blob: 243f7df24b87ed4a465f587ed87a486e72c1062a [file] [log] [blame]
Fang Deng5d518f42013-08-02 14:04:32 -07001# Copyright (c) 2013 The Chromium OS Authors. All rights reserved.
2# Use of this source code is governed by a BSD-style license that can be
3# found in the LICENSE file.
4#
5# Expects to be run in an environment with sudo and no interactive password
6# prompt, such as within the Chromium OS development chroot.
7
8
9"""This file provides core logic for servo verify/repair process."""
10
11
12import httplib
13import logging
14import socket
Fang Deng5d518f42013-08-02 14:04:32 -070015import xmlrpclib
16
17from autotest_lib.client.bin import utils
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -070018from autotest_lib.client.common_lib import control_data
Fang Deng5d518f42013-08-02 14:04:32 -070019from autotest_lib.client.common_lib import error
beeps5e8c45a2013-12-17 22:05:11 -080020from autotest_lib.client.common_lib import global_config
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -070021from autotest_lib.client.common_lib import host_states
Richard Barnette9a26ad62016-06-10 12:03:08 -070022from autotest_lib.client.common_lib import hosts
Dan Shi0942b1d2015-03-31 11:07:00 -070023from autotest_lib.client.common_lib import lsbrelease_utils
beeps5e8c45a2013-12-17 22:05:11 -080024from autotest_lib.client.common_lib.cros import autoupdater
25from autotest_lib.client.common_lib.cros import dev_server
Fang Deng5d518f42013-08-02 14:04:32 -070026from autotest_lib.client.common_lib.cros import retry
Gabe Black1e1c41b2015-02-04 23:55:15 -080027from autotest_lib.client.common_lib.cros.graphite import autotest_stats
Christopher Wileycef1f902014-06-19 11:11:23 -070028from autotest_lib.client.common_lib.cros.network import ping_runner
Hsinyu Chaoe0b08e62015-08-11 10:50:37 +000029from autotest_lib.client.cros import constants as client_constants
Richard Barnettee519dcd2016-08-15 17:37:17 -070030from autotest_lib.server import afe_utils
beeps5e8c45a2013-12-17 22:05:11 -080031from autotest_lib.server import site_utils as server_site_utils
Cheng-Yi Chiang22612862015-08-20 20:39:57 +080032from autotest_lib.server.cros import dnsname_mangler
Simran Basi0739d682015-02-25 16:22:56 -080033from autotest_lib.server.cros.dynamic_suite import frontend_wrappers
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -070034from autotest_lib.server.cros.dynamic_suite import control_file_getter
Richard Barnette9a26ad62016-06-10 12:03:08 -070035from autotest_lib.server.cros.servo import servo
36from autotest_lib.server.hosts import servo_repair
Fang Deng5d518f42013-08-02 14:04:32 -070037from autotest_lib.server.hosts import ssh_host
Fang Dengd4fe7392013-09-20 12:18:21 -070038from autotest_lib.site_utils.rpm_control_system import rpm_client
Fang Deng5d518f42013-08-02 14:04:32 -070039
40
Simran Basi0739d682015-02-25 16:22:56 -080041# Names of the host attributes in the database that represent the values for
42# the servo_host and servo_port for a servo connected to the DUT.
43SERVO_HOST_ATTR = 'servo_host'
44SERVO_PORT_ATTR = 'servo_port'
Richard Barnettee519dcd2016-08-15 17:37:17 -070045SERVO_BOARD_ATTR = 'servo_board'
Kevin Cheng643ce8a2016-09-15 15:42:12 -070046SERVO_SERIAL_ATTR = 'servo_serial'
Simran Basi0739d682015-02-25 16:22:56 -080047
Dan Shi3b2adf62015-09-02 17:46:54 -070048_CONFIG = global_config.global_config
xixuan6cf6d2f2016-01-29 15:29:00 -080049ENABLE_SSH_TUNNEL_FOR_SERVO = _CONFIG.get_config_value(
50 'CROS', 'enable_ssh_tunnel_for_servo', type=bool, default=False)
Simran Basi0739d682015-02-25 16:22:56 -080051
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -070052AUTOTEST_BASE = _CONFIG.get_config_value(
53 'SCHEDULER', 'drone_installation_directory',
54 default='/usr/local/autotest')
55
56_SERVO_HOST_REBOOT_TEST_NAME = 'servohost_Reboot'
Kevin Cheng55265902016-10-19 12:46:50 -070057_SERVO_HOST_FORCE_REBOOT_TEST_NAME = 'servohost_Reboot.force_reboot'
Fang Deng5d518f42013-08-02 14:04:32 -070058
Fang Deng5d518f42013-08-02 14:04:32 -070059class ServoHost(ssh_host.SSHHost):
60 """Host class for a host that controls a servo, e.g. beaglebone."""
