Merge tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-tracegrafted

Pull tracing fixes from Steven Rostedt:

 - Don't destroy user event fields when removal fails

   User event fields are destroyed before the event is removed from
   visibility. But that can fail leaving the still visible event with no
   fields. Move the destroying of the fields to after the event is
   successfully removed from visibility.

 - Initialize function graph state is fork before calling
   copy_exec_state()

   For non-CLONE_VM forks, copy_exec_state() allocates a new
   task_exec_state. If that allocation fails, ftrace_graph_exit_task()
   will free the tasks ret_stack pointer. Since that pointer is still
   using the parent's ret_stack, it mistakenly frees the parent's
   pointer too.

   Call ftrace_graph_init() on the task first which will NULL out the
   new tasks's ret_stack and if the copy fails, it will not free
   anything.

 - Remove FGRAPH_MAX_INDEX

   The macro FGRAPH_MAX_INDEX was added but never used. Remove it.

 - Save ent_size in function graph printing of nested functions

   The function graph tracer needs to look at the next event to see if
   the next event is the return of the current function entry. If it is,
   it prints a single line:

	ktime_get();

   Otherwise it prints it like a nested function:

	tick_nohz_irq_exit() {
	    ktime_get();
	    kcpustat_irq_exit();
	}

   In order to look at the next event, it must save the current event so
   that it has the information to print from it. It saves the event in
   the iterator descriptor called "ent". What it doesn't save is the
   ent_size of the event which is now used to know if the function graph
   arguments are to be printed. The peek doesn't save the size so the
   size used happens to be that of the size of the last event that was
   seen.

   Save the entry event size in the iterator descriptor so that the
   correct size is used.

 - Fix several errors with freeing data in the histogram code

   The histogram code had a lot of leaked or or incorrect accounting
   when failures happen. Correct them.

 - Fix histogram regression of .percent and .graph modifiers

   Up until 6.3 histogram values could have "percent" or "graph"
   modifiers that changed how they were printed. But a change that added
   restricting histograms values from being strings, stack traces and
   other modifiers inadvertently prevented them from using the percent
   and graph modifiers, which were legal use cases for values.

   Put back the percent and graph modifiers.

 - Fix various typos in the comments

 - Set the trace_clock before initializing a histogram with clock
   argument

   The histogram API allows the user to specific which trace clock to
   use via a "clock=" string. The histogram is set up first before the
   clock is checked. If the passed in clock is not valid, it exits
   without fully fixing up the histogram leaving it on the list and a
   use-after-free can trigger.

   Update the clock argument first and if it fails then exit gracefully
   before the histogram trigger is placed on any lists.

 - Restore :mod: trailer after parsing in ftrace_set_clr_event

   The function ftrace_set_clr_event() modifies the parse string and
   needs to put it back to what was passed in. It searches for ":mod:"
   via a strsep() but fails to put back the first ':' in the string.

   Add back the ':' in the passed in string.

 - Take trace_array reference when opening a tracer options file

   The options files are dynamically created and some tracers add their
   own options. When a tracer adds their own list of options, the
   trace_array holding them has an array to hold the list of options for
   each tracer. This array increases in size via a krealloc(), and the
   new entry gets a newly allocated array to hold the options of the new
   tracer being added.

   The element in each entry of the tracer's option array holds a
   pointer back to the trace_array, a pointer to the tracer it is
   associated to, a pointer to the flags of the option.

   The issue is that these arrays are freed when the trace_array is
   freed when its instance it represents is removed from the instances
   directory. There's a race that an open of one of these options files
   can happen when the instance is being removed.

   Add a new helper function to be called by the open function of the
   options file to iterate all existing trace_arrays under a lock and
   find the one that has the given option element in one of it's tracer
   arrays. If found, then update the associated trace_array's reference
   counter to keep it from being freed. If not found, have the open call
   return -ENODEV.

 - Disable interrupts when acquiring the lock in rb_wake_up_waiters()

   The function rb_wake_up_waiters() assumes it will be called in
   interrupt context and does not disable irqs when taking
   cpu_buffer->reader_lock, which can be called in hard interrupt
   context. The issue is in PREEMPT_RT, this function is called in
   thread context leaving this lock open to a deadlock.

   Take the lock with interrupts disabled.

 - Use rcu_assign_pointer() for tmp_ops filter hash

   The tmp_ops used in update_ftrace_direct_mod() assigns its
   filter_hash field directly, but that field is annotated as __rcu and
   sparse complains. Assign it with rcu_assign_pointer()

 - Fix use-after-free in enable_trigger_private_data_free()

   The trace_event_call is accessed through the event_trigger_data's
   trace_event_file pointer to put the trace_event_call on freeing. The
   issue is that the trace_event_file data may have been freed already
   causing a use-after-free. Add a field to the event_trigger_data that
   points directly to the trace_event_call so that it can decrement its
   reference directly without needing to go through the
   trace_event_file.

 - Fix accounting of buffer data remote headers

   trace_buffer_desc_size() and trace_remote_alloc_buffer() undercount
   the number of pages is needed for the asked for size as it doesn't
   take into account the meta data on each page. Add a helper function
   to do the calculation properly and use that in these functions.

 - Catch nr_page_va overflow in ring_buffer_desc sizing

   The number of pages per remote ring buffer is capped by
   ring_buffer_desc::nr_page_va (32 bits). A buffer_size large enough to
   overflow that field would silently allocate a descriptor smaller than
   what was asked for.

 - Do not resize the subbuf order if any per_cpu buffer is disabled

   The mmapping of ring buffers disables resizing the subbuffers, but it
   is done per-cpu whereas the subbuf size change is done for all the
   per_cpu buffers under the buffer->mutex. It could change the size of
   some while the mapping is happening on others. Have the resize of the
   subbuf order check all the per_cpu buffers under the lock to see if
   any of them is disabled before starting and causing an inconsistency
   between buffers that are being mapped.

* tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (25 commits)
  ring-buffer: Check resize_disabled before publishing the new subbuf order
  tracing/remotes: Catch nr_page_va overflow in ring_buffer_desc sizing
  tracing/remotes: Account for ring buffer page header in size calculation
  tracing: Don't dereference trace_event_file in deferred trigger free
  ftrace: Use rcu_assign_pointer() for tmp_ops filter hash
  ring-buffer: Acquire the lock with irqsave in rb_wake_up_waiters()
  tracing: Take trace_array reference when opening a tracer options file
  tracing: Fix ring_buffer_read_page_size() kernel-doc
  tracing: Restore :mod: trailer after parsing in ftrace_set_clr_event()
  tracing: Fix memory corruption from a "STACKTRACE" histogram key
  tracing: Fix memory corruption from the histogram stacktrace modifier
  tracing: Undo the registration when enabling the histogram trigger fails
  tracing: Take the reference before publishing the named histogram trigger
  tracing: Set the trace clock before registering the histogram trigger
  tracing: Fix typo "preceeded" in comment
  tracing: Fix typo "availabe" in comment
  tracing: Let histogram values keep the percent and graph modifiers
  tracing: Keep the entry count when the histogram stats allocation fails
  tracing: Free histogram the field rejected for a bad modifier
  tracing: Free histogram the var ref when its initialization fails
  ...

