1#!/usr/bin/env python 2# 3# Copyright (c) 2017 Red Hat Inc 4# 5# Author: 6# Eduardo Habkost <ehabkost@redhat.com> 7# 8# This program is free software; you can redistribute it and/or modify 9# it under the terms of the GNU General Public License as published by 10# the Free Software Foundation; either version 2 of the License, or 11# (at your option) any later version. 12# 13# This program is distributed in the hope that it will be useful, 14# but WITHOUT ANY WARRANTY; without even the implied warranty of 15# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 16# GNU General Public License for more details. 17# 18# You should have received a copy of the GNU General Public License along 19# with this program; if not, write to the Free Software Foundation, Inc., 20# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. 21 22""" 23Run QEMU with all combinations of -machine and -device types, 24check for crashes and unexpected errors. 25""" 26from __future__ import print_function 27 28import os 29import sys 30import glob 31import logging 32import traceback 33import re 34import random 35import argparse 36from itertools import chain 37 38sys.path.append(os.path.join(os.path.dirname(__file__), '..', 'python')) 39from qemu.machine import QEMUMachine 40 41logger = logging.getLogger('device-crash-test') 42dbg = logger.debug 43 44 45# Purposes of the following whitelist: 46# * Avoiding verbose log messages when we find known non-fatal 47# (exitcode=1) errors 48# * Avoiding fatal errors when we find known crashes 49# * Skipping machines/devices that are known not to work out of 50# the box, when running in --quick mode 51# 52# Keeping the whitelist updated is desirable, but not required, 53# because unexpected cases where QEMU exits with exitcode=1 will 54# just trigger a INFO message. 55 56# Valid whitelist entry keys: 57# * accel: regexp, full match only 58# * machine: regexp, full match only 59# * device: regexp, full match only 60# * log: regexp, partial match allowed 61# * exitcode: if not present, defaults to 1. If None, matches any exitcode 62# * warn: if True, matching failures will be logged as warnings 63# * expected: if True, QEMU is expected to always fail every time 64# when testing the corresponding test case 65# * loglevel: log level of log output when there's a match. 66ERROR_WHITELIST = [ 67 # Machines that won't work out of the box: 68 # MACHINE | ERROR MESSAGE 69 {'machine':'niagara', 'expected':True}, # Unable to load a firmware for -M niagara 70 {'machine':'boston', 'expected':True}, # Please provide either a -kernel or -bios argument 71 {'machine':'leon3_generic', 'expected':True}, # Can't read bios image (null) 72 73 # devices that don't work out of the box because they require extra options to "-device DEV": 74 # DEVICE | ERROR MESSAGE 75 {'device':'.*-(i386|x86_64)-cpu', 'expected':True}, # CPU socket-id is not set 76 {'device':'icp', 'expected':True}, # icp_realize: required link 'xics' not found: Property '.xics' not found 77 {'device':'ics', 'expected':True}, # ics_base_realize: required link 'xics' not found: Property '.xics' not found 78 # "-device ide-cd" does work on more recent QEMU versions, so it doesn't have expected=True 79 {'device':'ide-cd'}, # No drive specified 80 {'device':'ide-drive', 'expected':True}, # No drive specified 81 {'device':'ide-hd', 'expected':True}, # No drive specified 82 {'device':'ipmi-bmc-extern', 'expected':True}, # IPMI external bmc requires chardev attribute 83 {'device':'isa-debugcon', 'expected':True}, # Can't create serial device, empty char device 84 {'device':'isa-ipmi-bt', 'expected':True}, # IPMI device requires a bmc attribute to be set 85 {'device':'isa-ipmi-kcs', 'expected':True}, # IPMI device requires a bmc attribute to be set 86 {'device':'isa-parallel', 'expected':True}, # Can't