| 1 | #!/usr/bin/env python |
| 2 | |
| 3 | import os, sys |
| 4 | import subprocess |
| 5 | import threading |
| 6 | import Queue |
| 7 | import time |
| 8 | |
| 9 | from signal import signal, SIGTERM, SIGINT |
| 10 | |
| 11 | SESSIOND_BIN_NAME = "lttng-sessiond" |
| 12 | SESSIOND_BIN_PATH = "src/bin/lttng-sessiond/.libs/" |
| 13 | TESTDIR_PATH = "" |
| 14 | |
| 15 | PRINT_BRACKET = "\033[1;34m[\033[1;33m+\033[1;34m]\033[00m" |
| 16 | PRINT_RED_BRACKET = "\033[1;31m[+]\033[00m" |
| 17 | PRINT_GREEN_BRACKET = "\033[1;32m[+]\033[00m" |
| 18 | PRINT_ARROW = "\033[1;32m-->\033[00m" |
| 19 | |
| 20 | is_root = 1 |
| 21 | no_stats = 0 |
| 22 | stop_sampling = 1 |
| 23 | |
| 24 | top_cpu_legend = { 'us': "User CPU time", 'sy': "System CPU time", |
| 25 | 'id': "Idle CPU time", 'ni': "Nice CPU time", 'wa': "iowait", |
| 26 | 'hi': "Hardware IRQ", 'si': "Software Interrupts", 'st': "Steal Time", } |
| 27 | |
| 28 | cpu_ret_q = Queue.Queue() |
| 29 | mem_ret_q = Queue.Queue() |
| 30 | test_ret_q = Queue.Queue() |
| 31 | |
| 32 | global sdaemon_proc |
| 33 | global worker_proc |
| 34 | |
| 35 | def cpu_create_usage_dict(top_line): |
| 36 | """ |
| 37 | Return a dictionnary from a 'top' cpu line. |
| 38 | Ex: Cpu(s): 2.1%us, 1.2%sy, 0.0%ni, 96.2%id, 0.4%wa, 0.0%hi, 0.0%si, 0.0%st |
| 39 | """ |
| 40 | top_dict = {'us': 0, 'sy': 0, 'ni': 0, 'id': 0, 'wa': 0, 'hi': 0, 'si': 0, 'st': 0} |
| 41 | |
| 42 | # Split expression and remove first value which is "Cpu(s)" |
| 43 | top_line = top_line.replace(",","") |
| 44 | words = top_line.split()[1:] |
| 45 | |
| 46 | for word in words: |
| 47 | index = word.find('%') |
| 48 | # Add the value to the dictionnary |
| 49 | top_dict[word[index + 1:]] = float(word[:index]) |
| 50 | |
| 51 | return top_dict |
| 52 | |
| 53 | def cpu_average_usage(top_lines): |
| 54 | """ |
| 55 | Return a dictionnary of 'top' CPU stats but averaging all values. |
| 56 | """ |
| 57 | avg_dict = {'us': 0, 'sy': 0, 'ni': 0, 'id': 0, 'wa': 0, 'hi': 0, 'si': 0, 'st': 0} |
| 58 | # Average count |
| 59 | count = 0.0 |
| 60 | |
| 61 | for line in top_lines: |
| 62 | tmp_dict = cpu_create_usage_dict(line) |
| 63 | # Add value to avg dictionnary |
| 64 | for key in tmp_dict: |
| 65 | avg_dict[key] += tmp_dict[key] |
| 66 | |
| 67 | count += 1.0 |
| 68 | |
| 69 | for key in avg_dict: |
| 70 | avg_dict[key] = avg_dict[key] / count |
| 71 | |
| 72 | return (count, avg_dict) |
| 73 | |
| 74 | def cpu_sample_usage(pid=None): |
| 75 | """ |
| 76 | Sample CPU usage for num iterations. |
| 77 | If num is greater than 1, the average will be computed. |
| 78 | """ |
| 79 | args = ["top", "-b", "-n", "1"] |
| 80 | if pid: |
| 81 | args.append("-p") |
| 82 | args.append(str(pid)) |
| 83 | |
| 84 | # Spawn top process |
| 85 | top = subprocess.Popen(args, stdout = subprocess.PIPE) |
| 86 | |
| 87 | grep = subprocess.Popen(["grep", "^Cpu"], stdin = top.stdout, |
| 88 | stdout = subprocess.PIPE) |
| 89 | top.stdout.close() |
| 90 | |
| 91 | return grep.communicate()[0].strip("\n") |
| 92 | |
| 93 | def mem_sample_usage(pid): |
| 94 | """ |
| 95 | Sample memory usage using /proc and a pid |
| 96 | """ |
| 97 | args = ["cat", "/proc/" + str(pid) + "/status"] |
| 98 | |
| 99 | if not os.path.isfile(args[1]): |
| 100 | return -1 |
| 101 | |
| 102 | mem_proc = subprocess.Popen(args, stdout = subprocess.PIPE) |
| 103 | |