6 files changed, 1151 insertions, 0 deletions

diff --git a/tools/testing/selftests/amd-pstate/Makefile b/tools/testing/selftests/amd-pstate/Makefile
new file mode 100644
index 000000000..c382f579f
--- /dev/null
+++ b/tools/testing/selftests/amd-pstate/Makefile
@@ -0,0 +1,18 @@
+# SPDX-License-Identifier: GPL-2.0-only
+# Makefile for amd-pstate/ function selftests
+
+# No binaries, but make sure arg-less "make" doesn't trigger "run_tests"
+all:
+
+ARCH ?= $(shell uname -m 2>/dev/null || echo not)
+ARCH := $(shell echo $(ARCH) | sed -e s/i.86/x86/ -e s/x86_64/x86/)
+
+ifeq (x86,$(ARCH))
+TEST_FILES += ../../../power/x86/amd_pstate_tracer/amd_pstate_trace.py
+TEST_FILES += ../../../power/x86/intel_pstate_tracer/intel_pstate_tracer.py
+endif
+
+TEST_PROGS += run.sh
+TEST_FILES += basic.sh tbench.sh gitsource.sh
+
+include ../lib.mk
diff --git a/tools/testing/selftests/amd-pstate/basic.sh b/tools/testing/selftests/amd-pstate/basic.sh
new file mode 100755
index 000000000..e4c43193e
--- /dev/null
+++ b/tools/testing/selftests/amd-pstate/basic.sh
@@ -0,0 +1,38 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0
+
+# amd-pstate-ut is a test module for testing the amd-pstate driver.
+# It can only run on x86 architectures and current cpufreq driver
+# must be amd-pstate.
+# (1) It can help all users to verify their processor support
+# (SBIOS/Firmware or Hardware).
+# (2) Kernel can have a basic function test to avoid the kernel
+# regression during the update.
+# (3) We can introduce more functional or performance tests to align
+# the result together, it will benefit power and performance scale optimization.
+
+# protect against multiple inclusion
+if [ $FILE_BASIC ]; then
+	return 0
+else
+	FILE_BASIC=DONE
+fi
+
+amd_pstate_basic()
+{
+	printf "\n---------------------------------------------\n"
+	printf "*** Running AMD P-state ut                ***"
+	printf "\n---------------------------------------------\n"
+
+	if ! /sbin/modprobe -q -n amd-pstate-ut; then
+		echo "amd-pstate-ut: module amd-pstate-ut is not found [SKIP]"
+		exit $ksft_skip
+	fi
+	if /sbin/modprobe -q amd-pstate-ut; then
+		/sbin/modprobe -q -r amd-pstate-ut
+		echo "amd-pstate-basic: ok"
+	else
+		echo "amd-pstate-basic: [FAIL]"
+		exit 1
+	fi
+}
diff --git a/tools/testing/selftests/amd-pstate/config b/tools/testing/selftests/amd-pstate/config
new file mode 100644
index 000000000..f43103c9a
--- /dev/null
+++ b/tools/testing/selftests/amd-pstate/config
@@ -0,0 +1 @@
+CONFIG_X86_AMD_PSTATE_UT=m
diff --git a/tools/testing/selftests/amd-pstate/gitsource.sh b/tools/testing/selftests/amd-pstate/gitsource.sh
new file mode 100755
index 000000000..4cde62f90
--- /dev/null
+++ b/tools/testing/selftests/amd-pstate/gitsource.sh
@@ -0,0 +1,359 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+# Testing and monitor the cpu desire performance, frequency, load,
+# power consumption and throughput etc. when this script trigger
+# gitsource test.
+# 1) Download and tar gitsource codes.
+# 2) Run gitsource benchmark on specific governors, ondemand or schedutil.
+# 3) Run tbench benchmark comparative test on acpi-cpufreq kernel driver.
+# 4) Get desire performance, frequency, load by perf.
+# 5) Get power consumption and throughput by amd_pstate_trace.py.
+# 6) Get run time by /usr/bin/time.
+# 7) Analyse test results and save it in file selftest.gitsource.csv.
+#8) Plot png images about time, energy and performance per watt for each test.
+
+# protect against multiple inclusion
+if [ $FILE_GITSOURCE ]; then
+	return 0
+else
+	FILE_GITSOURCE=DONE
+fi
+
+git_name="git-2.15.1"
+git_tar="$git_name.tar.gz"
+gitsource_url="https://github.com/git/git/archive/refs/tags/v2.15.1.tar.gz"
+gitsource_governors=("ondemand" "schedutil")
+
+# $1: governor, $2: round, $3: des-perf, $4: freq, $5: load, $6: time $7: energy, $8: PPW
+store_csv_gitsource()
+{
+	echo "$1, $2, $3, $4, $5, $6, $7, $8" | tee -a $OUTFILE_GIT.csv > /dev/null 2>&1
+}
+
+# clear some special lines
+clear_csv_gitsource()
+{
+	if [ -f $OUTFILE_GIT.csv ]; then
+		sed -i '/Comprison(%)/d' $OUTFILE_GIT.csv
+		sed -i "/$(scaling_name)/d" $OUTFILE_GIT.csv
+	fi
+}
+
+# find string $1 in file csv and get the number of lines
+get_lines_csv_gitsource()
+{
+	if [ -f $OUTFILE_GIT.csv ]; then
+		return `grep -c "$1" $OUTFILE_GIT.csv`
+	else
+		return 0
+	fi
+}
+
+pre_clear_gitsource()
+{
+	post_clear_gitsource
+	rm -rf gitsource_*.png
+	clear_csv_gitsource
+}
+
+post_clear_gitsource()
+{
+	rm -rf results/tracer-gitsource*
+	rm -rf $OUTFILE_GIT*.log
+	rm -rf $OUTFILE_GIT*.result
+}
+
+install_gitsource()
+{
+	if [ ! -d $SCRIPTDIR/$git_name ]; then
+		pushd $(pwd) > /dev/null 2>&1
+		cd $SCRIPTDIR
+		printf "Download gitsource, please wait a moment ...\n\n"
+		wget -O $git_tar $gitsource_url > /dev/null 2>&1
+
+		printf "Tar gitsource ...\n\n"
+		tar -xzf $git_tar
+		popd > /dev/null 2>&1
+	fi
+}
+
+# $1: governor, $2: loop
+run_gitsource()
+{
+	echo "Launching amd pstate tracer for $1 #$2 tracer_interval: $TRACER_INTERVAL"
+	$TRACER -n tracer-gitsource-$1-$2 -i $TRACER_INTERVAL > /dev/null 2>&1 &
+
+	printf "Make and test gitsource for $1 #$2 make_cpus: $MAKE_CPUS\n"
+	BACKUP_DIR=$(pwd)
+	pushd $BACKUP_DIR > /dev/null 2>&1
+	cd $SCRIPTDIR/$git_name
+	$PERF stat -a --per-socket -I 1000 -e power/energy-pkg/ /usr/bin/time -o $BACKUP_DIR/$OUTFILE_GIT.time-gitsource-$1-$2.log make test -j$MAKE_CPUS > $BACKUP_DIR/$OUTFILE_GIT-perf-$1-$2.log 2>&1
+	popd > /dev/null 2>&1
+
+	for job in `jobs -p`
+	do
+		echo "Waiting for job id $job"
+		wait $job
+	done
+}
+
+# $1: governor, $2: loop
+parse_gitsource()
+{
+	awk '{print $5}' results/tracer-gitsource-$1-$2/cpu.csv | sed -e '1d' | sed s/,// > $OUTFILE_GIT-des-perf-$1-$2.log
+	avg_des_perf=$(awk 'BEGIN {i=0; sum=0};{i++; sum += $1};END {print sum/i}' $OUTFILE_GIT-des-perf-$1-$2.log)
+	printf "Gitsource-$1-#$2 avg des perf: $avg_des_perf\n" | tee -a $OUTFILE_GIT.result
+