create serial device, empty char device 87 {'device':'ivshmem-doorbell', 'expected':True}, # You must specify a 'chardev' 88 {'device':'ivshmem-plain', 'expected':True}, # You must specify a 'memdev' 89 {'device':'loader', 'expected':True}, # please include valid arguments 90 {'device':'nand', 'expected':True}, # Unsupported NAND block size 0x1 91 {'device':'nvdimm', 'expected':True}, # 'memdev' property is not set 92 {'device':'nvme', 'expected':True}, # Device initialization failed 93 {'device':'pc-dimm', 'expected':True}, # 'memdev' property is not set 94 {'device':'pci-bridge', 'expected':True}, # Bridge chassis not specified. Each bridge is required to be assigned a unique chassis id > 0. 95 {'device':'pci-bridge-seat', 'expected':True}, # Bridge chassis not specified. Each bridge is required to be assigned a unique chassis id > 0. 96 {'device':'pxb', 'expected':True}, # Bridge chassis not specified. Each bridge is required to be assigned a unique chassis id > 0. 97 {'device':'scsi-block', 'expected':True}, # drive property not set 98 {'device':'scsi-disk', 'expected':True}, # drive property not set 99 {'device':'scsi-generic', 'expected':True}, # drive property not set 100 {'device':'scsi-hd', 'expected':True}, # drive property not set 101 {'device':'spapr-pci-host-bridge', 'expected':True}, # BUID not specified for PHB 102 {'device':'spapr-rng', 'expected':True}, # spapr-rng needs an RNG backend! 103 {'device':'spapr-vty', 'expected':True}, # chardev property not set 104 {'device':'tpm-tis', 'expected':True}, # tpm_tis: backend driver with id (null) could not be found 105 {'device':'unimplemented-device', 'expected':True}, # property 'size' not specified or zero 106 {'device':'usb-braille', 'expected':True}, # Property chardev is required 107 {'device':'usb-mtp', 'expected':True}, # rootdir property must be configured 108 {'device':'usb-redir', 'expected':True}, # Parameter 'chardev' is missing 109 {'device':'usb-serial', 'expected':True}, # Property chardev is required 110 {'device':'usb-storage', 'expected':True}, # drive property not set 111 {'device':'vfio-amd-xgbe', 'expected':True}, # -device vfio-amd-xgbe: vfio error: wrong host device name 112 {'device':'vfio-calxeda-xgmac', 'expected':True}, # -device vfio-calxeda-xgmac: vfio error: wrong host device name 113 {'device':'vfio-pci', 'expected':True}, # No provided host device 114 {'device':'vfio-pci-igd-lpc-bridge', 'expected':True}, # VFIO dummy ISA/LPC bridge must have address 1f.0 115 {'device':'vhost-scsi.*', 'expected':True}, # vhost-scsi: missing wwpn 116 {'device':'vhost-vsock-device', 'expected':True}, # guest-cid property must be greater than 2 117 {'device':'vhost-vsock-pci', 'expected':True}, # guest-cid property must be greater than 2 118 {'device':'virtio-9p-ccw', 'expected':True}, # 9pfs device couldn't find fsdev with the id = NULL 119 {'device':'virtio-9p-device', 'expected':True}, # 9pfs device couldn't find fsdev with the id = NULL 120 {'device':'virtio-9p-pci', 'expected':True}, # 9pfs device couldn't find fsdev with the id = NULL 121 {'device':'virtio-blk-ccw', 'expected':True}, # drive property not set 122 {'device':'virtio-blk-device', 'expected':True}, # drive property not set 123 {'device':'virtio-blk-device', 'expected':True}, # drive property not set 124 {'device':'virtio-blk-pci', 'expected':True}, # drive property not set 125 {'device':'virtio-crypto-ccw', 'expected':True}, # 'cryptodev' parameter expects a valid object 126 {'device':'virtio-crypto-device', 'expected':True}, # 'cryptodev' parameter expects a valid object 127 {'device':'virtio-crypto-pci', 'expected':True}, # 'cryptodev' parameter expects a valid object 128 {'device':'virtio-input-host-device', 'expected':True}, # evdev property is required 129 {'device':'virtio-input-host-pci', 'expected':True}, # evdev property is required 130 {'device':'xen-pvdevice', 'expected':True}, # Device ID invalid, it must always be supplied 131 {'device':'vhost-vsock-ccw', 'expected':True}, # guest-cid property must be greater than 2 132 {'device':'zpci', 'expected':True}, # target must be defined 133 {'device':'pnv-(occ|icp|lpc)', 'expected':True}, # required link 'xics' not found: Property '.xics' not found 134 {'device':'powernv-cpu-.