Support up to 8192 processors
Add cpuidle governor debug telemetry, disabled by default
Update default output to exclude cpuidle invocation counts
Bug fixes
Signed-off-by: Len Brown <len.brown@intel.com>
Create "pct_idle" counter group, the sofware notion of residency
so it can now be singled out, independent of other counter groups.
Create "cpuidle" group, the cpuidle invocation counts.
Disable "cpuidle", by default.
Create "swidle" = "cpuidle" + "pct_idle".
Undocument "sysfs", the old name for "swidle", but keep it working
for backwards compatibilty.
Create "hwidle", all the HW idle counters
Modify "idle", enabled by default
"idle" = "hwidle" + "pct_idle" (and now excludes "cpuidle")
Signed-off-by: Len Brown <len.brown@intel.com>
Probe cpuidle "sysfs" residency and counts separately,
since soon we will make one disabled on, and the
other disabled off.
Clarify that some BIC (build-in-counters) are actually "groups".
since we're about to re-name some of those groups.
no functional change.
Signed-off-by: Len Brown <len.brown@intel.com>
Do fflush() to discard the buffered data, before each read of the
graphics sysfs knobs.
Fixes: ba99a4fc8c ("tools/power turbostat: Remove unnecessary fflush() call")
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
The CoreThr column displays total thermal throttling events
since boot time.
Change it to report events during the measurement interval.
This is more useful for showing a user the current conditions.
Total events since boot time are still available to the user via
/sys/devices/system/cpu/cpu*/thermal_throttle/*
Document CoreThr on turbostat.8
Fixes: eae97e053f ("turbostat: Support thermal throttle count print")
Reported-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Cc: Chen Yu <yu.c.chen@intel.com>
On systems with >= 1024 cpus (in my case 1152), turbostat fails with the error output:
"turbostat: /sys/fs/cgroup/cpuset.cpus.effective: cpu str malformat 0-1151"
A similar error appears with the use of turbostat --cpu when the inputted cpu
range contains a cpu number >= 1024:
# turbostat -c 1100-1151
"--cpu 1100-1151" malformed
...
Both errors are caused by parse_cpu_str() reaching its limit of CPU_SUBSET_MAXCPUS.
It's a good idea to limit the maximum cpu number being parsed, but 1024 is too low.
For a small increase in compute and allocated memory, increasing CPU_SUBSET_MAXCPUS
brings support for parsing cpu numbers >= 1024.
Increase CPU_SUBSET_MAXCPUS to 8192, a common setting for CONFIG_NR_CPUS on x86_64.
Signed-off-by: Justin Ernst <justin.ernst@hpe.com>
Signed-off-by: Len Brown <len.brown@intel.com>
The idle governor provides the following per-idle state sysfs files:
* above - Indicates overshoots, where a more shallow state should have
been requested (if avaliale and enabled).
* below - Indicates undershoots, where a deeper state should have been
requested (if available and enabled).
These files offer valuable insights into how effectively the Linux kernel
idle governor selects idle states for a given workload. This commit adds
support for these files in turbostat.
Expose the contents of these files with the following naming convention:
* C1: The number of times the C1 state was requested (existing counter).
* C1+: The number of times the idle governor selected C1, but a deeper
idle state should have been selected instead.
* C1-: The number of times the idle governor selected C1, but a shallower
idle state should have been selected instead.
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Fix the 'find_msrp_by_name()' function which returns incorrect matches for
cases like this:
s1 = "C1-";
find_msrp_by_name(head, s1);
Inside 'find_msrp_by_name()':
...
s2 = "C1"
if !(strcnmp(s1, s2, len(s2)))
// Incorrect match!
return mp;
Full strings should be match istead. Switch to 'strcmp()' to fix the problem.
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
turbostat aborted with below messages on a dual-package system,
turbostat: turbostat.c:3744: rapl_counter_accumulate: Assertion `dst->unit == src->unit' failed.
Aborted
This is because
1. the MSR_DRAM_PERF_STATUS returns Zero for one package, and non-Zero
for another package
2. probe_msr() treats Zero return value as a failure so this feature is
enabled on one package, and disabled for another package.
3. turbostat aborts because the feature is invalid on some package
Unlike the RAPL energy counter registers, MSR_DRAM_PERF_STATUS can
return Zero value, and this should not be treated as a failure.
Fix the problem by allowing Zero return value for RAPL registers other
than the energy counters.
Fixes: 7c6fee25bd ("tools/power turbostat: Check for non-zero value when MSR probing")
Reported-by: Artem Bityutskiy <artem.bityutskiy@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
The clustered uncore frequency counters, UMHz*.*
should honor the --show and --hide options.