+	awk '{print $7}' results/tracer-gitsource-$1-$2/cpu.csv | sed -e '1d' | sed s/,// > $OUTFILE_GIT-freq-$1-$2.log
+	avg_freq=$(awk 'BEGIN {i=0; sum=0};{i++; sum += $1};END {print sum/i}' $OUTFILE_GIT-freq-$1-$2.log)
+	printf "Gitsource-$1-#$2 avg freq: $avg_freq\n" | tee -a $OUTFILE_GIT.result
+
+	awk '{print $11}' results/tracer-gitsource-$1-$2/cpu.csv | sed -e '1d' | sed s/,// > $OUTFILE_GIT-load-$1-$2.log
+	avg_load=$(awk 'BEGIN {i=0; sum=0};{i++; sum += $1};END {print sum/i}' $OUTFILE_GIT-load-$1-$2.log)
+	printf "Gitsource-$1-#$2 avg load: $avg_load\n" | tee -a $OUTFILE_GIT.result
+
+	grep user $OUTFILE_GIT.time-gitsource-$1-$2.log | awk '{print $1}' | sed -e 's/user//' > $OUTFILE_GIT-time-$1-$2.log
+	time_sum=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum}' $OUTFILE_GIT-time-$1-$2.log)
+	printf "Gitsource-$1-#$2 user time(s): $time_sum\n" | tee -a $OUTFILE_GIT.result
+
+	grep Joules $OUTFILE_GIT-perf-$1-$2.log | awk '{print $4}' > $OUTFILE_GIT-energy-$1-$2.log
+	en_sum=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum}' $OUTFILE_GIT-energy-$1-$2.log)
+	printf "Gitsource-$1-#$2 power consumption(J): $en_sum\n" | tee -a $OUTFILE_GIT.result
+
+	# Permance is the number of run gitsource per second, denoted 1/t, where 1 is the number of run gitsource in t
+	# seconds. It is well known that P=E/t, where P is power measured in watts(W), E is energy measured in joules(J),
+	# and t is time measured in seconds(s). This means that performance per watt becomes
+	#        1/t     1/t     1
+	#       ----- = ----- = ---
+	#         P      E/t     E
+	# with unit given by 1 per joule.
+	ppw=`echo "scale=9;1/$en_sum" | bc | awk '{printf "%.9f", $0}'`
+	printf "Gitsource-$1-#$2 performance per watt(1/J): $ppw\n" | tee -a $OUTFILE_GIT.result
+	printf "\n" | tee -a $OUTFILE_GIT.result
+
+	driver_name=`echo $(scaling_name)`
+	store_csv_gitsource "$driver_name-$1" $2 $avg_des_perf $avg_freq $avg_load $time_sum $en_sum $ppw
+}
+
+# $1: governor
+loop_gitsource()
+{
+	printf "\nGitsource total test times is $LOOP_TIMES for $1\n\n"
+	for i in `seq 1 $LOOP_TIMES`
+	do
+		run_gitsource $1 $i
+		parse_gitsource $1 $i
+	done
+}
+
+# $1: governor
+gather_gitsource()
+{
+	printf "Gitsource test result for $1 (loops:$LOOP_TIMES)" | tee -a $OUTFILE_GIT.result
+	printf "\n--------------------------------------------------\n" | tee -a $OUTFILE_GIT.result
+
+	grep "Gitsource-$1-#" $OUTFILE_GIT.result | grep "avg des perf:" | awk '{print $NF}' > $OUTFILE_GIT-des-perf-$1.log
+	avg_des_perf=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_GIT-des-perf-$1.log)
+	printf "Gitsource-$1 avg des perf: $avg_des_perf\n" | tee -a $OUTFILE_GIT.result
+
+	grep "Gitsource-$1-#" $OUTFILE_GIT.result | grep "avg freq:" | awk '{print $NF}' > $OUTFILE_GIT-freq-$1.log
+	avg_freq=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_GIT-freq-$1.log)
+	printf "Gitsource-$1 avg freq: $avg_freq\n" | tee -a $OUTFILE_GIT.result
+
+	grep "Gitsource-$1-#" $OUTFILE_GIT.result | grep "avg load:" | awk '{print $NF}' > $OUTFILE_GIT-load-$1.log
+	avg_load=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_GIT-load-$1.log)
+	printf "Gitsource-$1 avg load: $avg_load\n" | tee -a $OUTFILE_GIT.result
+
+	grep "Gitsource-$1-#" $OUTFILE_GIT.result | grep "user time(s):" | awk '{print $NF}' > $OUTFILE_GIT-time-$1.log
+	time_sum=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum}' $OUTFILE_GIT-time-$1.log)
+	printf "Gitsource-$1 total user time(s): $time_sum\n" | tee -a $OUTFILE_GIT.result
+
+	avg_time=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_GIT-time-$1.log)
+	printf "Gitsource-$1 avg user times(s): $avg_time\n" | tee -a $OUTFILE_GIT.result
+
+	grep "Gitsource-$1-#" $OUTFILE_GIT.result | grep "power consumption(J):" | awk '{print $NF}' > $OUTFILE_GIT-energy-$1.log
+	en_sum=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum}' $OUTFILE_GIT-energy-$1.log)
+	printf "Gitsource-$1 total power consumption(J): $en_sum\n" | tee -a $OUTFILE_GIT.result
+
+	avg_en=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_GIT-energy-$1.log)
+	printf "Gitsource-$1 avg power consumption(J): $avg_en\n" | tee -a $OUTFILE_GIT.result
+
+	# Permance is the number of run gitsource per second, denoted 1/t, where 1 is the number of run gitsource in t
+	# seconds. It is well known that P=E/t, where P is power measured in watts(W), E is energy measured in joules(J),
+	# and t is time measured in seconds(s). This means that performance per watt becomes
+	#        1/t     1/t     1
+	#       ----- = ----- = ---
+	#         P      E/t     E
+	# with unit given by 1 per joule.
+	ppw=`echo "scale=9;1/$avg_en" | bc | awk '{printf "%.9f", $0}'`
+	printf "Gitsource-$1 performance per watt(1/J): $ppw\n" | tee -a $OUTFILE_GIT.result
+	printf "\n" | tee -a $OUTFILE_GIT.result
+
+	driver_name=`echo $(scaling_name)`
+	store_csv_gitsource "$driver_name-$1" "Average" $avg_des_perf $avg_freq $avg_load $avg_time $avg_en $ppw
+}
+
+# $1: base scaling_driver $2: base governor $3: comparison scaling_driver $4: comparison governor
+__calc_comp_gitsource()
+{
+	base=`grep "$1-$2" $OUTFILE_GIT.csv | grep "Average"`
+	comp=`grep "$3-$4" $OUTFILE_GIT.csv | grep "Average"`
+
+	if [ -n "$base" -a -n "$comp" ]; then
+		printf "\n==================================================\n" | tee -a $OUTFILE_GIT.result
+		printf "Gitsource comparison $1-$2 VS $3-$4" | tee -a $OUTFILE_GIT.result
+		printf "\n==================================================\n" | tee -a $OUTFILE_GIT.result
+
+		# get the base values
+		des_perf_base=`echo "$base" | awk '{print $3}' | sed s/,//`
+		freq_base=`echo "$base" | awk '{print $4}' | sed s/,//`
+		load_base=`echo "$base" | awk '{print $5}' | sed s/,//`
+		time_base=`echo "$base" | awk '{print $6}' | sed s/,//`
+		energy_base=`echo "$base" | awk '{print $7}' | sed s/,//`
+		ppw_base=`echo "$base" | awk '{print $8}' | sed s/,//`
+
+		# get the comparison values