*', 'expected':True}, # pnv_core_realize: required link 'xics' not found: Property '.xics' not found 135 136 # ioapic devices are already created by pc and will fail: 137 {'machine':'q35|pc.*', 'device':'kvm-ioapic', 'expected':True}, # Only 1 ioapics allowed 138 {'machine':'q35|pc.*', 'device':'ioapic', 'expected':True}, # Only 1 ioapics allowed 139 140 # "spapr-cpu-core needs a pseries machine" 141 {'machine':'(?!pseries).*', 'device':'.*-spapr-cpu-core', 'expected':True}, 142 143 # KVM-specific devices shouldn't be tried without accel=kvm: 144 {'accel':'(?!kvm).*', 'device':'kvmclock', 'expected':True}, 145 146 # xen-specific machines and devices: 147 {'accel':'(?!xen).*', 'machine':'xen.*', 'expected':True}, 148 {'accel':'(?!xen).*', 'device':'xen-.*', 'expected':True}, 149 150 # this fails on some machine-types, but not all, so they don't have expected=True: 151 {'device':'vmgenid'}, # vmgenid requires DMA write support in fw_cfg, which this machine type does not provide 152 153 # Silence INFO messages for errors that are common on multiple 154 # devices/machines: 155 {'log':r"No '[\w-]+' bus found for device '[\w-]+'"}, 156 {'log':r"images* must be given with the 'pflash' parameter"}, 157 {'log':r"(Guest|ROM|Flash|Kernel) image must be specified"}, 158 {'log':r"[cC]ould not load [\w ]+ (BIOS|bios) '[\w-]+\.bin'"}, 159 {'log':r"Couldn't find rom image '[\w-]+\.bin'"}, 160 {'log':r"speed mismatch trying to attach usb device"}, 161 {'log':r"Can't create a second ISA bus"}, 162 {'log':r"duplicate fw_cfg file name"}, 163 # sysbus-related error messages: most machines reject most dynamic sysbus devices: 164 {'log':r"Option '-device [\w.,-]+' cannot be handled by this machine"}, 165 {'log':r"Device [\w.,-]+ is not supported by this machine yet"}, 166 {'log':r"Device [\w.,-]+ can not be dynamically instantiated"}, 167 {'log':r"Platform Bus: Can not fit MMIO region of size "}, 168 # other more specific errors we will ignore: 169 {'device':'.*-spapr-cpu-core', 'log':r"CPU core type should be"}, 170 {'log':r"MSI(-X)? is not supported by interrupt controller"}, 171 {'log':r"pxb-pcie? devices cannot reside on a PCIe? bus"}, 172 {'log':r"Ignoring smp_cpus value"}, 173 {'log':r"sd_init failed: Drive 'sd0' is already in use because it has been automatically connected to another device"}, 174 {'log':r"This CPU requires a smaller page size than the system is using"}, 175 {'log':r"MSI-X support is mandatory in the S390 architecture"}, 176 {'log':r"rom check and register reset failed"}, 177 {'log':r"Unable to initialize GIC, CPUState for CPU#0 not valid"}, 178 {'log':r"Multiple VT220 operator consoles are not supported"}, 179 {'log':r"core 0 already populated"}, 180 {'log':r"could not find stage1 bootloader"}, 181 182 # other exitcode=1 failures not listed above will just generate INFO messages: 183 {'exitcode':1, 'loglevel':logging.INFO}, 184 185 # everything else (including SIGABRT and SIGSEGV) will be a fatal error: 186 {'exitcode':None, 'fatal':True, 'loglevel':logging.FATAL}, 187] 188 189 190def whitelistTestCaseMatch(wl, t): 191 """Check if a test case specification can match a whitelist entry 192 193 This only checks if a whitelist entry is a candidate match 194 for a given test case, it won't check if the test case 195 results/output match the entry. See whitelistResultMatch(). 