| 104 | grep = subprocess.Popen(["grep", "^VmRSS"], stdin = mem_proc.stdout, |
| 105 | stdout = subprocess.PIPE) |
| 106 | mem_proc.stdout.close() |
| 107 | |
| 108 | # Return virtual memory size in kilobytes (kB) |
| 109 | #ret = grep.communicate()[0].split() |
| 110 | ret = grep.communicate()[0].split() |
| 111 | |
| 112 | if len(ret) > 1: |
| 113 | ret = ret[1] |
| 114 | else: |
| 115 | ret = 0 |
| 116 | |
| 117 | return int(ret) |
| 118 | |
| 119 | class SamplingWorker(threading.Thread): |
| 120 | def __init__(self, s_type, worker = None, delay = 0.2, pid = 0): |
| 121 | threading.Thread.__init__ (self) |
| 122 | self.s_type = s_type |
| 123 | self.delay = delay |
| 124 | self.pid = pid |
| 125 | self.worker = worker |
| 126 | |
| 127 | def run(self): |
| 128 | count = 1 |
| 129 | lines = [] |
| 130 | |
| 131 | if self.s_type == "cpu": |
| 132 | while 1: |
| 133 | if self.worker == None: |
| 134 | cpu_line = cpu_sample_usage(self.pid) |
| 135 | lines.append(cpu_line) |
| 136 | break |
| 137 | elif self.worker.is_alive(): |
| 138 | cpu_line = cpu_sample_usage(self.pid) |
| 139 | lines.append(cpu_line) |
| 140 | else: |
| 141 | break |
| 142 | |
| 143 | # Delay sec per memory sampling |
| 144 | time.sleep(self.delay) |
| 145 | |
| 146 | count, stats = cpu_average_usage(lines) |
| 147 | cpu_ret_q.put((count, stats)) |
| 148 | # grep process has ended here |
| 149 | |
| 150 | elif self.s_type == "mem": |
| 151 | count = 0 |
| 152 | mem_stat = 0 |
| 153 | |
| 154 | while 1: |
| 155 | if self.worker == None: |
| 156 | cpu_line = cpu_sample_usage(self.pid) |
| 157 | lines.append(cpu_line) |
| 158 | break |
| 159 | elif self.worker.is_alive(): |
| 160 | mem_stat += get_mem_usage(self.pid) |
| 161 | count += 1 |
| 162 | else: |
| 163 | break |
| 164 | |
| 165 | # Delay sec per memory sampling |
| 166 | time.sleep(self.delay) |
| 167 | |
| 168 | mem_ret_q.put((count, mem_stat)) |
| 169 | |
| 170 | class TestWorker(threading.Thread): |
| 171 | def __init__(self, path, name): |
| 172 | threading.Thread.__init__(self) |
| 173 | self.path = path |
| 174 | self.name = name |
| 175 | |
| 176 | def run(self): |
| 177 | bin_path_name = os.path.join(self.path, self.name) |
| 178 | |
| 179 | env = os.environ |
| 180 | env['TEST_NO_SESSIOND'] = '1' |
| 181 | |
| 182 | test = subprocess.Popen([bin_path_name], env=env) |
| 183 | test.wait() |
| 184 | |
| 185 | # Send ret value to main thread |
| 186 | test_ret_q.put(test.returncode) |
| 187 | |
| 188 | def get_pid(procname): |
| 189 | """ |
| 190 | Return pid of process name using 'pidof' command |
| 191 | """ |
| 192 | pidof = subprocess.Popen(["pidof", procname], stdout = subprocess.PIPE) |
| 193 | pid = pidof.communicate()[0].split() |
| 194 | |
| 195 | if pid == []: |
| 196 | return 0 |
| 197 | |
| 198 | return int(pid[0]) |
| 199 | |
| 200 | def spawn_session_daemon(): |
| 201 | """ |
| 202 | Exec the session daemon and return PID |
| 203 | """ |
| 204 | global sdaemon_proc |
| 205 | |
| 206 | pid = get_pid(SESSIOND_BIN_NAME) |
| 207 | if pid != 0: |
| 208 | os.kill(pid, SIGTERM) |
| 209 | |
| 210 | bin_path = os.path.join(TESTDIR_PATH, "..", SESSIOND_BIN_PATH, SESSIOND_BIN_NAME) |
| 211 | |
| 212 | if not os.path.isfile(bin_path): |
| 213 | print "Error: No session daemon binary found. Compiled?" |
| 214 | return 0 |
| 215 | |
| 216 | try: |
| 217 | sdaemon_proc = subprocess.Popen([bin_path, "-d"], shell=False, |