+		des_perf_comp=`echo "$comp" | awk '{print $3}' | sed s/,//`
+		freq_comp=`echo "$comp" | awk '{print $4}' | sed s/,//`
+		load_comp=`echo "$comp" | awk '{print $5}' | sed s/,//`
+		time_comp=`echo "$comp" | awk '{print $6}' | sed s/,//`
+		energy_comp=`echo "$comp" | awk '{print $7}' | sed s/,//`
+		ppw_comp=`echo "$comp" | awk '{print $8}' | sed s/,//`
+
+		# compare the base and comp values
+		des_perf_drop=`echo "scale=4;($des_perf_comp-$des_perf_base)*100/$des_perf_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Gitsource-$1 des perf base: $des_perf_base comprison: $des_perf_comp percent: $des_perf_drop\n" | tee -a $OUTFILE_GIT.result
+
+		freq_drop=`echo "scale=4;($freq_comp-$freq_base)*100/$freq_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Gitsource-$1 freq base: $freq_base comprison: $freq_comp percent: $freq_drop\n" | tee -a $OUTFILE_GIT.result
+
+		load_drop=`echo "scale=4;($load_comp-$load_base)*100/$load_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Gitsource-$1 load base: $load_base comprison: $load_comp percent: $load_drop\n" | tee -a $OUTFILE_GIT.result
+
+		time_drop=`echo "scale=4;($time_comp-$time_base)*100/$time_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Gitsource-$1 time base: $time_base comprison: $time_comp percent: $time_drop\n" | tee -a $OUTFILE_GIT.result
+
+		energy_drop=`echo "scale=4;($energy_comp-$energy_base)*100/$energy_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Gitsource-$1 energy base: $energy_base comprison: $energy_comp percent: $energy_drop\n" | tee -a $OUTFILE_GIT.result
+
+		ppw_drop=`echo "scale=4;($ppw_comp-$ppw_base)*100/$ppw_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Gitsource-$1 performance per watt base: $ppw_base comprison: $ppw_comp percent: $ppw_drop\n" | tee -a $OUTFILE_GIT.result
+		printf "\n" | tee -a $OUTFILE_GIT.result
+
+		store_csv_gitsource "$1-$2 VS $3-$4" "Comprison(%)" "$des_perf_drop" "$freq_drop" "$load_drop" "$time_drop" "$energy_drop" "$ppw_drop"
+	fi
+}
+
+# calculate the comparison(%)
+calc_comp_gitsource()
+{
+	# acpi-cpufreq-ondemand VS acpi-cpufreq-schedutil
+	__calc_comp_gitsource ${all_scaling_names[0]} ${gitsource_governors[0]} ${all_scaling_names[0]} ${gitsource_governors[1]}
+
+	# amd-pstate-ondemand VS amd-pstate-schedutil
+	__calc_comp_gitsource ${all_scaling_names[1]} ${gitsource_governors[0]} ${all_scaling_names[1]} ${gitsource_governors[1]}
+
+	# acpi-cpufreq-ondemand VS amd-pstate-ondemand
+	__calc_comp_gitsource ${all_scaling_names[0]} ${gitsource_governors[0]} ${all_scaling_names[1]} ${gitsource_governors[0]}
+
+	# acpi-cpufreq-schedutil VS amd-pstate-schedutil
+	__calc_comp_gitsource ${all_scaling_names[0]} ${gitsource_governors[1]} ${all_scaling_names[1]} ${gitsource_governors[1]}
+}
+
+# $1: file_name, $2: title, $3: ylable, $4: column
+plot_png_gitsource()
+{
+	# all_scaling_names[1] all_scaling_names[0] flag
+	#    amd-pstate           acpi-cpufreq
+	#         N                   N             0
+	#         N                   Y             1
+	#         Y                   N             2
+	#         Y                   Y             3
+	ret=`grep -c "${all_scaling_names[1]}" $OUTFILE_GIT.csv`
+	if [ $ret -eq 0 ]; then
+		ret=`grep -c "${all_scaling_names[0]}" $OUTFILE_GIT.csv`
+		if [ $ret -eq 0 ]; then
+			flag=0
+		else
+			flag=1
+		fi
+	else
+		ret=`grep -c "${all_scaling_names[0]}" $OUTFILE_GIT.csv`
+		if [ $ret -eq 0 ]; then
+			flag=2
+		else
+			flag=3
+		fi
+	fi
+
+	gnuplot << EOF
+		set term png
+		set output "$1"
+
+		set title "$2"
+		set xlabel "Test Cycles (round)"
+		set ylabel "$3"
+
+		set grid
+		set style data histogram
+		set style fill solid 0.5 border
+		set boxwidth 0.8
+
+		if ($flag == 1) {
+			plot \
+			"<(sed -n -e 's/,//g' -e '/${all_scaling_names[0]}-${gitsource_governors[0]}/p' $OUTFILE_GIT.csv)" using $4:xtic(2) title "${all_scaling_names[0]}-${gitsource_governors[0]}", \
+			"<(sed -n -e 's/,//g' -e '/${all_scaling_names[0]}-${gitsource_governors[1]}/p' $OUTFILE_GIT.csv)" using $4:xtic(2) title "${all_scaling_names[0]}-${gitsource_governors[1]}"
+		} else {
+			if ($flag == 2) {
+				plot \
+				"<(sed -n -e 's/,//g' -e '/${all_scaling_names[1]}-${gitsource_governors[0]}/p' $OUTFILE_GIT.csv)" using $4:xtic(2) title "${all_scaling_names[1]}-${gitsource_governors[0]}", \
+				"<(sed -n -e 's/,//g' -e '/${all_scaling_names[1]}-${gitsource_governors[1]}/p' $OUTFILE_GIT.csv)" using $4:xtic(2) title "${all_scaling_names[1]}-${gitsource_governors[1]}"
+			} else {
+				if ($flag == 3 ) {
+					plot \
+					"<(sed -n -e 's/,//g' -e '/${all_scaling_names[0]}-${gitsource_governors[0]}/p' $OUTFILE_GIT.csv)" using $4:xtic(2) title "${all_scaling_names[0]}-${gitsource_governors[0]}", \
+					"<(sed -n -e 's/,//g' -e '/${all_scaling_names[0]}-${gitsource_governors[1]}/p' $OUTFILE_GIT.csv)" using $4:xtic(2) title "${all_scaling_names[0]}-${gitsource_governors[1]}", \
+					"<(sed -n -e 's/,//g' -e '/${all_scaling_names[1]}-${gitsource_governors[0]}/p' $OUTFILE_GIT.csv)" using $4:xtic(2) title "${all_scaling_names[1]}-${gitsource_governors[0]}", \
+					"<(sed -n -e 's/,//g' -e '/${all_scaling_names[1]}-${gitsource_governors[1]}/p' $OUTFILE_GIT.csv)" using $4:xtic(2) title "${all_scaling_names[1]}-${gitsource_governors[1]}"
+				}
+			}
+		}
+		quit
+EOF
+}
+
+amd_pstate_gitsource()
+{
+	printf "\n---------------------------------------------\n"
+	printf "*** Running gitsource                     ***"
+	printf "\n---------------------------------------------\n"
+
+	pre_clear_gitsource
+
+	install_gitsource
+
+	get_lines_csv_gitsource "Governor"
+	if [ $? -eq 0 ]; then
+		# add titles and unit for csv file
+		store_csv_gitsource "Governor" "Round" "Des-perf" "Freq" "Load" "Time" "Energy" "Performance Per Watt"
+		store_csv_gitsource "Unit" "" "" "GHz" "" "s" "J" "1/J"
+	fi
+
+	backup_governor
+	for governor in ${gitsource_governors[*]} ; do