All non-specified counters should be implicityly hidden.
But when --show was used, UMHz*.* showed up anyway:
$ sudo turbostat -q -S --show Busy%
Busy% UMHz0.0 UMHz1.0 UMHz2.0 UMHz3.0 UMHz4.0
Indeed, there was no string that can be used to explicitly
show or hide clustered uncore counters.
Even through they are dynamically probed and added,
group the clustered UMHz*.* counters with the legacy
built-in-counter "UncMHz" for show/hide.
turbostat --show Busy%
does not show UMHz*.*.
turbostat --show UncMHz
shows either UncMHz or UMHz*.*, if present
turbostat --hide UncMHz
hides either UncMHz or UMHz*.*, if present
Reported-by: Artem Bityutskiy <artem.bityutskiy@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Tested-by: Artem Bityutskiy <artem.bityutskiy@intel.com>
This CPU was mistakenly given the name INTEL_ATOM_AIRMONT_MID. But it
uses a Silvermont core, not Airmont.
Change #define name to INTEL_ATOM_SILVERMONT_MID2
Reported-by: Christian Ludloff <ludloff@gmail.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Link: https://lore.kernel.org/all/20241007165701.19693-1-tony.luck%40intel.com
Summary of Changes since 2024.11.30:
Fix regression in 2023.11.07 that affinitized forked child
in one-shot mode.
Harden one-shot mode against hotplug online/offline
Enable RAPL SysWatt column by default.
Add initial PTL, CWF platform support.
Harden initial PMT code in response to early use.
Enable first built-in PMT counter: CWF c1e residency
Refuse to run on unsupported platforms without --force,
to encourage updating to a version that supports the system,
and to avoid no-so-useful measurement results.
Signed-off-by: Len Brown <len.brown@intel.com>
Intel Clearwater Forest report PMT telemetry with GUID 0x14421519, which
can be used to obtain module c1e residency counter of type tcore clock.
Add early support for the counter by using heuristic that should work
for the Clearwater Forest platforms.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
when turbostat interval mode can't migrate to a CPU, it complains,
prints no data, re-initializes with the new CPU configuration
and starts a new interval.
But this strategy in the face of a CPU hotplug offline during an interval
doesn't help in one-shot mode. When the missing CPU is discovered
at the end of the interval, the forked program has already returned
and there is nothing left for a new interval to measure.
So instead of aborting get_coutners() and delta_cpu() if a missing CPU
is detected, complain, but carry on and output what statistics are
actually present.
Use the same strategy for delta_cpu when aperf:mperf are observed
to have been reset -- complain, but carry on and print data for
the CPUs that are still present.
Interval mode error handling is unchanged.
One-shot mode can now do this:
$ sudo chcpu -e 1 ; sudo ./turbostat --quiet --show PkgWatt,Busy%,CPU chcpu -d 1
CPU 1 enabled
CPU 1 disabled
get_counters: Could not migrate to CPU 1
./turbostat: Counter reset detected
0.036920 sec
CPU Busy% PkgWatt
- 0.00 10.00
0 99.73 10.00
1 0.00
2 91.53
3 16.83
Suggested-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
In "one-shot" mode, turbostat
1. takes a counter snapshot
2. forks and waits for a child
3. takes the end counter snapshot and prints the result.
But turbostat counter snapshots currently use affinity to travel
around the system so that counter reads are "local", and this
affinity must be cleared between #1 and #2 above.
The offending commit removed that reset that allowed the child
to run on cpu_present_set.
Fix that issue, and improve upon the original by using
cpu_possible_set for the child. This allows the child
to also run on CPUs that hotplug online during its runtime.
Reported-by: Zhang Rui <rui.zhang@intel.com>
Fixes: 7bb3fe27ad ("tools/power/turbostat: Obey allowed CPUs during startup")
Signed-off-by: Len Brown <len.brown@intel.com>
Some PMT counters, for example module c1e residency on Intel Clearwater
Forest, are reported using tcore clock type.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Fix checkpatch whitespace issues since 2024.11.30
Summary of Changes since 2024.11.30:
Enable SysWatt by default.
Add initial PTL, CWF platform support.
Refuse to run on unsupported platforms without --force
to avoid not-so-useful measurements mistakenly made
using obsolete versions.
Harden initial PMT code in response to early use.