| 218 | stderr = subprocess.PIPE) |
| 219 | except OSError, e: |
| 220 | print e |
| 221 | return 0 |
| 222 | |
| 223 | return get_pid(SESSIOND_BIN_NAME) |
| 224 | |
| 225 | def start_test(name): |
| 226 | """ |
| 227 | Spawn test and return exit code |
| 228 | """ |
| 229 | tw = TestWorker(".", name) |
| 230 | tw.start() |
| 231 | |
| 232 | return test_ret_q.get(True) |
| 233 | |
| 234 | def print_cpu_stats(stats, count): |
| 235 | """ |
| 236 | Pretty print on one line the CPU stats |
| 237 | """ |
| 238 | sys.stdout.write(PRINT_ARROW + " Cpu [sampled %d time(s)]:\n " % (count)) |
| 239 | for stat in stats: |
| 240 | sys.stdout.write(" %s: %.2f, " % (stat, stats[stat])) |
| 241 | print "" |
| 242 | |
| 243 | def get_cpu_usage(delay=1, pid=0): |
| 244 | """ |
| 245 | Spawn a worker thread to sample cpu usage. |
| 246 | """ |
| 247 | sw = SamplingWorker("cpu", delay = delay, pid = pid) |
| 248 | sw.start() |
| 249 | |
| 250 | return cpu_ret_q.get(True) |
| 251 | |
| 252 | def get_mem_usage(pid): |
| 253 | """ |
| 254 | Get memory usage for PID |
| 255 | """ |
| 256 | return mem_sample_usage(pid) |
| 257 | |
| 258 | def print_test_success(ret, expect): |
| 259 | """ |
| 260 | Print if test has failed or pass according to the expected value. |
| 261 | """ |
| 262 | if ret != expect: |
| 263 | print "\n" + PRINT_RED_BRACKET + \ |
| 264 | " Failed: ret = %d (expected %d)" % (ret, expect) |
| 265 | return 1 |
| 266 | else: |
| 267 | print "\n" + PRINT_BRACKET + \ |
| 268 | " Passed: ret = %d (expected %d)" % (ret, expect) |
| 269 | return 0 |
| 270 | |
| 271 | def run_test(test): |
| 272 | """ |
| 273 | Run test 'name' and output report of the test with stats. |
| 274 | """ |
| 275 | global worker_proc |
| 276 | global sdaemon_proc |
| 277 | dem_pid = 0 # Session daemon pid |
| 278 | |
| 279 | print PRINT_BRACKET + " %s" % (test['name']) |
| 280 | print PRINT_ARROW + " %s" % (test['desc']) |
| 281 | if no_stats: |
| 282 | print PRINT_ARROW + " Statistics will NOT be collected" |
| 283 | else: |
| 284 | print PRINT_ARROW + " Statistics of the session daemon will be collected" |
| 285 | |
| 286 | if test['kern'] and not is_root: |
| 287 | print "Needs root for kernel tracing. Skipping" |
| 288 | return 0 |
| 289 | |
| 290 | if not os.path.isfile(test['bin']): |
| 291 | print "Unable to find test file '%s'. Skipping" % (test['bin']) |
| 292 | return 0 |
| 293 | |
| 294 | # No session daemon needed |
| 295 | if not test['daemon']: |
| 296 | print PRINT_ARROW + " No session daemon needed" |
| 297 | ret = start_test(test['bin']) |
| 298 | print_test_success(ret, test['success']) |
| 299 | return 0 |
| 300 | else: |
| 301 | print PRINT_ARROW + " Session daemon needed" |
| 302 | |
| 303 | dem_pid = spawn_session_daemon() |
| 304 | if dem_pid <= 0: |
| 305 | print "Unable to start %s. Stopping" % (SESSIOND_BIN_NAME) |
| 306 | print sdaemon_proc.communicate()[1] |
| 307 | return 0 |
| 308 | |
| 309 | print PRINT_BRACKET + " Session daemon spawned (pid: %d)\n" % (dem_pid) |
| 310 | |
| 311 | if not no_stats: |
| 312 | mem_before = get_mem_usage(dem_pid) |
| 313 | print PRINT_BRACKET + " Stats *before* test:" |
| 314 | print PRINT_ARROW + " Mem (kB): %d" % (mem_before) |
| 315 | cpu_count, cpu_stats = get_cpu_usage(pid = dem_pid) |
| 316 | print_cpu_stats(cpu_stats, cpu_count) |
| 317 | |
| 318 | tw = TestWorker(".", test['bin']) |
| 319 | tw.start() |
| 320 | |
