Error: Method parse/execution failed \_SB.PCI0.LPCB.EC._REG, AE_NOT_EXIST (20180810/psparse-516) [ +0,002183] ACPI: [Firmware Bug]: BIOS _OSI(Linux) query ignored And the same behavior: When CPU C states are enabled (BIOS), the CPU frequency is very high 3.1 GHz. This did not happen with version 4.15 and earlier. More information: After fully system update/upgrade (sudo pacman -Syyuu), at each restart the following error appeared: [Firmware Bug]: TSC_DEADLINE disabled due to Errata; please update microcode to version: 0xb2 (or later) This was solved after BIOS update (from msi website). Note: I'm using intel-ucode but it was not enough without a BIOS update.
(In reply to mvasi90 from comment #12) > (In reply to Zhang Rui from comment #6) > > what's the latest kernel you have tried? > > can you please confirm if the problem still exists in the latest -rc, say, > > 4.19-rc4? > > > > for this error > > sep 13 14:57:31 SYSTEM kernel: ACPI Error: No handler for Region [EC__] > > ((____ptrval____)) [EmbeddedControl] (20180531/evregion-132) > > sep 13 14:57:31 SYSTEM kernel: ACPI Error: Region EmbeddedControl (ID=3) > has > > no handler (20180531/exfldio-265) > > sep 13 14:57:31 SYSTEM kernel: ACPI Error: Method parse/execution failed > > \_SB.PCI0.LPCB.EC._REG, AE_NOT_EXIST (20180531/psparse-516) > > I recalled that it has been fixed in some time between 4.18-rc7/-rc8 > > Now I'm using the version 4.19.2-arch1-1-ARCH. Thanks. > The same ACPI error: > > [ +0,004109] ACPI Error: No handler for Region [EC__] (00000000a3d39ad4) > [EmbeddedControl] (20180810/evregion-132) > [ +0,000011] fbcon: Taking over console > [ +0,000003] ACPI Error: Region EmbeddedControl (ID=3) has no handler > (20180810/exfldio-265) > [ +0,000006] ACPI Error: Method parse/execution failed > \_SB.PCI0.LPCB.EC._REG, AE_NOT_EXIST (20180810/psparse-516) @Rui, does it mean, since there's no ec handler, the system would behave insanely? > [ +0,002183] ACPI: [Firmware Bug]: BIOS _OSI(Linux) query ignored > > And the same behavior: > When CPU C states are enabled (BIOS), the CPU frequency is very high 3.1 GHz. > This did not happen with version 4.15 and earlier. > One of the recommend way is to use git bisect, although it might take a while to do that. > More information: > After fully system update/upgrade (sudo pacman -Syyuu), at each restart the > following error appeared: > [Firmware Bug]: TSC_DEADLINE disabled due to Errata; please update microcode > to version: 0xb2 (or later) > This reminds me of a tsc bug previously which causes high cpu busy utilization. Could you please provide(on latest kernel of cause): 1.turbostat -i 5 --out turbostat_4_19.log you can download the turbostat src from here: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/power/x86/turbostat?h=v4.20 and you can compile it manually. 2. top -b -d 5 > top_4_19.log 3. grep . /sys/devices/system/cpu/cpu*/cpuidle/state3/* > cstates_4_19.log > This was solved after BIOS update (from msi website). > Note: I'm using intel-ucode but it was not enough without a BIOS update.
(In reply to Chen Yu from comment #13) > (In reply to mvasi90 from comment #12) > > (In reply to Zhang Rui from comment #6) > > > what's the latest kernel you have tried? > > > can you please confirm if the problem still exists in the latest -rc, > say, > > > 4.19-rc4? > > > > > > for this error > > > sep 13 14:57:31 SYSTEM kernel: ACPI Error: No handler for Region [EC__] > > > ((____ptrval____)) [EmbeddedControl] (20180531/evregion-132) > > > sep 13 14:57:31 SYSTEM kernel: ACPI Error: Region EmbeddedControl (ID=3) > > has > > > no handler (20180531/exfldio-265) > > > sep 13 14:57:31 SYSTEM kernel: ACPI Error: Method parse/execution failed > > > \_SB.PCI0.LPCB.EC._REG, AE_NOT_EXIST (20180531/psparse-516) > > > I recalled that it has been fixed in some time between 4.18-rc7/-rc8 > > > > Now I'm using the version 4.19.2-arch1-1-ARCH. > Thanks. > > The same ACPI error: > > > > [ +0,004109] ACPI Error: No handler for Region [EC__] (00000000a3d39ad4) > > [EmbeddedControl] (20180810/evregion-132) > > [ +0,000011] fbcon: Taking over console > > [ +0,000003] ACPI Error: Region EmbeddedControl (ID=3) has no handler > > (20180810/exfldio-265) > > [ +0,000006] ACPI Error: Method parse/execution failed > > \_SB.PCI0.LPCB.EC._REG, AE_NOT_EXIST (20180810/psparse-516) > @Rui, does it mean, since there's no ec handler, the system would behave > insanely? > > [ +0,002183] ACPI: [Firmware Bug]: BIOS _OSI(Linux) query ignored > > > > And the same behavior: > > When CPU C states are enabled (BIOS), the CPU frequency is very high 3.1 > GHz. > > This did not happen with version 4.15 and earlier. > > > One of the recommend way is to use git bisect, although it might take a > while to do that. > > More information: > > After fully system update/upgrade (sudo pacman -Syyuu), at each restart the > > following error appeared: > > [Firmware Bug]: TSC_DEADLINE disabled due to Errata; please update > microcode > > to version: 0xb2 (or later) > > > This reminds me of a tsc bug previously which causes high cpu busy > utilization. > Could you please provide(on latest kernel of cause): > 1.turbostat -i 5 --out turbostat_4_19.log > you can download the turbostat src from here: > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/ > tools/power/x86/turbostat?h=v4.20 > and you can compile it manually. > 2. top -b -d 5 > top_4_19.log > 3. grep . /sys/devices/system/cpu/cpu*/cpuidle/state3/* > cstates_4_19.log Should be: grep . /sys/devices/system/cpu/cpu*/cpuidle/state*/* > cstates_4_19.log > > This was solved after BIOS update (from msi website). > > Note: I'm using intel-ucode but it was not enough without a BIOS update.
Created attachment 280333 [details] turbostat output
Created attachment 280335 [details] top output
Created attachment 280337 [details] cpuidle is missing
(In reply to mvasi90 from comment #17) > Created attachment 280337 [details] > cpuidle is missing The turbostat/top/cpuidle data is expected to be collected with c-state enabled in BIOS actually, which will be used to track the issue reported.
Created attachment 280789 [details] turbostat - 4.19 [CPU C state enabled]
Created attachment 280791 [details] top - 4.19 [CPU C state enabled]
Created attachment 280793 [details] cstates - 4.19 [CPU C state enabled]
I think this is probably caused by the wrong ECDT probe order issue. And the problem should have been fixed in 5.0. Bug closed, please feel free to reopen it if you can reproduce it in the latest upstream kernel