196 """ 197 return (('machine' not in wl or 198 'machine' not in t or 199 re.match(wl['machine'] + '$', t['machine'])) and 200 ('accel' not in wl or 201 'accel' not in t or 202 re.match(wl['accel'] + '$', t['accel'])) and 203 ('device' not in wl or 204 'device' not in t or 205 re.match(wl['device'] + '$', t['device']))) 206 207 208def whitelistCandidates(t): 209 """Generate the list of candidates that can match a test case""" 210 for i, wl in enumerate(ERROR_WHITELIST): 211 if whitelistTestCaseMatch(wl, t): 212 yield (i, wl) 213 214 215def findExpectedResult(t): 216 """Check if there's an expected=True whitelist entry for a test case 217 218 Returns (i, wl) tuple, where i is the index in 219 ERROR_WHITELIST and wl is the whitelist entry itself. 220 """ 221 for i, wl in whitelistCandidates(t): 222 if wl.get('expected'): 223 return (i, wl) 224 225 226def whitelistResultMatch(wl, r): 227 """Check if test case results/output match a whitelist entry 228 229 It is valid to call this function only if 230 whitelistTestCaseMatch() is True for the entry (e.g. on 231 entries returned by whitelistCandidates()) 232 """ 233 assert whitelistTestCaseMatch(wl, r['testcase']) 234 return ((wl.get('exitcode', 1) is None or 235 r['exitcode'] == wl.get('exitcode', 1)) and 236 ('log' not in wl or 237 re.search(wl['log'], r['log'], re.MULTILINE))) 238 239 240def checkResultWhitelist(r): 241 """Look up whitelist entry for a given test case result 242 243 Returns (i, wl) tuple, where i is the index in 244 ERROR_WHITELIST and wl is the whitelist entry itself. 245 """ 246 for i, wl in whitelistCandidates(r['testcase']): 247 if whitelistResultMatch(wl, r): 248 return i, wl 249 250 raise Exception("this should never happen") 251 252 253def qemuOptsEscape(s): 254 """Escape option value QemuOpts""" 255 return s.replace(",", ",,") 256 257 258def formatTestCase(t): 259 """Format test case info as "key=value key=value" for prettier logging output""" 260 return ' '.join('%s=%s' % (k, v) for k, v in t.items()) 261 262 263def qomListTypeNames(vm, **kwargs): 264 """Run qom-list-types QMP command, return type names""" 265 types = vm.command('qom-list-types', **kwargs) 266 return [t['name'] for t in types] 267 268 269def infoQDM(vm): 270 """Parse 'info qdm' output""" 271 args = {'command-line': 'info qdm'} 272 devhelp = vm.command('human-monitor-command', **args) 273 for l in devhelp.split('\n'): 274 l = l.strip() 275 if l == '' or l.endswith(':'): 276 continue 277 d = {'name': re.search(r'name "([^"]+)"', l).group(1), 278 'no-user': (re.search(', no-user', l) is not None)} 279 yield d 280 281 282class QemuBinaryInfo(object): 283 def __init__(self, binary, devtype): 284 if devtype is None: 285 devtype = 'device' 286 287 self.binary = binary 288 self._machine_info = {} 289 290 dbg("devtype: %r", devtype) 291 args = ['-S', '-machine', 'none,accel=kvm:tcg'] 292 dbg("querying info for QEMU binary: %s", binary) 293 vm = QEMUMachine(binary=binary, args=args) 294 vm.launch() 295 try: 296 self.alldevs = set(qomListTypeNames(vm, implements=devtype, abstract=False)) 297 # there's no way to query DeviceClass::user_creatable using QMP, 298 # so use 'info qdm': 299 self.no_user_devs = set([d['name'] for d in infoQDM(vm, ) if d['no-user']]) 300 self.machines = list(m['name'] for m in vm.command('query-machines')) 301 self.user_devs = self.alldevs.difference(self.no_user_devs) 302 self.kvm_available = vm.command('query-kvm')['enabled'] 303 finally: 304 vm.shutdown() 305 306 def machineInfo(self, machine): 307 """Query for information on a specific machine-type 308 309 Results are cached internally, in case the same machine- 310 type is queried multiple times. 