client: fixed a bug where AMD CPUs were not correctly fingerprinting base speed

[mvegter]

Relates to: #19468

Description

Within one of our production sites we are running on a mixed server fleet, where some are running baremetal Nomad and others are part of the virtualised fleet. We noticed a big difference in the reported CPU . On the baremetal systems they were using the boost clock at 3.5 GHz whereas on the virtualised load they were using the QEMU default for dmidecode at 2GHz

nomad version
Nomad v1.8.2
BuildDate 2024-07-16T08:50:09Z
Revision 7f0822c1e4f25907d9f60e2d595411950dd1bd28

Testing & Reproduction steps

grep -H . /sys/devices/system/cpu/cpu0/cpufreq/*
/sys/devices/system/cpu/cpu0/cpufreq/affected_cpus:0
/sys/devices/system/cpu/cpu0/cpufreq/bios_limit:2450000
/sys/devices/system/cpu/cpu0/cpufreq/cpb:1
/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq:2450000
/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq:3529052
/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_min_freq:1500000
/sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_transition_latency:0
/sys/devices/system/cpu/cpu0/cpufreq/freqdomain_cpus:0 128
/sys/devices/system/cpu/cpu0/cpufreq/related_cpus:0
/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies:2450000 2000000 1500000
/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors:conservative ondemand userspace powersave performance schedutil
/sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq:3241647
/sys/devices/system/cpu/cpu0/cpufreq/scaling_driver:acpi-cpufreq
/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor:performance
/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq:3529052
/sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq:1500000
/sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed:<unsupported>

grep -H . /sys/devices/system/cpu/cpu0/acpi_cppc/*
/sys/devices/system/cpu/cpu0/acpi_cppc/feedback_ctrs:ref:4782881970238275 del:6173496996912339
/sys/devices/system/cpu/cpu0/acpi_cppc/highest_perf:255
/sys/devices/system/cpu/cpu0/acpi_cppc/lowest_freq:400
/sys/devices/system/cpu/cpu0/acpi_cppc/lowest_nonlinear_perf:109
/sys/devices/system/cpu/cpu0/acpi_cppc/lowest_perf:29
/sys/devices/system/cpu/cpu0/acpi_cppc/nominal_freq:2451
/sys/devices/system/cpu/cpu0/acpi_cppc/nominal_perf:177
/sys/devices/system/cpu/cpu0/acpi_cppc/reference_perf:177
/sys/devices/system/cpu/cpu0/acpi_cppc/wraparound_time:18446744073709551615

Quick search through our inventory :

AMD EPYC 7313 16-Core Processor
/sys/devices/system/cpu/cpu0/acpi_cppc/nominal_freq:3001

AMD EPYC 7713 64-Core Processor
/sys/devices/system/cpu/cpu0/acpi_cppc/nominal_freq:2000

AMD EPYC 7742 64-Core Processor
/sys/devices/system/cpu/cpu0/acpi_cppc/nominal_freq:2251

AMD EPYC 7763 64-Core Processor
/sys/devices/system/cpu/cpu0/acpi_cppc/nominal_freq:2451

Links

• https://docs.kernel.org/admin-guide/pm/intel_pstate.html#interpretation-of-policy-attributes
• https://docs.kernel.org/admin-guide/acpi/cppc_sysfs.html

Contributor Checklist

• Changelog Entry If this PR changes user-facing behavior, please generate and add a
  changelog entry using the make cl command.
• Testing Please add tests to cover any new functionality or to demonstrate bug fixes and
  ensure regressions will be caught.
• Documentation If the change impacts user-facing functionality such as the CLI, API, UI,
  and job configuration, please update the Nomad website documentation to reflect this. Refer to
  the website README for docs guidelines. Please also consider whether the
  change requires notes within the upgrade guide.

Reviewer Checklist

• Backport Labels Please add the correct backport labels as described by the internal
  backporting document.
• Commit Type Ensure the correct merge method is selected which should be "squash and merge"
  in the majority of situations. The main exceptions are long-lived feature branches or merges where
  history should be preserved.
• Enterprise PRs If this is an enterprise only PR, please add any required changelog entry
  within the public repository.