61
Richard Barnette9a26ad62016-06-10 12:03:08 -070062 DEFAULT_PORT = 9999
63
Dan Shie5b3c512014-08-21 12:12:09 -070064 # Timeout for initializing servo signals.
65 INITIALIZE_SERVO_TIMEOUT_SECS = 30
Richard Barnette9a26ad62016-06-10 12:03:08 -070066
xixuan6cf6d2f2016-01-29 15:29:00 -080067 # Ready test function
68 SERVO_READY_METHOD = 'get_version'
Fang Deng5d518f42013-08-02 14:04:32 -070069
Gabe Black1e1c41b2015-02-04 23:55:15 -080070 _timer = autotest_stats.Timer('servo_host')
Fang Dengd4fe7392013-09-20 12:18:21 -070071
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -070072 REBOOT_CMD = 'sleep 1; reboot & sleep 10; reboot -f'
73
Fang Deng5d518f42013-08-02 14:04:32 -070074
Richard Barnette17bfc6c2016-08-04 18:41:43 -070075 def _initialize(self, servo_host='localhost',
Richard Barnettee519dcd2016-08-15 17:37:17 -070076 servo_port=DEFAULT_PORT, servo_board=None,
Kevin Cheng643ce8a2016-09-15 15:42:12 -070077 servo_serial=None, is_in_lab=None, *args, **dargs):
Fang Deng5d518f42013-08-02 14:04:32 -070078 """Initialize a ServoHost instance.
79
80 A ServoHost instance represents a host that controls a servo.
81
82 @param servo_host: Name of the host where the servod process
83 is running.
84 @param servo_port: Port the servod process is listening on.
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -070085 @param servo_board: Board that the servo is connected to.
Dan Shi4d478522014-02-14 13:46:32 -080086 @param is_in_lab: True if the servo host is in Cros Lab. Default is set
87 to None, for which utils.host_is_in_lab_zone will be
88 called to check if the servo host is in Cros lab.
Fang Deng5d518f42013-08-02 14:04:32 -070089
90 """
91 super(ServoHost, self)._initialize(hostname=servo_host,
92 *args, **dargs)
Richard Barnettee519dcd2016-08-15 17:37:17 -070093 self.servo_port = servo_port
94 self.servo_board = servo_board
Kevin Cheng643ce8a2016-09-15 15:42:12 -070095 self.servo_serial = servo_serial
Richard Barnettee519dcd2016-08-15 17:37:17 -070096 self._servo = None
Richard Barnette9a26ad62016-06-10 12:03:08 -070097 self._repair_strategy = (
98 servo_repair.create_servo_repair_strategy())
Richard Barnettee519dcd2016-08-15 17:37:17 -070099 self._is_localhost = (self.hostname == 'localhost')
100 if self._is_localhost:
101 self._is_in_lab = False
102 elif is_in_lab is None:
Dan Shi4d478522014-02-14 13:46:32 -0800103 self._is_in_lab = utils.host_is_in_lab_zone(self.hostname)
104 else:
105 self._is_in_lab = is_in_lab
xixuan6cf6d2f2016-01-29 15:29:00 -0800106
Richard Barnettee519dcd2016-08-15 17:37:17 -0700107 # Commands on the servo host must be run by the superuser.
108 # Our account on a remote host is root, but if our target is
109 # localhost then we might be running unprivileged. If so,
110 # `sudo` will have to be added to the commands.
Fang Deng5d518f42013-08-02 14:04:32 -0700111 if self._is_localhost:
112 self._sudo_required = utils.system_output('id -u') != '0'
113 else:
114 self._sudo_required = False
Richard Barnettee519dcd2016-08-15 17:37:17 -0700115
Richard Barnette9a26ad62016-06-10 12:03:08 -0700116
117 def connect_servo(self):
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700118 """Establish a connection to the servod server on this host.