311 """ 312 if machine in self._machine_info: 313 return self._machine_info[machine] 314 315 mi = {} 316 args = ['-S', '-machine', '%s' % (machine)] 317 dbg("querying machine info for binary=%s machine=%s", self.binary, machine) 318 vm = QEMUMachine(binary=self.binary, args=args) 319 try: 320 vm.launch() 321 mi['runnable'] = True 322 except KeyboardInterrupt: 323 raise 324 except: 325 dbg("exception trying to run binary=%s machine=%s", self.binary, machine, exc_info=sys.exc_info()) 326 dbg("log: %r", vm.get_log()) 327 mi['runnable'] = False 328 329 vm.shutdown() 330 self._machine_info[machine] = mi 331 return mi 332 333 334BINARY_INFO = {} 335 336 337def getBinaryInfo(args, binary): 338 if binary not in BINARY_INFO: 339 BINARY_INFO[binary] = QemuBinaryInfo(binary, args.devtype) 340 return BINARY_INFO[binary] 341 342 343def checkOneCase(args, testcase): 344 """Check one specific case 345 346 Returns a dictionary containing failure information on error, 347 or None on success 348 """ 349 binary = testcase['binary'] 350 accel = testcase['accel'] 351 machine = testcase['machine'] 352 device = testcase['device'] 353 354 dbg("will test: %r", testcase) 355 356 args = ['-S', '-machine', '%s,accel=%s' % (machine, accel), 357 '-device', qemuOptsEscape(device)] 358 cmdline = ' '.join([binary] + args) 359 dbg("will launch QEMU: %s", cmdline) 360 vm = QEMUMachine(binary=binary, args=args) 361 362 exc_traceback = None 363 try: 364 vm.launch() 365 except KeyboardInterrupt: 366 raise 367 except: 368 exc_traceback = traceback.format_exc() 369 dbg("Exception while running test case") 370 finally: 371 vm.shutdown() 372 ec = vm.exitcode() 373 log = vm.get_log() 374 375 if exc_traceback is not None or ec != 0: 376 return {'exc_traceback':exc_traceback, 377 'exitcode':ec, 378 'log':log, 379 'testcase':testcase, 380 'cmdline':cmdline} 381 382 383def binariesToTest(args, testcase): 384 if args.qemu: 385 r = args.qemu 386 else: 387 r = glob.glob('./*-softmmu/qemu-system-*') 388 return r 389 390 391def accelsToTest(args, testcase): 392 if getBinaryInfo(args, testcase['binary']).kvm_available: 393 yield 'kvm' 394 yield 'tcg' 395 396 397def machinesToTest(args, testcase): 398 return getBinaryInfo(args, testcase['binary']).machines 399 400 401def devicesToTest(args, testcase): 402 return getBinaryInfo(args, testcase['binary']).user_devs 403 404 405TESTCASE_VARIABLES = [ 406 ('binary', binariesToTest), 407 ('accel', accelsToTest), 408 ('machine', machinesToTest), 409 ('device', devicesToTest), 410] 411 412 413def genCases1(args, testcases, var, fn): 414 """Generate new testcases for one variable 415 416 If an existing item already has a variable set, don't 417 generate new items and just return it directly. This 418 allows the "-t" command-line option to be used to choose 419 a specific test case. 420 """ 421 for testcase in testcases: 422 if var in testcase: 423 yield testcase.copy() 424 else: 425 for i in fn(args, testcase): 426 t = testcase.copy() 427 t[var] = i 428 yield t 429 430 431def genCases(args, testcase): 432 """Generate test cases for all variables 433 """ 434 cases = [testcase.copy()] 435 for var, fn in TESTCASE_VARIABLES: 436 dbg("var: %r, fn: %r", var, fn) 437 cases = genCases1(args, cases, var, fn) 438 return cases 439 440 441def casesToTest(args, testcase): 442 cases = genCases(args, testcase) 443 if args.random: 444 cases = list(cases) 445 cases = random.sample(cases, min(args.random, len(cases))) 446 if args.debug: 447 cases = list(cases) 448 dbg("%d test cases to test", len(cases)) 449 if args.shuffle: 450 cases = list(cases) 451 random.shuffle(cases) 452 return cases 453 454 455def logFailure(f, level): 456 t = f['testcase'] 457 logger.log(level, "failed: %s", formatTestCase(t)) 458 logger.log(level, "cmdline: %s", f['cmdline']) 459 for l in f['log'].strip().split('\n'): 460 logger.log(level, "log: %s", l) 461 logger.log(level, "exit code: %r", f['exitcode']) 462 if f['exc_traceback']: 463 