Signed-off-by: Len Brown <len.brown@intel.com>
Allow user to add PMT counters by either identifying the source with:
guid=%u,seq=%u
or, since this patch, with direct sysfs path:
path=%s, for example path=/sys/class/intel_pmt/telem5
In the later case, the guid and sequence number will be infered
by turbostat.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Some platforms may expose multiple telemetry files identified with the
same GUID. Interpreting it correctly, to associate given counter with a
CPU, core or a package requires more metadata from the user.
Parse and create ordered, linked list of those PMT aggregators, so that
we can identify specific aggregator with GUID + sequence number.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
PMT directories exposed in sysfs use the following pattern:
telem%u
for example:
telem0, telem2, telem3, ..., telem15, telem16
This naming scheme preserves the ordering from the PCIe discovery, which
is important to correctly map the telemetry directory to the specific
domain (cpu, core, package etc).
Because readdir() traverses the entries in alphabetical order, causing
for example "telem13" to be traversed before "telem3", it is necessary
to use scandir() with custom compare() callback to preserve the PCIe
ordering.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
When platforms expose multiple PMT aggregators with the same GUID, the
only way to identify them and map to specific domain is by reading them
in an order they were exposed via PCIe. Intel PMT kernel driver does
keep the same order and numbers the telemetry directories accordingly.
Use GUID and sequence number (order) to uniquely identify PMT
aggregators.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
When requesting PMT counters with --add command, user may want to skip
specifying values for all the domains (that is, cpu, core, package etc).
For the domains that user did not provide information on how to read the
counter, return default value - zero.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
For some MSRs, for example, the Platform Energy Counter (RAPL PSYS), it
is required to additionally check for a non-zero value to confirm that
it is present.
From Intel SDM vol. 4:
Platform Energy Counter (R/O)
This MSR is valid only if both platform vendor hardware
implementation and BIOS enablement support it.
This MSR will read 0 if not valid.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Include procfs and sysfs data collection time in the system summary
row of the "usec" column. This is useful for isolating where the
time goes during turbostat data collection.
Background:
Column "usec" shows
1. the number of microseconds elapsed during counter collection,
including thread migration -- if any, for each CPU row.
2. total elapsed time to collect the counters on all cpus, for the
summary row.
This can be used to check the time cost of a give column. For example,
run below commands separately
turbostat --show usec sleep 1
turbostat --show usec,CoreTmp sleep 1
and the delta in the usec column will tell the time cost for CoreTmp
(Thermal MSR read)
Problem:
Some of the kernel procfs/sysfs accesses are expensive, especially on
high core count systems. "usec" column cannot tell this because it only
includes the time cost of the counters.
Solution:
Leave the per CPU "usec" as it is and modify the summary "usec" to
include the time cost of the procfs/sysfs snapshot.
With it, the "usec" column can be used to get
1. the baseline, e.g.
turbostat --show usec sleep 1
2. the baseline + some per CPU counter cost, e.g.
turbostat --show usec,CoreTmp sleep 1
3. the baseline + some per CPU sysfs cost, e.g.
turbostat --show usec,C1 sleep 1
4. the baseline + /proc/interrupts cost, e.g
turbostat --show usec,IRQ sleep 1
Man-page update is also included.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Intel Sapphire Rapids is an exception and has fixed divisor for RAPL PSYS
counter set to 1.0. Add a platform bit and enable it for SPR.
Reported-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
This (the old code) is just not how you round up to a page size.
Noticed on a recent Intel platform. Previous ones must have been
reporting sizes already aligned to a page and so the bug was missed when
testing.
Fixes: f0e4ed752f ("tools/power turbostat: Add early support for PMT counters")
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
The counter is present on most supporting Intel platforms and provides
useful data to the user. There is no reason to disable the counter by
default.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Add an NMI column, a proper sub-set of the IRQ column.
It would be preferable if the kernel exported
/sys/kernel/irq/NMI/per_cpu_count.
But since we are already forced to parse /proc/interrupts,
noticing which row is the NMI is simple enough.
Suggested-by: Artem Bityutskiy <artem.bityutskiy@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Turbostat currently exits under the following conditions:
1. When running on non-Intel/AMD/Hygon x86 vendors.
2. When running on Intel models that lack specific platform features.
Introduce a new `--force` parameter that allows turbostat to run on
these unsupported platforms with minimal default feature support. This
provides users with the flexibility to gather basic information even on
unsupported systems.
[lenb: updated warning message text]
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Improve the `--help` output of turbostat by standardizing the format
and enhancing readability. The following changes are made to ensure
consistency and clarity in the help message:
1. Use a consistent pattern for each parameter's help message:
- Display the parameter and its input (if any) on the same line,
separated by a space.
- Provide the detailed description on a separate line.