Richard Barnette9a26ad62016-06-10 12:03:08 -0700119
120 Initializes `self._servo` and then verifies that all network
121 connections are working. This will create an ssh tunnel if
122 it's required.
123
124 As a side effect of testing the connection, all signals on the
125 target servo are reset to default values, and the USB stick is
126 set to the neutral (off) position.
127 """
Kevin Cheng643ce8a2016-09-15 15:42:12 -0700128 servo_obj = servo.Servo(servo_host=self, servo_serial=self.servo_serial)
Richard Barnette9a26ad62016-06-10 12:03:08 -0700129 timeout, _ = retry.timeout(
130 servo_obj.initialize_dut,
131 timeout_sec=self.INITIALIZE_SERVO_TIMEOUT_SECS)
132 if timeout:
133 raise hosts.AutoservVerifyError(
134 'Servo initialize timed out.')
135 self._servo = servo_obj
136
137
138 def disconnect_servo(self):
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700139 """Disconnect our servo if it exists.
Richard Barnette9a26ad62016-06-10 12:03:08 -0700140
141 If we've previously successfully connected to our servo,
142 disconnect any established ssh tunnel, and set `self._servo`
143 back to `None`.
144 """
145 if self._servo:
146 # N.B. This call is safe even without a tunnel:
147 # rpc_server_tracker.disconnect() silently ignores
148 # unknown ports.
149 self.rpc_server_tracker.disconnect(self.servo_port)
150 self._servo = None
Fang Deng5d518f42013-08-02 14:04:32 -0700151
152
153 def is_in_lab(self):
154 """Check whether the servo host is a lab device.
155
156 @returns: True if the servo host is in Cros Lab, otherwise False.
157
158 """
159 return self._is_in_lab
160
161
162 def is_localhost(self):
163 """Checks whether the servo host points to localhost.
164
165 @returns: True if it points to localhost, otherwise False.
166
167 """
168 return self._is_localhost
169
170
171 def get_servod_server_proxy(self):
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700172 """Return a proxy that can be used to communicate with servod server.
Fang Deng5d518f42013-08-02 14:04:32 -0700173
174 @returns: An xmlrpclib.ServerProxy that is connected to the servod
175 server on the host.
Fang Deng5d518f42013-08-02 14:04:32 -0700176 """
Richard Barnette9a26ad62016-06-10 12:03:08 -0700177 if ENABLE_SSH_TUNNEL_FOR_SERVO and not self.is_localhost():
178 return self.rpc_server_tracker.xmlrpc_connect(
179 None, self.servo_port,
180 ready_test_name=self.SERVO_READY_METHOD,
181 timeout_seconds=60)
182 else:
183 remote = 'http://%s:%s' % (self.hostname, self.servo_port)
184 return xmlrpclib.ServerProxy(remote)
Fang Deng5d518f42013-08-02 14:04:32 -0700185
186
Richard Barnette9a26ad62016-06-10 12:03:08 -0700187 def is_cros_host(self):
beeps5e8c45a2013-12-17 22:05:11 -0800188 """Check if a servo host is running chromeos.
189
190 @return: True if the servo host is running chromeos.
191 False if it isn't, or we don't have enough information.
192 """
193 try:
194 result = self.run('grep -q CHROMEOS /etc/lsb-release',
195 ignore_status=True, timeout=10)
196 except (error.AutoservRunError, error.AutoservSSHTimeout):
197 return False
198 return result.exit_status == 0
199
200
Fang Deng5d518f42013-08-02 14:04:32 -0700201 def make_ssh_command(self, user='root', port=22, opts='', hosts_file=None,
202 connect_timeout=None, alive_interval=None):
203 """Override default make_ssh_command to use tuned options.
204
205 Tuning changes:
206 - ConnectTimeout=30; maximum of 30 seconds allowed for an SSH
207 connection failure. Consistency with remote_access.py.
208
209 - ServerAliveInterval=180; which causes SSH to ping connection every
210 180 seconds. In conjunction with ServerAliveCountMax ensures
211 that if the connection dies, Autotest will bail out quickly.
212
213 - ServerAliveCountMax=3; consistency with remote_access.py.
214
215 - ConnectAttempts=4; reduce flakiness in connection errors;
216 consistency with remote_access.py.