+		printf "\nSpecified governor is $governor\n\n"
+		switch_governor $governor
+		loop_gitsource $governor
+		gather_gitsource $governor
+	done
+	restore_governor
+
+	plot_png_gitsource "gitsource_time.png" "Gitsource Benchmark Time" "Time (s)" 6
+	plot_png_gitsource "gitsource_energy.png" "Gitsource Benchmark Energy" "Energy (J)" 7
+	plot_png_gitsource "gitsource_ppw.png" "Gitsource Benchmark Performance Per Watt" "Performance Per Watt (1/J)" 8
+
+	calc_comp_gitsource
+
+	post_clear_gitsource
+}
diff --git a/tools/testing/selftests/amd-pstate/run.sh b/tools/testing/selftests/amd-pstate/run.sh
new file mode 100755
index 000000000..b053eea8b
--- /dev/null
+++ b/tools/testing/selftests/amd-pstate/run.sh
@@ -0,0 +1,396 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+# protect against multiple inclusion
+if [ $FILE_MAIN ]; then
+	return 0
+else
+	FILE_MAIN=DONE
+fi
+
+SCRIPTDIR=`dirname "$0"`
+TRACER=$SCRIPTDIR/../../../power/x86/amd_pstate_tracer/amd_pstate_trace.py
+
+source $SCRIPTDIR/basic.sh
+source $SCRIPTDIR/tbench.sh
+source $SCRIPTDIR/gitsource.sh
+
+# amd-pstate-ut only run on x86/x86_64 AMD systems.
+ARCH=$(uname -m 2>/dev/null | sed -e 's/i.86/x86/' -e 's/x86_64/x86/')
+VENDOR=$(cat /proc/cpuinfo | grep -m 1 'vendor_id' | awk '{print $NF}')
+
+msg="Skip all tests:"
+FUNC=all
+OUTFILE=selftest
+OUTFILE_TBENCH="$OUTFILE.tbench"
+OUTFILE_GIT="$OUTFILE.gitsource"
+
+PERF=/usr/bin/perf
+SYSFS=
+CPUROOT=
+CPUFREQROOT=
+MAKE_CPUS=
+
+TIME_LIMIT=100
+PROCESS_NUM=128
+LOOP_TIMES=3
+TRACER_INTERVAL=10
+CURRENT_TEST=amd-pstate
+COMPARATIVE_TEST=
+
+# Kselftest framework requirement - SKIP code is 4.
+ksft_skip=4
+all_scaling_names=("acpi-cpufreq" "amd-pstate")
+
+# Get current cpufreq scaling driver name
+scaling_name()
+{
+	if [ "$COMPARATIVE_TEST" = "" ]; then
+		echo "$CURRENT_TEST"
+	else
+		echo "$COMPARATIVE_TEST"
+	fi
+}
+
+# Counts CPUs with cpufreq directories
+count_cpus()
+{
+	count=0;
+
+	for cpu in `ls $CPUROOT | grep "cpu[0-9].*"`; do
+		if [ -d $CPUROOT/$cpu/cpufreq ]; then
+			let count=count+1;
+		fi
+	done
+
+	echo $count;
+}
+
+# $1: policy
+find_current_governor()
+{
+	cat $CPUFREQROOT/$1/scaling_governor
+}
+
+backup_governor()
+{
+	policies=$(ls $CPUFREQROOT| grep "policy[0-9].*")
+	for policy in $policies; do
+		cur_gov=$(find_current_governor $policy)
+		echo "$policy $cur_gov" >> $OUTFILE.backup_governor.log
+	done
+
+	printf "Governor $cur_gov backup done.\n"
+}
+
+restore_governor()
+{
+	i=0;
+
+	policies=$(awk '{print $1}' $OUTFILE.backup_governor.log)
+	for policy in $policies; do
+		let i++;
+		governor=$(sed -n ''$i'p' $OUTFILE.backup_governor.log | awk '{print $2}')
+
+		# switch governor
+		echo $governor > $CPUFREQROOT/$policy/scaling_governor
+	done
+
+	printf "Governor restored to $governor.\n"
+}
+
+# $1: governor
+switch_governor()
+{
+	policies=$(ls $CPUFREQROOT| grep "policy[0-9].*")
+	for policy in $policies; do
+		filepath=$CPUFREQROOT/$policy/scaling_available_governors
+
+		# Exit if cpu isn't managed by cpufreq core
+		if [ ! -f $filepath ]; then
+			return;
+		fi
+
+		echo $1 > $CPUFREQROOT/$policy/scaling_governor
+	done
+
+	printf "Switched governor to $1.\n"
+}
+
+# All amd-pstate tests
+amd_pstate_all()
+{
+	printf "\n=============================================\n"
+	printf "***** Running AMD P-state Sanity Tests  *****\n"
+	printf "=============================================\n\n"
+
+	count=$(count_cpus)
+	if [ $count = 0 ]; then
+		printf "No cpu is managed by cpufreq core, exiting\n"
+		exit;
+	else
+		printf "AMD P-state manages: $count CPUs\n"
+	fi
+
+	# unit test for amd-pstate kernel driver
+	amd_pstate_basic
+
+	# tbench
+	amd_pstate_tbench
+
+	# gitsource
+	amd_pstate_gitsource
+}
+
+help()
+{
+	printf "Usage: $0 [OPTION...]
+	[-h <help>]
+	[-o <output-file-for-dump>]
+	[-c <all: All testing,
+	     basic: Basic testing,
+	     tbench: Tbench testing,
+	     gitsource: Gitsource testing.>]
+	[-t <tbench time limit>]
+	[-p <tbench process number>]
+	[-l <loop times for tbench>]
+	[-i <amd tracer interval>]
+	[-b <perf binary>]
+	[-m <comparative test: acpi-cpufreq>]
+	\n"
+	exit 2
+}
+
+parse_arguments()
+{
+	while getopts ho:c:t:p:l:i:b:m: arg
+	do
+		case $arg in
+			h) # --help
+				help
+				;;
+
+			c) # --func_type (Function to perform: basic, tbench, gitsource (default: all))
+				FUNC=$OPTARG
+				;;
+
+			o) # --output-file (Output file to store dumps)
+				OUTFILE=$OPTARG
+				;;
+
+			t) # --tbench-time-limit
+				TIME_LIMIT=$OPTARG
+				;;
+
+			p) # --tbench-process-number
+				PROCESS_NUM=$OPTARG
+				;;
+
+			l) # --tbench/gitsource-loop-times
+				LOOP_TIMES=$OPTARG
+				;;
+
+			i) # --amd-tracer-interval
+				TRACER_INTERVAL=$OPTARG
+				;;
+
+			b) # --perf-binary
+				PERF=`realpath $OPTARG`
+				;;
+
+			m) # --comparative-test
+				COMPARATIVE_TEST=$OPTARG
+				;;
+
+			*)
+				help
+				;;
+		esac
+	done
+}
+
+command_perf()
+{
+	if ! $PERF -v; then
+		echo $msg please install perf or provide perf binary path as argument >&2
+		exit $ksft_skip
+	fi
+}
+
+command_tbench()
+{
+	if ! command -v tbench > /dev/null; then
+		if apt policy dbench > /dev/null 2>&1; then
+			echo $msg apt install dbench >&2
+			exit $ksft_skip
+		elif yum list available | grep dbench > /dev/null 2>&1; then
+			echo $msg yum install dbench >&2
+			exit $ksft_skip
+		fi
+	fi
+
+	if ! command -v tbench > /dev/null; then
+		echo $msg please install tbench. >&2
+		exit $ksft_skip
+	fi
+}
+
+prerequisite()
+{
+	if ! echo "$ARCH" | grep -q x86; then
+		echo "$0 # Skipped: Test can only run on x86 architectures."
+		exit $ksft_skip
+	fi
+
+	if ! echo "$VENDOR" | grep -iq amd; then
+		echo "$0 # Skipped: Test can only run on AMD CPU."
+		echo "$0 # Current cpu vendor is $VENDOR."
+		exit $ksft_skip
+	fi
+
+	scaling_driver=$(cat /sys/devices/system/cpu/cpufreq/policy0/scaling_driver)
+	if [ "$COMPARATIVE_TEST" = "" ]; then
