Add SPDX identifier to the gitignore. Remove the comment and .i file
since the file it references was removed in another patch. This patch
depends on Min-Hua Chen's 'pm: cpupower: rename raw_pylibcpupower.i'.

Signed-off-by: John B. Wyatt IV <jwyatt@redhat.com>
Signed-off-by: John B. Wyatt IV <sageofredondo@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

The raw_pylibcpupower.i is removed unexpectedly after 'make mrproper'

We can reproduce the error by performing the following steps:
cd linux-next
make mrproper
cd tools/power/cpupower/bindings/python
make

We will get an error message:
make: *** No rule to make target 'raw_pylibcpupower.i', needed by 'raw_pylibcpupower_wrap.c'.  Stop.

The root cause:

The *.i files are already used for pre-processor output files and
the kernel removes all the *.i files by 'make mrproper'.

That explains why the raw_pylibcpupower.i is removed by 'make mrproper'.

To fix it, Follow John's suggestion to rename raw_pylibcpupower.i to
raw_pylibcpupower.swg.

See:
https://www.swig.org/Doc4.2/SWIG.html



Reviewed-by: John B. Wyatt IV <jwyatt@redhat.com>
Reviewed-by: John B. Wyatt IV <sageofredondo@gmail.com>
Tested-by: John B. Wyatt IV <jwyatt@redhat.com>
Tested-by: John B. Wyatt IV <sageofredondo@gmail.com>
Signed-off-by: Min-Hua Chen <minhuadotchen@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

Add error message to better explain to the user when SWIG and
python-config is missing from the path. Makefile was cleaned up
and unneeded elements were removed.

Suggested-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: John B. Wyatt IV <jwyatt@redhat.com>
Signed-off-by: John B. Wyatt IV <sageofredondo@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

This script demonstrates how to make use of, and tests, the bindings.

In the future, this script could become part of a larger test suite to
test the bindings and libcpupower.

Signed-off-by: John B. Wyatt IV <jwyatt@redhat.com>
Signed-off-by: John B. Wyatt IV <sageofredondo@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

SWIG is a tool packaged in Fedora and other distros that can generate
bindings from C and C++ code for several languages including Python,
Perl, and Go.

These bindings allows users to easily write scripts that use and extend
libcpupower's functionality. Currently, only Python is provided in the
makefile, but additional languages may be added if there is demand.

Added suggestions from Shuah Khan for the README and license discussion.

Note that while SWIG itself is GPL v3+ licensed; the resulting output,
the bindings code, is permissively licensed + the license of the .o
files. Please see
https://swig.org/legal.html and [1] for more details.

[1]
https://lore.kernel.org/linux-pm/Zqv9BOjxLAgyNP5B@hatbackup/



Suggested-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: John B. Wyatt IV <jwyatt@redhat.com>
Signed-off-by: John B. Wyatt IV <sageofredondo@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

There was a symbol listed in the powercap.h file that was not implemented.
Implement it with a stub return of 0.

Programs like SWIG require that functions that are defined in the
headers be implemented.

Fixes: c2294c14 ("cpupower: Introduce powercap intel-rapl library and powercap-info command")
Suggested-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: John B. Wyatt IV <jwyatt@redhat.com>
Signed-off-by: John B. Wyatt IV <sageofredondo@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

 - Standardize directory variables to support more flexible installations.
 - Add copyright and licensing information to the Makefile.
 - Introduce ".PHONY" declarations to ensure that specific targets are always
   executed, regardless of the presence of files with matching names.
 - Add a help target to provide usage instructions.

Signed-off-by: Amit Vadhavana <av2082000@gmail.com>
Acked-by: Todd Brandt <todd.e.brandt@linux.intel.com>
Link: https://patch.msgid.link/Update

 directory handling and installation process in Makefile
[ rjw: Changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

By default, sleepgraph.py creates suspend-{date}-{time} directories
to store artifacts, or suspend-{date}-{time}-xN if the --multi option
is used.