217
218 - UserKnownHostsFile=/dev/null; we don't care about the keys.
219
220 - SSH protocol forced to 2; needed for ServerAliveInterval.
221
222 @param user User name to use for the ssh connection.
223 @param port Port on the target host to use for ssh connection.
224 @param opts Additional options to the ssh command.
225 @param hosts_file Ignored.
226 @param connect_timeout Ignored.
227 @param alive_interval Ignored.
228
229 @returns: An ssh command with the requested settings.
230
231 """
232 base_command = ('/usr/bin/ssh -a -x %s -o StrictHostKeyChecking=no'
233 ' -o UserKnownHostsFile=/dev/null -o BatchMode=yes'
234 ' -o ConnectTimeout=30 -o ServerAliveInterval=180'
235 ' -o ServerAliveCountMax=3 -o ConnectionAttempts=4'
236 ' -o Protocol=2 -l %s -p %d')
237 return base_command % (opts, user, port)
238
239
240 def _make_scp_cmd(self, sources, dest):
241 """Format scp command.
242
243 Given a list of source paths and a destination path, produces the
244 appropriate scp command for encoding it. Remote paths must be
245 pre-encoded. Overrides _make_scp_cmd in AbstractSSHHost
246 to allow additional ssh options.
247
248 @param sources: A list of source paths to copy from.
249 @param dest: Destination path to copy to.
250
251 @returns: An scp command that copies |sources| on local machine to
252 |dest| on the remote servo host.
253
254 """
255 command = ('scp -rq %s -o BatchMode=yes -o StrictHostKeyChecking=no '
256 '-o UserKnownHostsFile=/dev/null -P %d %s "%s"')
257 return command % (self.master_ssh_option,
258 self.port, ' '.join(sources), dest)
259
260
261 def run(self, command, timeout=3600, ignore_status=False,
262 stdout_tee=utils.TEE_TO_LOGS, stderr_tee=utils.TEE_TO_LOGS,
263 connect_timeout=30, options='', stdin=None, verbose=True, args=()):
264 """Run a command on the servo host.
265
266 Extends method `run` in SSHHost. If the servo host is a remote device,
267 it will call `run` in SSHost without changing anything.
268 If the servo host is 'localhost', it will call utils.system_output.
269
270 @param command: The command line string.
271 @param timeout: Time limit in seconds before attempting to
272 kill the running process. The run() function
273 will take a few seconds longer than 'timeout'
274 to complete if it has to kill the process.
275 @param ignore_status: Do not raise an exception, no matter
276 what the exit code of the command is.
277 @param stdout_tee/stderr_tee: Where to tee the stdout/stderr.
278 @param connect_timeout: SSH connection timeout (in seconds)
279 Ignored if host is 'localhost'.
280 @param options: String with additional ssh command options
281 Ignored if host is 'localhost'.
282 @param stdin: Stdin to pass (a string) to the executed command.
283 @param verbose: Log the commands.
284 @param args: Sequence of strings to pass as arguments to command by
285 quoting them in " and escaping their contents if necessary.
286
287 @returns: A utils.CmdResult object.
288
289 @raises AutoservRunError if the command failed.
290 @raises AutoservSSHTimeout SSH connection has timed out. Only applies
291 when servo host is not 'localhost'.
292
293 """
294 run_args = {'command': command, 'timeout': timeout,
295 'ignore_status': ignore_status, 'stdout_tee': stdout_tee,
296 'stderr_tee': stderr_tee, 'stdin': stdin,
297 'verbose': verbose, 'args': args}
298 if self.is_localhost():
299 if self._sudo_required:
300 run_args['command'] = 'sudo -n %s' % command
301 try:
302 return utils.run(**run_args)
303 except error.CmdError as e:
304 logging.error(e)
305 raise error.AutoservRunError('command execution error',
306 e.result_obj)
307 else:
308 run_args['connect_timeout'] = connect_timeout
309 run_args['options'] = options
310 return super(ServoHost, self).run(**run_args)
311
312
Richard Barnette9a26ad62016-06-10 12:03:08 -0700313 def _get_release_version(self):
Dan Shi0942b1d2015-03-31 11:07:00 -0700314 """Get the value of attribute CHROMEOS_RELEASE_VERSION from lsb-release.