+		if [ "$scaling_driver" != "$CURRENT_TEST" ]; then
+			echo "$0 # Skipped: Test can only run on $CURRENT_TEST driver or run comparative test."
+			echo "$0 # Please set X86_AMD_PSTATE enabled or run comparative test."
+			echo "$0 # Current cpufreq scaling driver is $scaling_driver."
+			exit $ksft_skip
+		fi
+	else
+		case "$FUNC" in
+			"tbench" | "gitsource")
+				if [ "$scaling_driver" != "$COMPARATIVE_TEST" ]; then
+					echo "$0 # Skipped: Comparison test can only run on $COMPARISON_TEST driver."
+					echo "$0 # Current cpufreq scaling driver is $scaling_driver."
+					exit $ksft_skip
+				fi
+				;;
+
+			*)
+				echo "$0 # Skipped: Comparison test are only for tbench or gitsource."
+				echo "$0 # Current comparative test is for $FUNC."
+				exit $ksft_skip
+				;;
+		esac
+	fi
+
+	if [ ! -w /dev ]; then
+		echo $msg please run this as root >&2
+		exit $ksft_skip
+	fi
+
+	case "$FUNC" in
+		"all")
+			command_perf
+			command_tbench
+			;;
+
+		"tbench")
+			command_perf
+			command_tbench
+			;;
+
+		"gitsource")
+			command_perf
+			;;
+	esac
+
+	SYSFS=`mount -t sysfs | head -1 | awk '{ print $3 }'`
+
+	if [ ! -d "$SYSFS" ]; then
+		echo $msg sysfs is not mounted >&2
+		exit 2
+	fi
+
+	CPUROOT=$SYSFS/devices/system/cpu
+	CPUFREQROOT="$CPUROOT/cpufreq"
+
+	if ! ls $CPUROOT/cpu* > /dev/null 2>&1; then
+		echo $msg cpus not available in sysfs >&2
+		exit 2
+	fi
+
+	if ! ls $CPUROOT/cpufreq > /dev/null 2>&1; then
+		echo $msg cpufreq directory not available in sysfs >&2
+		exit 2
+	fi
+}
+
+do_test()
+{
+	# Check if CPUs are managed by cpufreq or not
+	count=$(count_cpus)
+	MAKE_CPUS=$((count*2))
+
+	if [ $count = 0 ]; then
+		echo "No cpu is managed by cpufreq core, exiting"
+		exit 2;
+	fi
+
+	case "$FUNC" in
+		"all")
+			amd_pstate_all
+			;;
+
+		"basic")
+			amd_pstate_basic
+			;;
+
+		"tbench")
+			amd_pstate_tbench
+			;;
+
+		"gitsource")
+			amd_pstate_gitsource
+			;;
+
+		*)
+			echo "Invalid [-f] function type"
+			help
+			;;
+	esac
+}
+
+# clear dumps
+pre_clear_dumps()
+{
+	case "$FUNC" in
+		"all")
+			rm -rf $OUTFILE.log
+			rm -rf $OUTFILE.backup_governor.log
+			rm -rf *.png
+			;;
+
+		"tbench")
+			rm -rf $OUTFILE.log
+			rm -rf $OUTFILE.backup_governor.log
+			rm -rf tbench_*.png
+			;;
+
+		"gitsource")
+			rm -rf $OUTFILE.log
+			rm -rf $OUTFILE.backup_governor.log
+			rm -rf gitsource_*.png
+			;;
+
+		*)
+			;;
+	esac
+}
+
+post_clear_dumps()
+{
+	rm -rf $OUTFILE.log
+	rm -rf $OUTFILE.backup_governor.log
+}
+
+# Parse arguments
+parse_arguments $@
+
+# Make sure all requirements are met
+prerequisite
+
+# Run requested functions
+pre_clear_dumps
+do_test | tee -a $OUTFILE.log
+post_clear_dumps
diff --git a/tools/testing/selftests/amd-pstate/tbench.sh b/tools/testing/selftests/amd-pstate/tbench.sh
new file mode 100755
index 000000000..2a98d9c92
--- /dev/null
+++ b/tools/testing/selftests/amd-pstate/tbench.sh
@@ -0,0 +1,339 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0
+
+# Testing and monitor the cpu desire performance, frequency, load,
+# power consumption and throughput etc.when this script trigger tbench
+# test cases.
+# 1) Run tbench benchmark on specific governors, ondemand or schedutil.
+# 2) Run tbench benchmark comparative test on acpi-cpufreq kernel driver.
+# 3) Get desire performance, frequency, load by perf.
+# 4) Get power consumption and throughput by amd_pstate_trace.py.
+# 5) Analyse test results and save it in file selftest.tbench.csv.
+# 6) Plot png images about performance, energy and performance per watt for each test.
+
+# protect against multiple inclusion
+if [ $FILE_TBENCH ]; then
+	return 0
+else
+	FILE_TBENCH=DONE
+fi
+
+tbench_governors=("ondemand" "schedutil")
+
+# $1: governor, $2: round, $3: des-perf, $4: freq, $5: load, $6: performance, $7: energy, $8: performance per watt
+store_csv_tbench()
+{
+	echo "$1, $2, $3, $4, $5, $6, $7, $8" | tee -a $OUTFILE_TBENCH.csv > /dev/null 2>&1
+}
+
+# clear some special lines
+clear_csv_tbench()
+{
+	if [ -f $OUTFILE_TBENCH.csv ]; then
+		sed -i '/Comprison(%)/d' $OUTFILE_TBENCH.csv
+		sed -i "/$(scaling_name)/d" $OUTFILE_TBENCH.csv
+	fi
+}
+
+# find string $1 in file csv and get the number of lines
+get_lines_csv_tbench()
+{
+	if [ -f $OUTFILE_TBENCH.csv ]; then
+		return `grep -c "$1" $OUTFILE_TBENCH.csv`
+	else
+		return 0
+	fi
+}
+
+pre_clear_tbench()
+{
+	post_clear_tbench
+	rm -rf tbench_*.png
+	clear_csv_tbench
+}
+
+post_clear_tbench()
+{
+	rm -rf results/tracer-tbench*
+	rm -rf $OUTFILE_TBENCH*.log
+	rm -rf $OUTFILE_TBENCH*.result
+
+}
+
+# $1: governor, $2: loop
+run_tbench()
+{
+	echo "Launching amd pstate tracer for $1 #$2 tracer_interval: $TRACER_INTERVAL"
+	$TRACER -n tracer-tbench-$1-$2 -i $TRACER_INTERVAL > /dev/null 2>&1 &
+
+	printf "Test tbench for $1 #$2 time_limit: $TIME_LIMIT procs_num: $PROCESS_NUM\n"
+	tbench_srv > /dev/null 2>&1 &
+	$PERF stat -a --per-socket -I 1000 -e power/energy-pkg/ tbench -t $TIME_LIMIT $PROCESS_NUM > $OUTFILE_TBENCH-perf-$1-$2.log 2>&1
+
+	pid=`pidof tbench_srv`
+	kill $pid
+
+	for job in `jobs -p`
+	do
+		echo "Waiting for job id $job"
+		wait $job
+	done
+}
+
+# $1: governor, $2: loop
+parse_tbench()
+{
+	awk '{print $5}' results/tracer-tbench-$1-$2/cpu.csv | sed -e '1d' | sed s/,// > $OUTFILE_TBENCH-des-perf-$1-$2.log
+	avg_des_perf=$(awk 'BEGIN {i=0; sum=0};{i++; sum += $1};END {print sum/i}' $OUTFILE_TBENCH-des-perf-$1-$2.log)
+	printf "Tbench-$1-#$2 avg des perf: $avg_des_perf\n" | tee -a $OUTFILE_TBENCH.result
+
+	awk '{print $7}' results/tracer-tbench-$1-$2/cpu.csv | sed -e '1d' | sed s/,// > $OUTFILE_TBENCH-freq-$1-$2.log
+	avg_freq=$(awk 'BEGIN {i=0; sum=0};{i++; sum += $1};END {print sum/i}' $OUTFILE_TBENCH-freq-$1-$2.log)
+	printf "Tbench-$1-#$2 avg freq: $avg_freq\n" | tee -a $OUTFILE_TBENCH.result
+
+	awk '{print $11}' results/tracer-tbench-$1-$2/cpu.csv | sed -e '1d' | sed s/,// > $OUTFILE_TBENCH-load-$1-$2.log
