Hi Len,
find attached fix to avoid exiting with -13 on Zen. Patch is against turbostat as included in Linux-5.10.2.
Please merge.
PS: This is probably material for -stable, as it used to work before on Zen (Zen2 aka Ryzen 3000 in my case).
--
Kurt Garloff <[email protected]>
Cologne, Germany
Hi Len,
Issue persists on Ryzen in 5.11-rc6:
kvmadmin@KurtSrv2018(//):~ [0]$ sudo /casa/src/linux-stable/tools/power/x86/turbostat/turbostat
turbostat version 20.09.30 - Len Brown <[email protected]>
CPUID(0): AuthenticAMD 0x10 CPUID levels; 0x80000023 xlevels; family:model:stepping 0x19:21:0 (25:33:0)
CPUID(1): SSE3 MONITOR - - - TSC MSR - HT -
CPUID(6): APERF, No-TURBO, No-DTS, No-PTM, No-HWP, No-HWPnotify, No-HWPwindow, No-HWPepp, No-HWPpkg, No-EPB
CPUID(7): No-SGX
RAPL: 234 sec. Joule Counter Range, at 280 Watts
/dev/cpu_dma_latency: 2000000000 usec (default)
current_driver: acpi_idle
current_governor: menu
current_governor_ro: menu
cpu22: POLL: CPUIDLE CORE POLL IDLE
cpu22: C1: ACPI FFH MWAIT 0x0
cpu22: C2: ACPI IOPORT 0x414
cpu22: cpufreq driver: acpi-cpufreq
cpu22: cpufreq governor: schedutil
cpufreq boost: 1
cpu0: MSR_RAPL_PWR_UNIT: 0x000a1003 (0.125000 Watts, 0.000015 Joules, 0.000977 sec.)
kvmadmin@KurtSrv2018(//):~ [243]$
^^^ Exit code
With the patch:
kvmadmin@KurtSrv2018(//):~ [243]$ sudo /casa/src/linux-stable/tools/power/x86/turbostat/turbostat
turbostat version 20.09.30 - Len Brown <[email protected]>
CPUID(0): AuthenticAMD 0x10 CPUID levels; 0x80000023 xlevels; family:model:stepping 0x19:21:0 (25:33:0)
CPUID(1): SSE3 MONITOR - - - TSC MSR - HT -
CPUID(6): APERF, No-TURBO, No-DTS, No-PTM, No-HWP, No-HWPnotify, No-HWPwindow, No-HWPepp, No-HWPpkg, No-EPB
CPUID(7): No-SGX
RAPL: 234 sec. Joule Counter Range, at 280 Watts
/dev/cpu_dma_latency: 2000000000 usec (default)
current_driver: acpi_idle
current_governor: menu
current_governor_ro: menu
cpu28: POLL: CPUIDLE CORE POLL IDLE
cpu28: C1: ACPI FFH MWAIT 0x0
cpu28: C2: ACPI IOPORT 0x414
cpu28: cpufreq driver: acpi-cpufreq
cpu28: cpufreq governor: schedutil
cpufreq boost: 1
cpu0: MSR_RAPL_PWR_UNIT: 0x000a1003 (0.125000 Watts, 0.000015 Joules, 0.000977 sec.)
Core CPU Avg_MHz Busy% Bzy_MHz TSC_MHz IRQ POLL C1 C2 POLL% C1% C2% CorWatt PkgWatt
- - 27 1.04 2562 3411 16046 33 2931 12895 0.00 0.85 98.48 1.57 18.81
0 0 12 0.55 2193 3400 885 1 111 757 0.00 1.12 98.42 0.04 18.74
0 16 1 0.05 2351 3400 53 0 3 54 0.00 0.05 99.92
1 1 20 0.89 2261 3400 478 0 39 427 0.00 0.37 98.80 0.06
1 17 9 0.40 2329 3400 308 0 38 282 0.00 0.35 99.29
[...]
--
Kurt Garloff <[email protected]>
Cologne, Germany
On 26/12/2020 13:13, Kurt Garloff wrote:
> Hi Len,
>
> find attached fix to avoid exiting with -13 on Zen. Patch is against turbostat as included in Linux-5.10.2.
> Please merge.
>
> PS: This is probably material for -stable, as it used to work before on Zen (Zen2 aka Ryzen 3000 in my case).
>
commit 5d399d05df42ffcaa2b3836b580631c4024487a0
Author: Kurt Garloff <[email protected]>
Date: Mon Feb 1 09:01:47 2021 +0000
turbostat: Fix Pkg Power tracking on Zen
AMD Zen processors use a different MSR (MSR_PKG_ENERGY_STAT) than intel
(MSR_PKG_ENERGY_STATUS) to track package power; however we want to record
it at the same offset in our package_data.
offset_to_idx() however only recognized the intel MSR, erroring
out with -13 on Zen.
With this fix, it will support the Zen MSR.
Tested successfully on Ryzen 3000 & 5000.