2. Ensure that the first character of each description is in lower-case.
These changes make the help output more uniform and easier to read,
helping users quickly understand the available options and their usage.
No functional change.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Turbostat currently supports x86 processors from Intel, AMD, and Hygon.
The behavior of turbostat on CPUs from other vendors has not been
evaluated and may lead to incorrect or undefined behavior.
Enhance turbostat to exit by default when running on an unsupported CPU
vendor. This ensures that users are aware that their CPU is not
currently supported by turbostat, guiding them to seek support for their
specific hardware through future patches.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Turbostat requires per-platform enabling for Intel CPU models due to
platform-specific features. When running on unsupported Intel CPU
models, turbostat currently operates with limited default features,
which can lead to users unknowingly using an outdated version of the
tool.
Enhance turbostat to exit by default when run on unsupported Intel CPU
models, with a clear message to users, informing them that their CPU
model is not supported and advising them to update to the latest version
of turbostat for full functionality.
[lenb: updated error message wording]
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Add initial support for ClearwaterForest.
It shares the same features with SierraForest.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Add initial support for PantherLake.
It shares the same features with Lunarlake.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
since 2024.07.26:
assorted minor bug fixes
assorted platform specific tweaks
initial RAPL PSYS (SysWatt) support
Signed-off-by: Len Brown <len.brown@intel.com>
Introduce the counter as a part of global, platform counters structure.
We open the counter for only one cpu, but otherwise treat it as an
ordinary RAPL counter, allowing for grouped perf read.
The counter is disabled by default, because it's interpretation may
require additional, platform specific information, making it unsuitable
for general use.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Add '+' to optstring when early scanning for --no-msr and --no-perf.
It causes option processing to stop as soon as a nonoption argument is
encountered, effectively skipping child's arguments.
Fixes: 3e4048466c ("tools/power turbostat: Add --no-msr option")
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Force the --no-perf early to prevent using it as a source. User asks for
raw values, but perf returns them relative to the opening of the file
descriptor.
Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
On some machines, the graphics device is enumerated as
/sys/class/drm/card1 instead of /sys/class/drm/card0. The current
implementation does not handle this scenario, resulting in the loss of
graphics C6 residency and frequency information.
Add support for /sys/class/drm/card1, ensuring that turbostat can
retrieve and display the graphics columns for these platforms.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Snapshots of the graphics sysfs knobs are taken based on file
descriptors. To optimize this process, open the files and cache the file
descriptors during the graphics probe phase. As a result, the previously
cached pathnames become redundant and are removed.
This change aims to streamline the code without altering its functionality.
No functional change intended.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Currently, there is an inconsistency in how graphics sysfs knobs are
accessed: graphics residency sysfs knobs are opened and closed for each
read, while graphics frequency sysfs knobs are opened once and remain
open until turbostat exits. This inconsistency is confusing and adds
unnecessary code complexity.
Consolidate the access method by opening the sysfs files once and
reusing the file pointers for subsequent accesses. This approach
simplifies the code and ensures a consistent method for accessing
graphics sysfs knobs.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
The graphics sysfs knobs are read-only, making the use of fflush()
before reading them redundant.
Remove the unnecessary fflush() call.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
In various generations, platforms often share a majority of features,
diverging only in a few specific aspects. The current approach of using
hardcoded values in 'platform_features' structure fails to effectively
represent these divergences.
To improve the description of platform divergence:
1. Each newly introduced 'platform_features' structure must have a base,
typically derived from the previous generation.
2. Platform feature values should be inherited from the base structure
rather than being hardcoded.
This approach ensures a more accurate and maintainable representation of
platform-specific features across different generations.
Converts `adl_features` and `lnl_features` to follow this new scheme.
No functional change.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Add initial support for GraniteRapids-D. It shares the same features
with SapphireRapids.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Lunarlake supports CC1/CC6/CC7/PC2/PC6/PC10.
Remove PC3 support on Lunarlake.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
As ARL shares the same features with ADL/RPL/MTL, now 'arl_features' is
used by Lunarlake platform only.
Rename 'arl_features' to 'lnl_features'.
No functional change.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Similar to ADL/RPL/MTL, ARL supports CC1/CC6/CC7/PC2/PC3/PC6/PC8/PC10.
Add back PC8 support on Arrowlake.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Similar to ADL/RPL, MTL support CC1/CC6/CC7/PC2/PC3/PC6/PC8/CP10.
Remove PC7/PC9 support on MTL.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Linus Torvalds
Len Brown
Zhang Rui
Justin Ernst
Artem Bityutskiy
Tony Luck
Patryk Wlazlyn