perf scripting python: intel-pt-events.py: Add --insn-trace and --src-trace

Add an instruction trace and a source trace to the intel-pt-events.py
script.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: https://lore.kernel.org/r/20210530192308.7382-14-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools/perf/Documentation/perf-intel-pt.txt

@@ -174,7 +174,11 @@ Refer to script export-to-sqlite.py or export-to-postgresql.py for more details,
 and to script exported-sql-viewer.py for an example of using the database.
 
 There is also script intel-pt-events.py which provides an example of how to
-unpack the raw data for power events and PTWRITE.
+unpack the raw data for power events and PTWRITE. The script also displays
+branches, and supports 2 additional modes selected by option:
+
+ --insn-trace - instruction trace
+ --src-trace - source trace
 
 As mentioned above, it is easy to capture too much data.  One way to limit the
 data captured is to use 'snapshot' mode which is explained further below.

tools/perf/scripts/python/intel-pt-events.py

@@ -16,13 +16,16 @@ from __future__ import print_function
 import os
 import sys
 import struct
+import argparse
+
+from libxed import LibXED
+from ctypes import create_string_buffer, addressof
 
 sys.path.append(os.environ['PERF_EXEC_PATH'] + \
 	'/scripts/python/Perf-Trace-Util/lib/Perf/Trace')
 
-# These perf imports are not used at present
+from perf_trace_context import perf_set_itrace_options, \
-#from perf_trace_context import *
+	perf_sample_insn, perf_sample_srccode
-#from Core import *
 
 try:
 	broken_pipe_exception = BrokenPipeError
@@ -31,6 +34,12 @@ except:
 
 glb_switch_str		= None
 glb_switch_printed	= True
+glb_insn		= False
+glb_disassembler	= None
+glb_src			= False
+glb_source_file_name	= None
+glb_line_number		= None
+glb_dso			= None
 
 def get_optional_null(perf_dict, field):
 	if field in perf_dict:
@@ -42,6 +51,11 @@ def get_optional_zero(perf_dict, field):
 		return perf_dict[field]
 	return 0
 
+def get_optional_bytes(perf_dict, field):
+	if field in perf_dict:
+		return perf_dict[field]
+	return bytes()
+
 def get_optional(perf_dict, field):
 	if field in perf_dict:
 		return perf_dict[field]
@@ -53,7 +67,31 @@ def get_offset(perf_dict, field):
 	return ""
 
 def trace_begin():
+	ap = argparse.ArgumentParser(usage = "", add_help = False)
+	ap.add_argument("--insn-trace", action='store_true')
+	ap.add_argument("--src-trace", action='store_true')
+	global glb_args
+	global glb_insn
+	global glb_src
+	glb_args = ap.parse_args()
+	if glb_args.insn_trace:
+		print("Intel PT Instruction Trace")
+		itrace = "i0nsepwx"
+		glb_insn = True
+	elif glb_args.src_trace:
+		print("Intel PT Source Trace")
+		itrace = "i0nsepwx"
+		glb_insn = True
+		glb_src = True
+	else:
 		print("Intel PT Branch Trace, Power Events and PTWRITE")
+		itrace = "bepwx"
+	global glb_disassembler
+	try:
+		glb_disassembler = LibXED()
+	except:
+		glb_disassembler = None
+	perf_set_itrace_options(perf_script_context, itrace)
 
 def trace_end():
 	print("End")
@@ -111,11 +149,14 @@ def print_psb(raw_buf):
 	offset = data[1]
 	print("offset: %#x" % (offset), end=' ')
 
-def print_common_start(comm, sample, name):
+def common_start_str(comm, sample):
 	ts = sample["time"]
 	cpu = sample["cpu"]
 	pid = sample["pid"]
 	tid = sample["tid"]
+	return "%16s %5u/%-5u [%03u] %9u.%09u  " % (comm, pid, tid, cpu, ts / 1000000000, ts %1000000000)
+
+def print_common_start(comm, sample, name):
 	flags_disp = get_optional_null(sample, "flags_disp")
 	# Unused fields:
 	# period      = sample["period"]
@@ -123,22 +164,96 @@ def print_common_start(comm, sample, name):
 	# weight      = sample["weight"]
 	# transaction = sample["transaction"]
 	# cpumode     = get_optional_zero(sample, "cpumode")
-	print("%16s %5u/%-5u [%03u] %9u.%09u  %7s  %19s" %
+	print(common_start_str(comm, sample) + "%7s  %19s" % (name, flags_disp), end=' ')
-		(comm, pid, tid, cpu, ts / 1000000000, ts %1000000000, name, flags_disp),
+
-		end=' ')
+def print_instructions_start(comm, sample):
+	if "x" in get_optional_null(sample, "flags"):
+		print(common_start_str(comm, sample) + "x", end=' ')
+	else:
+		print(common_start_str(comm, sample), end='  ')
+
+def disassem(insn, ip):
+	inst = glb_disassembler.Instruction()
+	glb_disassembler.SetMode(inst, 0) # Assume 64-bit
+	buf = create_string_buffer(64)
+	buf.value = insn
+	return glb_disassembler.DisassembleOne(inst, addressof(buf), len(insn), ip)
 