+	avg_load=$(awk 'BEGIN {i=0; sum=0};{i++; sum += $1};END {print sum/i}' $OUTFILE_TBENCH-load-$1-$2.log)
+	printf "Tbench-$1-#$2 avg load: $avg_load\n" | tee -a $OUTFILE_TBENCH.result
+
+	grep Throughput $OUTFILE_TBENCH-perf-$1-$2.log | awk '{print $2}' > $OUTFILE_TBENCH-throughput-$1-$2.log
+	tp_sum=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum}' $OUTFILE_TBENCH-throughput-$1-$2.log)
+	printf "Tbench-$1-#$2 throughput(MB/s): $tp_sum\n" | tee -a $OUTFILE_TBENCH.result
+
+	grep Joules $OUTFILE_TBENCH-perf-$1-$2.log | awk '{print $4}' > $OUTFILE_TBENCH-energy-$1-$2.log
+	en_sum=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum}' $OUTFILE_TBENCH-energy-$1-$2.log)
+	printf "Tbench-$1-#$2 power consumption(J): $en_sum\n" | tee -a $OUTFILE_TBENCH.result
+
+	# Permance is throughput per second, denoted T/t, where T is throught rendered in t seconds.
+	# It is well known that P=E/t, where P is power measured in watts(W), E is energy measured in joules(J),
+	# and t is time measured in seconds(s). This means that performance per watt becomes
+	#       T/t   T/t    T
+	#       --- = --- = ---
+	#        P    E/t    E
+	# with unit given by MB per joule.
+	ppw=`echo "scale=4;($TIME_LIMIT-1)*$tp_sum/$en_sum" | bc | awk '{printf "%.4f", $0}'`
+	printf "Tbench-$1-#$2 performance per watt(MB/J): $ppw\n" | tee -a $OUTFILE_TBENCH.result
+	printf "\n" | tee -a $OUTFILE_TBENCH.result
+
+	driver_name=`echo $(scaling_name)`
+	store_csv_tbench "$driver_name-$1" $2 $avg_des_perf $avg_freq $avg_load $tp_sum $en_sum $ppw
+}
+
+# $1: governor
+loop_tbench()
+{
+	printf "\nTbench total test times is $LOOP_TIMES for $1\n\n"
+	for i in `seq 1 $LOOP_TIMES`
+	do
+		run_tbench $1 $i
+		parse_tbench $1 $i
+	done
+}
+
+# $1: governor
+gather_tbench()
+{
+	printf "Tbench test result for $1 (loops:$LOOP_TIMES)" | tee -a $OUTFILE_TBENCH.result
+	printf "\n--------------------------------------------------\n" | tee -a $OUTFILE_TBENCH.result
+
+	grep "Tbench-$1-#" $OUTFILE_TBENCH.result | grep "avg des perf:" | awk '{print $NF}' > $OUTFILE_TBENCH-des-perf-$1.log
+	avg_des_perf=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_TBENCH-des-perf-$1.log)
+	printf "Tbench-$1 avg des perf: $avg_des_perf\n" | tee -a $OUTFILE_TBENCH.result
+
+	grep "Tbench-$1-#" $OUTFILE_TBENCH.result | grep "avg freq:" | awk '{print $NF}' > $OUTFILE_TBENCH-freq-$1.log
+	avg_freq=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_TBENCH-freq-$1.log)
+	printf "Tbench-$1 avg freq: $avg_freq\n" | tee -a $OUTFILE_TBENCH.result
+
+	grep "Tbench-$1-#" $OUTFILE_TBENCH.result | grep "avg load:" | awk '{print $NF}' > $OUTFILE_TBENCH-load-$1.log
+	avg_load=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_TBENCH-load-$1.log)
+	printf "Tbench-$1 avg load: $avg_load\n" | tee -a $OUTFILE_TBENCH.result
+
+	grep "Tbench-$1-#" $OUTFILE_TBENCH.result | grep "throughput(MB/s):" | awk '{print $NF}' > $OUTFILE_TBENCH-throughput-$1.log
+	tp_sum=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum}' $OUTFILE_TBENCH-throughput-$1.log)
+	printf "Tbench-$1 total throughput(MB/s): $tp_sum\n" | tee -a $OUTFILE_TBENCH.result
+
+	avg_tp=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_TBENCH-throughput-$1.log)
+	printf "Tbench-$1 avg throughput(MB/s): $avg_tp\n" | tee -a $OUTFILE_TBENCH.result
+
+	grep "Tbench-$1-#" $OUTFILE_TBENCH.result | grep "power consumption(J):" | awk '{print $NF}' > $OUTFILE_TBENCH-energy-$1.log
+	en_sum=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum}' $OUTFILE_TBENCH-energy-$1.log)
+	printf "Tbench-$1 total power consumption(J): $en_sum\n" | tee -a $OUTFILE_TBENCH.result
+
+	avg_en=$(awk 'BEGIN {sum=0};{sum += $1};END {print sum/'$LOOP_TIMES'}' $OUTFILE_TBENCH-energy-$1.log)
+	printf "Tbench-$1 avg power consumption(J): $avg_en\n" | tee -a $OUTFILE_TBENCH.result
+
+	# Permance is throughput per second, denoted T/t, where T is throught rendered in t seconds.
+	# It is well known that P=E/t, where P is power measured in watts(W), E is energy measured in joules(J),
+	# and t is time measured in seconds(s). This means that performance per watt becomes
+	#       T/t   T/t    T
+	#       --- = --- = ---
+	#        P    E/t    E
+	# with unit given by MB per joule.
+	ppw=`echo "scale=4;($TIME_LIMIT-1)*$avg_tp/$avg_en" | bc | awk '{printf "%.4f", $0}'`
+	printf "Tbench-$1 performance per watt(MB/J): $ppw\n" | tee -a $OUTFILE_TBENCH.result
+	printf "\n" | tee -a $OUTFILE_TBENCH.result
+
+	driver_name=`echo $(scaling_name)`
+	store_csv_tbench "$driver_name-$1" "Average" $avg_des_perf $avg_freq $avg_load $avg_tp $avg_en $ppw
+}
+
+# $1: base scaling_driver $2: base governor $3: comparative scaling_driver $4: comparative governor
+__calc_comp_tbench()
+{
+	base=`grep "$1-$2" $OUTFILE_TBENCH.csv | grep "Average"`
+	comp=`grep "$3-$4" $OUTFILE_TBENCH.csv | grep "Average"`
+
+	if [ -n "$base" -a -n "$comp" ]; then
+		printf "\n==================================================\n" | tee -a $OUTFILE_TBENCH.result
+		printf "Tbench comparison $1-$2 VS $3-$4" | tee -a $OUTFILE_TBENCH.result
+		printf "\n==================================================\n" | tee -a $OUTFILE_TBENCH.result
+
+		# get the base values
+		des_perf_base=`echo "$base" | awk '{print $3}' | sed s/,//`
+		freq_base=`echo "$base" | awk '{print $4}' | sed s/,//`
+		load_base=`echo "$base" | awk '{print $5}' | sed s/,//`
+		perf_base=`echo "$base" | awk '{print $6}' | sed s/,//`
+		energy_base=`echo "$base" | awk '{print $7}' | sed s/,//`
+		ppw_base=`echo "$base" | awk '{print $8}' | sed s/,//`
+
+		# get the comparative values
+		des_perf_comp=`echo "$comp" | awk '{print $3}' | sed s/,//`
+		freq_comp=`echo "$comp" | awk '{print $4}' | sed s/,//`
+		load_comp=`echo "$comp" | awk '{print $5}' | sed s/,//`
+		perf_comp=`echo "$comp" | awk '{print $6}' | sed s/,//`
+		energy_comp=`echo "$comp" | awk '{print $7}' | sed s/,//`
+		ppw_comp=`echo "$comp" | awk '{print $8}' | sed s/,//`
+
+		# compare the base and comp values