Ignore those directories by adding a .gitignore file.

Signed-off-by: Yo-Jung (Leo) Lin <0xff07@gmail.com>
Acked-by: Todd Brandt <todd.e.brandt@linux.intel.com>
Link: https://patch.msgid.link/20240825095353.7578-1-0xff07@gmail.com


[ rjw: Subject and changelog edits ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Update cpuidle tool to display the residency value of cpuidle
states. This addition provides a clearer and more detailed view
of idle state information when using cpuidle-info.

 --------------------------------
 Before Patch:
 --------------------------------
 $ cpupower idle-info
 CPUidle driver: intel_idle
 CPUidle governor: menu
 analyzing CPU 28:

 Number of idle states: 3
 Available idle states: POLL C1 C1E
 POLL:
 Flags/Description: CPUIDLE CORE POLL IDLE
 Latency: 0
 Usage: 7448
 Duration: 207170
 C1:
 Flags/Description: MWAIT 0x00
 Latency: 2
 Usage: 7023
 Duration: 3736853
 C1E:
 Flags/Description: MWAIT 0x01
 Latency: 10
 Usage: 18468
 Duration: 11396212

 --------------------------------
 After Patch:
 --------------------------------
 $ cpupower idle-info
 CPUidle driver: intel_idle
 CPUidle governor: menu
 analyzing CPU 12:

 Number of idle states: 3
 Available idle states: POLL C1 C1E
 POLL:
 Flags/Description: CPUIDLE CORE POLL IDLE
 Latency: 0
 Residency: 0
 Usage: 1950
 Duration: 38458
 C1:
 Flags/Description: MWAIT 0x00
 Latency: 2
 Residency: 2
 Usage: 10688
 Duration: 7133020
 C1E:
 Flags/Description: MWAIT 0x01
 Latency: 10
 Residency: 20
 Usage: 22356
 Duration: 15687259
 --------------------------------

Signed-off-by: Aboorva Devarajan <aboorvad@linux.ibm.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

Release 2024.07.26:

Enable turbostat extensions to add both perf and PMT
(Intel Platform Monitoring Technology) counters from the cmdline.

Demonstrate PMT access with built-in support for Meteor Lake's Die%c6 counter.

This commit:

Clean up white-space nits introduced since version 2024.05.10

Signed-off-by: Len Brown <len.brown@intel.com>

Some counters, like cpu/cache-misses/, expose and require umask=%x
parameter alongside event=%x in the sysfs perf counter's event file.
This change make sure we parse and use it when opening user added
counters.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Add a general description of the user interface for adding PMT
counters with the new --add pmt,... option.

Provide a complete example for requesting two counters.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Provide a definition for metadata that allows reading DC6 residency
counter via PMT and exposes it as a builtin counter.

Note that this residency counter is updated and read via
entirely different mechanisms vs the MSR-based residency counters.
On MTL processors, there are times when Die%c6 will report above 100%.
This is still useful, but don't expect 3 digits of precision...

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Allows users to read Intel PMT (Platform Monitoring Technology)
counters, providing interface similar to one used to add MSR and perf
counters. Because PMT is exposed as a raw MMIO range, without metadata,
user has to supply the necessary information to find and correctly
display the requested counter.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Printing information about the source and value during initialization and
reading of the counter for each cpu, while useful when debugging,
results in too verbose output.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

This leaves the stdout cleaner, having only counter data. It makes it
easier for programs to parse the output of turbostat, for example
selftests.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

"After" was missing an "r", nothing to see here.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

We had few lines about the feature, but without any complete examples.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

We had an extra "+" at the beginning of some lines that look like a
poorly formated patch.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

User can now read perf counters using "--add perf/<device>/<event>".
Other details work similarly to how --add works with MSRs.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

These three counters now are treated similar to other perf counters
groups. This simplifies and gets rid of a lot of special cases for APERF
and MPERF.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

This version addresses one issue:
- Fix updating TRL MSR after SST-TF is disabled in auto mode.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

When SST-TF is disabled in auto mode, the performance is getting
limited.