 def print_common_ip(param_dict, sample, symbol, dso):
 	ip   = sample["ip"]
 	offs = get_offset(param_dict, "symoff")
+	if "cyc_cnt" in sample:
+		cyc_cnt = sample["cyc_cnt"]
+		insn_cnt = get_optional_zero(sample, "insn_cnt")
+		ipc_str = "  IPC: %#.2f (%u/%u)" % (insn_cnt / cyc_cnt, insn_cnt, cyc_cnt)
+	else:
+		ipc_str = ""
+	if glb_insn and glb_disassembler is not None:
+		insn = perf_sample_insn(perf_script_context)
+		if insn and len(insn):
+			cnt, text = disassem(insn, ip)
+			byte_str = ("%x" % ip).rjust(16)
+			if sys.version_info.major >= 3:
+				for k in range(cnt):
+					byte_str += " %02x" % insn[k]
+			else:
+				for k in xrange(cnt):
+					byte_str += " %02x" % ord(insn[k])
+			print("%-40s  %-30s" % (byte_str, text), end=' ')
+		print("%s%s (%s)" % (symbol, offs, dso), end=' ')
+	else:
 		print("%16x %s%s (%s)" % (ip, symbol, offs, dso), end=' ')
 	if "addr_correlates_sym" in sample:
 		addr   = sample["addr"]
 		dso    = get_optional(sample, "addr_dso")
 		symbol = get_optional(sample, "addr_symbol")
 		offs   = get_offset(sample, "addr_symoff")
-		print("=> %x %s%s (%s)" % (addr, symbol, offs, dso))
+		print("=> %x %s%s (%s)%s" % (addr, symbol, offs, dso, ipc_str))
 	else:
-		print()
+		print(ipc_str)
+
+def print_srccode(comm, param_dict, sample, symbol, dso, with_insn):
+	ip = sample["ip"]
+	if symbol == "[unknown]":
+		start_str = common_start_str(comm, sample) + ("%x" % ip).rjust(16).ljust(40)
+	else:
+		offs = get_offset(param_dict, "symoff")
+		start_str = common_start_str(comm, sample) + (symbol + offs).ljust(40)
+
+	if with_insn and glb_insn and glb_disassembler is not None:
+		insn = perf_sample_insn(perf_script_context)
+		if insn and len(insn):
+			cnt, text = disassem(insn, ip)
+		start_str += text.ljust(30)
+
+	global glb_source_file_name
+	global glb_line_number
+	global glb_dso
+
+	source_file_name, line_number, source_line = perf_sample_srccode(perf_script_context)
+	if source_file_name:
+		if glb_line_number == line_number and glb_source_file_name == source_file_name:
+			src_str = ""
+		else:
+			if len(source_file_name) > 40:
+				src_file = ("..." + source_file_name[-37:]) + " "
+			else:
+				src_file = source_file_name.ljust(41)
+			if source_line is None:
+				src_str = src_file + str(line_number).rjust(4) + " <source not found>"
+			else:
+				src_str = src_file + str(line_number).rjust(4) + " " + source_line
+		glb_dso = None
+	elif dso == glb_dso:
+		src_str = ""
+	else:
+		src_str = dso
+		glb_dso = dso
+
+	glb_line_number = line_number
+	glb_source_file_name = source_file_name
+
+	print(start_str, src_str)
 
 def do_process_event(param_dict):
 	global glb_switch_printed
@@ -159,23 +274,48 @@ def do_process_event(param_dict):
 	dso    = get_optional(param_dict, "dso")
 	symbol = get_optional(param_dict, "symbol")
 
+	if name[0:12] == "instructions":
+		if glb_src:
+			print_srccode(comm, param_dict, sample, symbol, dso, True)
+		else:
+			print_instructions_start(comm, sample)
+			print_common_ip(param_dict, sample, symbol, dso)
+	elif name[0:8] == "branches":
+		if glb_src:
+			print_srccode(comm, param_dict, sample, symbol, dso, False)
+		else:
+			print_common_start(comm, sample, name)
+			print_common_ip(param_dict, sample, symbol, dso)
+	elif name == "ptwrite":
 		print_common_start(comm, sample, name)
-
-	if name == "ptwrite":
 		print_ptwrite(raw_buf)
+		print_common_ip(param_dict, sample, symbol, dso)
 	elif name == "cbr":
+		print_common_start(comm, sample, name)
 		print_cbr(raw_buf)
+		print_common_ip(param_dict, sample, symbol, dso)
 	elif name == "mwait":
+		print_common_start(comm, sample, name)
 		print_mwait(raw_buf)
+		print_common_ip(param_dict, sample, symbol, dso)
 	elif name == "pwre":
+		print_common_start(comm, sample, name)
 		print_pwre(raw_buf)
+		print_common_ip(param_dict, sample, symbol, dso)
 	elif name == "exstop":
+		print_common_start(comm, sample, name)
 		print_exstop(raw_buf)
+		print_common_ip(param_dict, sample, symbol, dso)
 	elif name == "pwrx":
+		print_common_start(comm, sample, name)
 		print_pwrx(raw_buf)
+		print_common_ip(param_dict, sample, symbol, dso)
 	elif name == "psb":
+		print_common_start(comm, sample, name)
 		print_psb(raw_buf)
-
+		print_common_ip(param_dict, sample, symbol, dso)
+	else:
+		print_common_start(comm, sample, name)
 		print_common_ip(param_dict, sample, symbol, dso)
 
 def process_event(param_dict):