+		des_perf_drop=`echo "scale=4;($des_perf_comp-$des_perf_base)*100/$des_perf_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Tbench-$1 des perf base: $des_perf_base comprison: $des_perf_comp percent: $des_perf_drop\n" | tee -a $OUTFILE_TBENCH.result
+
+		freq_drop=`echo "scale=4;($freq_comp-$freq_base)*100/$freq_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Tbench-$1 freq base: $freq_base comprison: $freq_comp percent: $freq_drop\n" | tee -a $OUTFILE_TBENCH.result
+
+		load_drop=`echo "scale=4;($load_comp-$load_base)*100/$load_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Tbench-$1 load base: $load_base comprison: $load_comp percent: $load_drop\n" | tee -a $OUTFILE_TBENCH.result
+
+		perf_drop=`echo "scale=4;($perf_comp-$perf_base)*100/$perf_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Tbench-$1 perf base: $perf_base comprison: $perf_comp percent: $perf_drop\n" | tee -a $OUTFILE_TBENCH.result
+
+		energy_drop=`echo "scale=4;($energy_comp-$energy_base)*100/$energy_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Tbench-$1 energy base: $energy_base comprison: $energy_comp percent: $energy_drop\n" | tee -a $OUTFILE_TBENCH.result
+
+		ppw_drop=`echo "scale=4;($ppw_comp-$ppw_base)*100/$ppw_base" | bc | awk '{printf "%.4f", $0}'`
+		printf "Tbench-$1 performance per watt base: $ppw_base comprison: $ppw_comp percent: $ppw_drop\n" | tee -a $OUTFILE_TBENCH.result
+		printf "\n" | tee -a $OUTFILE_TBENCH.result
+
+		store_csv_tbench "$1-$2 VS $3-$4" "Comprison(%)" "$des_perf_drop" "$freq_drop" "$load_drop" "$perf_drop" "$energy_drop" "$ppw_drop"
+	fi
+}
+
+# calculate the comparison(%)
+calc_comp_tbench()
+{
+	# acpi-cpufreq-ondemand VS acpi-cpufreq-schedutil
+	__calc_comp_tbench ${all_scaling_names[0]} ${tbench_governors[0]} ${all_scaling_names[0]} ${tbench_governors[1]}
+
+	# amd-pstate-ondemand VS amd-pstate-schedutil
+	__calc_comp_tbench ${all_scaling_names[1]} ${tbench_governors[0]} ${all_scaling_names[1]} ${tbench_governors[1]}
+
+	# acpi-cpufreq-ondemand VS amd-pstate-ondemand
+	__calc_comp_tbench ${all_scaling_names[0]} ${tbench_governors[0]} ${all_scaling_names[1]} ${tbench_governors[0]}
+
+	# acpi-cpufreq-schedutil VS amd-pstate-schedutil
+	__calc_comp_tbench ${all_scaling_names[0]} ${tbench_governors[1]} ${all_scaling_names[1]} ${tbench_governors[1]}
+}
+
+# $1: file_name, $2: title, $3: ylable, $4: column
+plot_png_tbench()
+{
+	# all_scaling_names[1] all_scaling_names[0] flag
+	#    amd-pstate           acpi-cpufreq
+	#         N                   N             0
+	#         N                   Y             1
+	#         Y                   N             2
+	#         Y                   Y             3
+	ret=`grep -c "${all_scaling_names[1]}" $OUTFILE_TBENCH.csv`
+	if [ $ret -eq 0 ]; then
+		ret=`grep -c "${all_scaling_names[0]}" $OUTFILE_TBENCH.csv`
+		if [ $ret -eq 0 ]; then
+			flag=0
+		else
+			flag=1
+		fi
+	else
+		ret=`grep -c "${all_scaling_names[0]}" $OUTFILE_TBENCH.csv`
+		if [ $ret -eq 0 ]; then
+			flag=2
+		else
+			flag=3
+		fi
+	fi
+
+	gnuplot << EOF
+		set term png
+		set output "$1"
+
+		set title "$2"
+		set xlabel "Test Cycles (round)"
+		set ylabel "$3"
+
+		set grid
+		set style data histogram
+		set style fill solid 0.5 border
+		set boxwidth 0.8
+
+		if ($flag == 1) {
+			plot \
+			"<(sed -n -e 's/,//g' -e '/${all_scaling_names[0]}-${tbench_governors[0]}/p' $OUTFILE_TBENCH.csv)" using $4:xtic(2) title "${all_scaling_names[0]}-${tbench_governors[0]}", \
+			"<(sed -n -e 's/,//g' -e '/${all_scaling_names[0]}-${tbench_governors[1]}/p' $OUTFILE_TBENCH.csv)" using $4:xtic(2) title "${all_scaling_names[0]}-${tbench_governors[1]}"
+		} else {
+			if ($flag == 2) {
+				plot \
+				"<(sed -n -e 's/,//g' -e '/${all_scaling_names[1]}-${tbench_governors[0]}/p' $OUTFILE_TBENCH.csv)" using $4:xtic(2) title "${all_scaling_names[1]}-${tbench_governors[0]}", \
+				"<(sed -n -e 's/,//g' -e '/${all_scaling_names[1]}-${tbench_governors[1]}/p' $OUTFILE_TBENCH.csv)" using $4:xtic(2) title "${all_scaling_names[1]}-${tbench_governors[1]}"
+			} else {
+				if ($flag == 3 ) {
+					plot \
+					"<(sed -n -e 's/,//g' -e '/${all_scaling_names[0]}-${tbench_governors[0]}/p' $OUTFILE_TBENCH.csv)" using $4:xtic(2) title "${all_scaling_names[0]}-${tbench_governors[0]}", \
+					"<(sed -n -e 's/,//g' -e '/${all_scaling_names[0]}-${tbench_governors[1]}/p' $OUTFILE_TBENCH.csv)" using $4:xtic(2) title "${all_scaling_names[0]}-${tbench_governors[1]}", \
+					"<(sed -n -e 's/,//g' -e '/${all_scaling_names[1]}-${tbench_governors[0]}/p' $OUTFILE_TBENCH.csv)" using $4:xtic(2) title "${all_scaling_names[1]}-${tbench_governors[0]}", \
+					"<(sed -n -e 's/,//g' -e '/${all_scaling_names[1]}-${tbench_governors[1]}/p' $OUTFILE_TBENCH.csv)" using $4:xtic(2) title "${all_scaling_names[1]}-${tbench_governors[1]}"
+				}
+			}
+		}
+		quit
+EOF
+}
+
+amd_pstate_tbench()
+{
+	printf "\n---------------------------------------------\n"
+	printf "*** Running tbench                        ***"
+	printf "\n---------------------------------------------\n"
+
+	pre_clear_tbench
+
+	get_lines_csv_tbench "Governor"
+	if [ $? -eq 0 ]; then
+		# add titles and unit for csv file
+		store_csv_tbench "Governor" "Round" "Des-perf" "Freq" "Load" "Performance" "Energy" "Performance Per Watt"
+		store_csv_tbench "Unit" "" "" "GHz" "" "MB/s" "J" "MB/J"
+	fi
+
+	backup_governor
+	for governor in ${tbench_governors[*]} ; do
+		printf "\nSpecified governor is $governor\n\n"
+		switch_governor $governor
+		loop_tbench $governor
+		gather_tbench $governor
+	done
+	restore_governor
+
+	plot_png_tbench "tbench_perfromance.png" "Tbench Benchmark Performance" "Performance" 6
+	plot_png_tbench "tbench_energy.png" "Tbench Benchmark Energy" "Energy (J)" 7
+	plot_png_tbench "tbench_ppw.png" "Tbench Benchmark Performance Per Watt" "Performance Per Watt (MB/J)" 8
+
+	calc_comp_tbench
+
+	post_clear_tbench
+}