Signed-off-by: Kurt Garloff <[email protected]>
diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c
index 389ea5209a83..cb830e73d899 100644
--- a/tools/power/x86/turbostat/turbostat.c
+++ b/tools/power/x86/turbostat/turbostat.c
@@ -325,6 +325,7 @@ int offset_to_idx(int offset)
int idx;
switch (offset) {
+ case MSR_PKG_ENERGY_STAT:
case MSR_PKG_ENERGY_STATUS:
idx = IDX_PKG_ENERGY;
break;
--
Kurt Garloff <[email protected]>
Cologne, Germany
Hi,
On 01.02.21 10:01, Kurt Garloff wrote:
> Issue persists on Ryzen in 5.11-rc6:
>
> kvmadmin@KurtSrv2018(//):~ [0]$ sudo /casa/src/linux-stable/tools/power/x86/turbostat/turbostat
> [...]
> kvmadmin@KurtSrv2018(//):~ [243]$
>
> ^^^ Exit code
>
> With the patch:
>
> kvmadmin@KurtSrv2018(//):~ [243]$ sudo /casa/src/linux-stable/tools/power/x86/turbostat/turbostat
> [...]
> Core CPU Avg_MHz Busy% Bzy_MHz TSC_MHz IRQ POLL C1 C2 POLL% C1% C2% CorWatt PkgWatt
> - - 27 1.04 2562 3411 16046 33 2931 12895 0.00 0.85 98.48 1.57 18.81
> 0 0 12 0.55 2193 3400 885 1 111 757 0.00 1.12 98.42 0.04 18.74
> 0 16 1 0.05 2351 3400 53 0 3 54 0.00 0.05 99.92
> 1 1 20 0.89 2261 3400 478 0 39 427 0.00 0.37 98.80 0.06
> 1 17 9 0.40 2329 3400 308 0 38 282 0.00 0.35 99.29
> [...]
I was seeing the same issue (no stats, program just exits with 243), and
Kurt's simple patch resolved it for me.
Best,
Christian
On Wed, Mar 24, 2021 at 5:38 AM Salvatore Bonaccorso <[email protected]> wrote:
> On Mon, Mar 15, 2021 at 10:54:24PM +0100, Christian Kastner wrote:
> > On 01.02.21 10:01, Kurt Garloff wrote:
> > > Issue persists on Ryzen in 5.11-rc6:
> > > kvmadmin@KurtSrv2018(//):~ [0]$ sudo /casa/src/linux-stable/tools/power/x86/turbostat/turbostat
... deleted stuff that doesn't display properly ...
> > I was seeing the same issue (no stats, program just exits with 243), and
> > Kurt's simple patch resolved it for me.
>
> Does Kurt's patch seems good to you and can be applied or is there
> anything missing?
There were multiple patch submissions. Chen Yu kindly merged them into
one, which was put out for testing a couple of weeks ago. Try it and
report back. I'll forward it in a moment.
Hi Len,
On Mon, Mar 15, 2021 at 10:54:24PM +0100, Christian Kastner wrote:
> Hi,
>
> On 01.02.21 10:01, Kurt Garloff wrote:
> > Issue persists on Ryzen in 5.11-rc6:
> >
> > kvmadmin@KurtSrv2018(//):~ [0]$ sudo /casa/src/linux-stable/tools/power/x86/turbostat/turbostat
> > [...]
> > kvmadmin@KurtSrv2018(//):~ [243]$
> >
> > ?????????????????????????????????????????????????????? ^^^ Exit code
> >
> > With the patch:
> >
> > kvmadmin@KurtSrv2018(//):~ [243]$ sudo /casa/src/linux-stable/tools/power/x86/turbostat/turbostat??????
> > [...]????????????????????????????????????
> > Core?????? CPU???????? Avg_MHz Busy%???? Bzy_MHz TSC_MHz IRQ???????? POLL?????? C1?????????? C2?????????? POLL%???? C1%???????? C2%???????? CorWatt PkgWatt
> > -???????????? -???????????? 27?????????? 1.04?????? 2562?????? 3411?????? 16046???? 33?????????? 2931?????? 12895???? 0.00?????? 0.85?????? 98.48???? 1.57?????? 18.81
> > 0???????????? 0???????????? 12?????????? 0.55?????? 2193?????? 3400?????? 885???????? 1???????????? 111???????? 757???????? 0.00?????? 1.12?????? 98.42???? 0.04?????? 18.74
> > 0???????????? 16?????????? 1???????????? 0.05?????? 2351?????? 3400?????? 53?????????? 0???????????? 3???????????? 54?????????? 0.00?????? 0.05?????? 99.92????????????
> > 1???????????? 1???????????? 20?????????? 0.89?????? 2261?????? 3400?????? 478???????? 0???????????? 39?????????? 427???????? 0.00?????? 0.37?????? 98.80???? 0.06
> > 1???????????? 17?????????? 9???????????? 0.40?????? 2329?????? 3400?????? 308???????? 0???????????? 38?????????? 282???????? 0.00?????? 0.35?????? 99.29????????????
> > [...]
>
> I was seeing the same issue (no stats, program just exits with 243), and
> Kurt's simple patch resolved it for me.
Does Kurt's patch seems good to you and can be applied or is there
anything missing?
Regards,
Salvatore