| 321 | if not no_stats: |
| 322 | # Start CPU sampling for test |
| 323 | sw_cpu = SamplingWorker("cpu", worker = tw, pid = dem_pid) |
| 324 | sw_cpu.start() |
| 325 | sw_mem = SamplingWorker("mem", worker = tw, pid = dem_pid) |
| 326 | sw_mem.start() |
| 327 | |
| 328 | ret = test_ret_q.get(True) |
| 329 | |
| 330 | if not no_stats: |
| 331 | time.sleep(2) |
| 332 | # Compute memory average |
| 333 | mem_count, mem_during = mem_ret_q.get(True) |
| 334 | mem_during = float(mem_during) / float(mem_count) |
| 335 | cpu_count, cpu_stats = cpu_ret_q.get(True) |
| 336 | |
| 337 | print "\n" + PRINT_BRACKET + " Stats *during* test:" |
| 338 | print PRINT_ARROW + " Mem (kB): %.0f [sampled %d time(s)]" % (mem_during, mem_count) |
| 339 | print_cpu_stats(cpu_stats, cpu_count) |
| 340 | |
| 341 | mem_after = get_mem_usage(dem_pid) |
| 342 | print "\n" + PRINT_BRACKET + " Stats *after* test:" |
| 343 | print PRINT_ARROW + " Mem (kB): %d" % (mem_after) |
| 344 | cpu_count, cpu_stats = get_cpu_usage(pid = dem_pid) |
| 345 | print_cpu_stats(cpu_stats, cpu_count) |
| 346 | |
| 347 | print "\n" + PRINT_BRACKET + " Memory usage differences:" |
| 348 | print PRINT_ARROW + " Diff during and before (kB): %d" % (mem_during - mem_before) |
| 349 | print PRINT_ARROW + " Diff during and after (kB): %d" % (mem_during - mem_after) |
| 350 | print PRINT_ARROW + " Diff before and after (kB): %d" % (mem_after - mem_before) |
| 351 | |
| 352 | # Return value of 0 means that is passed else it failed |
| 353 | ret = print_test_success(ret, test['success']) |
| 354 | |
| 355 | # Stop session daemon |
| 356 | if dem_pid > 0: |
| 357 | print PRINT_BRACKET + " Stopping session daemon (pid: %d)..." % (dem_pid) |
| 358 | try: |
| 359 | os.kill(dem_pid, SIGTERM) |
| 360 | # This call simply does not work... It seems python does not relay the signal |
| 361 | # to the child processes of sdaemon_proc. |
| 362 | # sdaemon_proc.terminate() |
| 363 | if ret != 0: |
| 364 | print sdaemon_proc.communicate()[1] |
| 365 | elif sdaemon_proc.returncode == None: |
| 366 | sdaemon_proc.communicate() |
| 367 | except OSError, e: |
| 368 | print e |
| 369 | |
| 370 | # Make sure all thread are released |
| 371 | if not no_stats: |
| 372 | tw.join() |
| 373 | sw_cpu.join() |
| 374 | sw_mem.join() |
| 375 | |
| 376 | return ret |
| 377 | |
| 378 | def main(): |
| 379 | for test in Tests: |
| 380 | if not test['enabled']: |
| 381 | continue |
| 382 | |
| 383 | ret = run_test(test) |
| 384 | if ret != 0: |
| 385 | # Stop all tests, the last one failed |
| 386 | return |
| 387 | print "" |
| 388 | |
| 389 | def cleanup(signo, stack): |
| 390 | """ Cleanup function """ |
| 391 | sys.exit(0) |
| 392 | |
| 393 | if __name__ == "__main__": |
| 394 | if not os.getuid() == 0: |
| 395 | is_root = 0 |
| 396 | print "NOTICE: Not root. No kernel tracing will be tested\n" |
| 397 | |
| 398 | if os.path.isfile("test_list.py"): |
| 399 | from test_list import Tests |
| 400 | else: |
| 401 | print "No test_list.py found. Stopping" |
| 402 | cleanup(0, 0) |
| 403 | |
| 404 | TESTDIR_PATH = os.getcwd() |
| 405 | |
| 406 | if len(sys.argv) > 1: |
| 407 | if sys.argv[1] == "--no-stats": |
| 408 | no_stats = 1 |
| 409 | |
| 410 | try: |
| 411 | signal(SIGTERM, cleanup) |
| 412 | signal(SIGINT, cleanup) |
| 413 | main() |
| 414 | cleanup(0, 0) |
| 415 | except KeyboardInterrupt: |
| 416 | cleanup(0, 0) |