This is caused by wrong programming of Turbo Ratio Limit (TRL) MSR.

This MSR always accepts the frequency ratio in 100 MHz unit. When the
TPMI is sending TRL in 1 MHz unit, change to 100 MHz, before updating
TRL MSR.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

Invocation the tool built with the default settings fails:
$ cpupower
cpupower: error while loading shared libraries: libcpupower.so.1: cannot
open shared object file: No such file or directory

The issue is that Makefile puts the library to "/usr/lib64" dir for a 64
bit machine. This is wrong. According to the "File hierarchy standard
specification:
https://en.wikipedia.org/wiki/Filesystem_Hierarchy_Standard
https://refspecs.linuxfoundation.org/FHS_3.0/fhs-3.0.pdf



"/usr/lib<qual>" dirs are intended for alternative-format libraries
(e.g., "/usr/lib32" for 32-bit libraries on a 64-bit machine (optional)).

The utility is built for the current machine and doesn't handle
'CROSS_COMPILE' and 'ARCH' env variables. It also doesn't change bit
depth. So the result is always the same - binary for x86_64
architecture. Therefore the library should be put in the '/usr/lib'
dir regardless of the build options.
This is the case for all the distros that comply with the
'File Hierarchy Standard 3.0" by Linux Foundation. Most of the distros
comply with it. For example, one can check this by examining the
"/usr/lb64" dir on debian-based distros and find that it contains only
"/usr/lib64/ld-linux-x86-64.so.2". And examine that "/usr/lib" contains
both 32 and 64 bit code:
find /usr/lib -name "*.so*" -type f | xargs file | grep 32-bit
find /usr/lib -name "*.so*" -type f | xargs file | grep 64-bit

Fix the issue by changing library destination dir to "/usr/lib".

Signed-off-by: Roman Storozhenko <romeusmeister@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

Execution of the 'make' command in the 'bench' subfolder causes the
following error:

$ make O=cpupower/build/ DESTDIR=cpupower/install/ -j8
"  CC      " cpupower/build//main.o
"  CC      " cpupower/build//parse.o
/bin/sh: 1: "  CC      "cpupower/build//system.o
  CC      : not found
  make: *** [Makefile:21: cpupower/build//main.o] Error 127
  make: *** Waiting for unfinished jobs....
  /bin/sh: 1:   CC      : not found
  /bin/sh: 1:   CC      : not found
  make: *** [Makefile:21: cpupower/build//parse.o] Error 127
  make: *** [Makefile:21: cpupower/build//system.o] Error 127

The makefile uses variables defined in the main project makefile and it
is not intended to run standalone. The reason is that 'bench' subproject
depends on the 'libcpupower' library, see the 'compile-bench' target in
the main makefile.
Add a check that prevents standalone execution of the 'bench' makefile.

Signed-off-by: Roman Storozhenko <romeusmeister@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

It makes it harder to shoot yourself in the foot, by using
additional __must_be_array() check.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Reuse the enum. It means the same thing in both cases.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

fd_perf field used to be part of the union, but later moved out of it,
because we test it with fd_perf != -1 to determine if any perf counter
is opened, making the union unused.

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Fixes compilation errors for Makefile snapshot target described in:
commit 231ce08b ("tools/power turbostat: Add "snapshot:" Makefile target")

Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

New CPU #defines encode vendor and family as well as model.

N.B. Copied VFM_*() defines here from <asm/cpu_device_id.h> to avoid
an application picking a second internal kernel header file.

Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Commit 78464d76 ("tools/power turbostat: Add columns for clustered
uncore frequency") introduced 'probe_intel_uncore_frequency_cluster()'
in a way which prevents printing uncore frequency columns if either of
the '-q' or '-l' options are used. Systems which do not have multiple
uncore frequencies per package are unaffected by this regression.