logger.log(level, "exception:") 464 for l in f['exc_traceback'].split('\n'): 465 logger.log(level, " %s", l.rstrip('\n')) 466 467 468def main(): 469 parser = argparse.ArgumentParser(description="QEMU -device crash test") 470 parser.add_argument('-t', metavar='KEY=VALUE', nargs='*', 471 help="Limit test cases to KEY=VALUE", 472 action='append', dest='testcases', default=[]) 473 parser.add_argument('-d', '--debug', action='store_true', 474 help='debug output') 475 parser.add_argument('-v', '--verbose', action='store_true', default=True, 476 help='verbose output') 477 parser.add_argument('-q', '--quiet', dest='verbose', action='store_false', 478 help='non-verbose output') 479 parser.add_argument('-r', '--random', type=int, metavar='COUNT', 480 help='run a random sample of COUNT test cases', 481 default=0) 482 parser.add_argument('--shuffle', action='store_true', 483 help='Run test cases in random order') 484 parser.add_argument('--dry-run', action='store_true', 485 help="Don't run any tests, just generate list") 486 parser.add_argument('-D', '--devtype', metavar='TYPE', 487 help="Test only device types that implement TYPE") 488 parser.add_argument('-Q', '--quick', action='store_true', default=True, 489 help="Quick mode: skip test cases that are expected to fail") 490 parser.add_argument('-F', '--full', action='store_false', dest='quick', 491 help="Full mode: test cases that are expected to fail") 492 parser.add_argument('--strict', action='store_true', dest='strict', 493 help="Treat all warnings as fatal") 494 parser.add_argument('qemu', nargs='*', metavar='QEMU', 495 help='QEMU binary to run') 496 args = parser.parse_args() 497 498 if args.debug: 499 lvl = logging.DEBUG 500 elif args.verbose: 501 lvl = logging.INFO 502 else: 503 lvl = logging.WARN 504 logging.basicConfig(stream=sys.stdout, level=lvl, format='%(levelname)s: %(message)s') 505 506 fatal_failures = [] 507 wl_stats = {} 508 skipped = 0 509 total = 0 510 511 tc = {} 512 dbg("testcases: %r", args.testcases) 513 if args.testcases: 514 for t in chain(*args.testcases): 515 for kv in t.split(): 516 k, v = kv.split('=', 1) 517 tc[k] = v 518 519 if len(binariesToTest(args, tc)) == 0: 520 print("No QEMU binary found", file=sys.stderr) 521 parser.print_usage(sys.stderr) 522 return 1 523 524 for t in casesToTest(args, tc): 525 logger.info("running test case: %s", formatTestCase(t)) 526 total += 1 527 528 expected_match = findExpectedResult(t) 529 if (args.quick and 530 (expected_match or 531 not getBinaryInfo(args, t['binary']).machineInfo(t['machine'])['runnable'])): 532 dbg("skipped: %s", formatTestCase(t)) 533 skipped += 1 534 continue 535 536 if args.dry_run: 537 continue 538 539 try: 540 f = checkOneCase(args, t) 541 except KeyboardInterrupt: 542 break 543 544 if f: 545 i, wl = checkResultWhitelist(f) 546 dbg("testcase: %r, whitelist match: %r", t, wl) 547 wl_stats.setdefault(i, []).append(f) 548 level = wl.get('loglevel', logging.DEBUG) 549 logFailure(f, level) 550 if wl.get('fatal') or (args.strict and level >= logging.WARN): 551 fatal_failures.append(f) 552 else: 553 dbg("success: %s", formatTestCase(t)) 554 if expected_match: 555 logger.warn("Didn't fail as expected: %s", formatTestCase(t)) 556 557 logger.info("Total: %d test cases", total) 558 if skipped: 559 logger.info("Skipped %d test cases", skipped) 560 561 if args.debug: 562 stats = sorted([(len(wl_stats.get(i, [])), wl) for i, wl in 563 enumerate(ERROR_WHITELIST)], key=lambda x: x[0]) 564 for count, wl in stats: 565 dbg("whitelist entry stats: %d: %r", count, wl) 566 567 if fatal_failures: 568 for f in fatal_failures: 569 t = f['testcase'] 570 logger.error("Fatal failure: %s", formatTestCase(t)) 571 logger.error("Fatal failures on some machine/device combinations") 572 return 1 573 574if __name__ == '__main__': 575 sys.exit(main()) 576