Verify failed on the version 2.27

[mr1jingles]

I have updated one of the servers on which it is installed
CPU: Model name: Intel(R) Xeon(R) Silver 4114 CPU @ 2.20GHz

And tuned-adm verify failed.

...
2026-03-25 16:21:41,115 INFO     tuned.plugins.plugin_cpu: We are running on an x86 GenuineIntel platform
2026-03-25 16:21:41,146 INFO     tuned.plugins.plugin_cpu: intel_pstate detected
...
2026-03-25 16:21:41,627 INFO     tuned.plugins.plugin_cpu: Setting no_turbo to '0' for intel_pstate driver (boost=1)
2026-03-25 16:21:41,627 INFO     tuned.plugins.plugin_cpu: Setting no_turbo to '0' for intel_pstate driver (boost=1)
...
2026-03-25 16:21:46,311 ERROR    tuned.plugins.base: verify: failed: device cpu36: 'boost' = 'None', expected '1'
2026-03-25 16:21:46,312 ERROR    tuned.plugins.base: verify: failed: device cpu11: 'boost' = 'None', expected '1'

After downgrade to version 2.26, evetything is ok

# tuned -v
tuned 2.26.0
# tuned-adm verify
Verification succeeded, current system settings match the preset profile.
See TuneD log file ('/var/log/tuned/tuned.log') for details.

[rsp4jack]

Happens to me. "/sys/devices/system/cpu/cpufreq/policy*/boost" matches nothing on my machine, but no_turbo exists.

[bvngee]

More refined bug report:

Afaict, #797 did not correctly fix the intel_pstate boost support. TuneD currently expects that both the cpufreq boost file AND the intel_pstate no_turbo files are created and updated, whereas only the no_turbo file is create when using intel_pstate. Straight from the kernel cpufreq docs:

The boost File in sysfs

This file is located under /sys/devices/system/cpu/cpufreq/ and controls the “boost” setting for the whole system. It is not present if the underlying scaling driver does not support the frequency boost mechanism (or supports it, but provides a driver-specific interface for controlling it, like intel_pstate).

In other words, log.debug("boost file missing, which can happen on pre 6.11 kernels.") is incorrect, as the boost file will not be present with intel_pstate in use.

Note that TuneD does correctly set intel_pstate's no_turbo option, it's just the verify command fails (afaik). Perhaps something like this is needed?

diff --git a/plugin_cpu.py b/plugin_cpu.py
--- a/plugin_cpu.py
+++ b/plugin_cpu.py
@@ -784,12 +784,9 @@ class CPULatencyPlugin(hotplug.Plugin):
 					boost_set = True
 				else:
 					log.error("Failed to set boost on cpu '%s'. Is the value in the profile correct?" % device)
-		else:
-			log.debug("boost file missing, which can happen on pre 6.11 kernels.")
-
 		# For Intel systems with intel_pstate driver, handle no_turbo
 		# This ensures boost=0 works properly on Intel CPUs
-		if self._has_intel_pstate and boost in ["0", "1"]:
+		elif self._has_intel_pstate and boost in ["0", "1"]:
 			no_turbo_val = "1" if boost == "0" else "0"
 
 			if not sim:
@@ -815,8 +812,12 @@ class CPULatencyPlugin(hotplug.Plugin):
 		cpu_id = device.lstrip("cpu")
 		if os.path.exists(self._pstate_boost_path(cpu_id)):
 			return self._cmd.read_file(self._pstate_boost_path(cpu_id)).strip()
+		elif self._has_intel_pstate:
+			no_turbo = self._get_intel_pstate_attr("no_turbo")
+			if no_turbo is not None:
+				return no_turbo
 		else:
-			log.debug("boost file missing, which can happen on pre 6.11 kernels.")
+			log.warning("Unable to get boost on cpu '%s'. Neither per-policy boost nor intel_pstate no_turbo is available." % device)
 		return None
 
 	@command_set("energy_performance_preference", per_device=True)

Pinging @sunalawa who made the initial change and @yarda who merged it. Unfortunately I don't have time atm to test this, but hopefully it helps somehow!