Fix the function so that uncore frequency columns are shown when either
the '-l' or '-q' option is used by checking if 'quiet' is true after
adding counters for the uncore frequency columns.

Fixes: 78464d76 ("tools/power turbostat: Add columns for clustered uncore frequency")

Signed-off-by: Adam Hawley <adam.james.hawley@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

In some cases specifying the '-n' command line argument will cause
turbostat to fail.  For instance 'turbostat -n 1' works fine; however,
'turbostat -n 1 -d' will fail.  This is the result of the first call
to getopt_long_only() where "MP" is specified as the optstring.  This can
be easily fixed by changing the optstring from "MP" to "MPn:" to remove
ambiguity between the arguments.

tools/power turbostat: option '-n' is ambiguous; possibilities: '-num_iterations' '-no-msr' '-no-perf'

Fixes: a0e86c90 ("tools/power turbostat: Add --no-perf option")

Signed-off-by: David Arcari <darcari@redhat.com>
Signed-off-by: Len Brown <len.brown@intel.com>

There is a type 'enum operation_mode_e' contains the display modes of
the 'monitor' subcommand. This type isn't used though, instead the
variable 'mode' is of a simple 'int' type.
Change 'mode' variable type from 'int' to 'enum operation_mode_e' in
order to improve compiler type checking.
Built and tested this with different monitor cmdline params. Everything
works as expected, that is nothing changed and no regressions encountered.

Signed-off-by: Roman Storozhenko <romeusmeister@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

Remove not supported '-v' parameter from the cpupower's 'monitor'
command description.
There is a '-v' parameter described in cpupower's 'monitor' command man
page. It isn't supported at the moment, and perhaps has never been
supported. When I run the monitor with this parameter I get the
following:

$ sudo LD_LIBRARY_PATH=lib64/ bin/cpupower monitor -v
monitor: invalid option -- 'v'
invalid or unknown argument
$ sudo LD_LIBRARY_PATH=lib64/ bin/cpupower monitor -V
monitor: invalid option -- 'V'
invalid or unknown argument

Signed-off-by: Roman Storozhenko <romeusmeister@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

Enhance cpupower build process description with the information on
building and installing the utility to the user defined directories
as well as with the information on the way of running the utility from
the custom defined installation directory.

Signed-off-by: Roman Storozhenko <romeusmeister@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

Make "cpupower" building process more user friendly by adding 'help'
target to the main makefile. This target describes various build
and cleaning options available to the user.

Signed-off-by: Roman Storozhenko <romeusmeister@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

Replace a dead reference link to a turbo boost technology description with
a reference to a root page of the technology on the Intel site, and add
another one, describing power management technology, which includes short
description of the turbo boost.

Signed-off-by: Roman Storozhenko <romeusmeister@gmail.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>

sleepgraph/bootgraph function correctly in python3.12 but include a slew
of deprecation warnings for unsupported regexes. This patch fixes up all
the strings in the code so that it comforms with python3.12 standards.

Signed-off-by: Todd Brandt <todd.e.brandt@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

- fix S3 suspend fail double run by using fp.flush to /sys/power/state
- when running turbostat print the return value
- handle case where html files have binary data
- max issues in summary-issues is now 100 (in case there are thousands)
- add backup to dmidecode, use /sys/class/dmi/id/ in case /dev/mem fails
- update summary page to use full mode (disk-platform instead of disk)

Signed-off-by: Todd Brandt <todd.e.brandt@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Update cpupower's P-State frequency calculation and reporting with AMD
Family 1Ah+ processors, when using the acpi-cpufreq driver. This is due
to a change in the PStateDef MSR layout in AMD Family 1Ah+.

Tested on 4th and 5th Gen AMD EPYC system

Signed-off-by: Ananth Narayan <Ananth.Narayan@amd.com>
Signed-off-by: Dhananjay Ugwekar <Dhananjay.Ugwekar@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>