315
316 @returns The version string in lsb-release, under attribute
317 CHROMEOS_RELEASE_VERSION.
318 """
319 lsb_release_content = self.run(
320 'cat "%s"' % client_constants.LSB_RELEASE).stdout.strip()
321 return lsbrelease_utils.get_chromeos_release_version(
322 lsb_release_content=lsb_release_content)
323
324
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700325 def get_attached_duts(self, afe):
326 """Gather a list of duts that use this servo host.
327
328 @param afe: afe instance.
329
330 @returns list of duts.
Richard Barnette3a7697f2016-04-20 11:33:27 -0700331 """
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700332 return afe.get_hosts_by_attribute(
333 attribute=SERVO_HOST_ATTR, value=self.hostname)
334
335
336 def get_board(self):
337 """Determine the board for this servo host.
338
339 @returns a string representing this servo host's board.
340 """
341 return lsbrelease_utils.get_current_board(
342 lsb_release_content=self.run('cat /etc/lsb-release').stdout)
343
344
345 def _choose_dut_for_synchronized_reboot(self, dut_list, afe):
346 """Choose which dut to schedule servo host reboot job.
347
348 We'll want a semi-deterministic way of selecting which host should be
349 scheduled for the servo host reboot job. For now we'll sort the
350 list with the expectation the dut list will stay consistent.
351 From there we'll grab the first dut that is available so we
352 don't schedule a job on a dut that will never run.
353
354 @param dut_list: List of the dut hostnames to choose from.
355 @param afe: Instance of the AFE.
356
357 @return hostname of dut to schedule job on.
358 """
359 afe_hosts = afe.get_hosts(dut_list)
360 afe_hosts.sort()
361 for afe_host in afe_hosts:
362 if afe_host.status not in host_states.UNAVAILABLE_STATES:
363 return afe_host.hostname
364 # If they're all unavailable, just return the first sorted dut.
365 dut_list.sort()
366 return dut_list[0]
367
368
369 def _sync_job_scheduled_for_duts(self, dut_list, afe):
370 """Checks if a synchronized reboot has been scheduled for these duts.
371
372 Grab all the host queue entries that aren't completed for the duts and
373 see if any of them have the expected job name.
374
375 @param dut_list: List of duts to check on.
376 @param afe: Instance of the AFE.
377
378 @returns True if the job is scheduled, False otherwise.
379 """
380 afe_hosts = afe.get_hosts(dut_list)
381 for afe_host in afe_hosts:
382 hqes = afe.get_host_queue_entries(host=afe_host.id, complete=0)
383 for hqe in hqes:
384 job = afe.get_jobs(id=hqe.job.id)
Kevin Cheng55265902016-10-19 12:46:50 -0700385 if job and job[0].name in (_SERVO_HOST_REBOOT_TEST_NAME,
386 _SERVO_HOST_FORCE_REBOOT_TEST_NAME):
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700387 return True
388 return False
389
390
Kevin Cheng55265902016-10-19 12:46:50 -0700391 def schedule_synchronized_reboot(self, dut_list, afe, force_reboot=False):
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700392 """Schedule a job to reboot the servo host.
393
394 When we schedule a job, it will create a ServoHost object which will
395 go through this entire flow of checking if a reboot is needed and
396 trying to schedule it. There is probably a better approach to setting
397 up a synchronized reboot but I'm coming up short on better ideas so I
398 apologize for this circus show.
399
Kevin Cheng55265902016-10-19 12:46:50 -0700400 @param dut_list: List of duts that need to be locked.
401 @param afe: Instance of afe.
402 @param force_reboot: Boolean to indicate if a forced reboot should be
403 scheduled or not.
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700404 """
405 # If we've already scheduled job on a dut, we're done here.
406 if self._sync_job_scheduled_for_duts(dut_list, afe):
407 return
408
409 # Looks like we haven't scheduled a job yet.
Kevin Cheng55265902016-10-19 12:46:50 -0700410 test = (_SERVO_HOST_REBOOT_TEST_NAME if not force_reboot
411 else _SERVO_HOST_FORCE_REBOOT_TEST_NAME)
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700412 dut = self._choose_dut_for_synchronized_reboot(dut_list, afe)
413 getter = control_file_getter.FileSystemGetter([AUTOTEST_BASE])
Kevin Cheng55265902016-10-19 12:46:50 -0700414 control_file = getter.get_control_file_contents_by_name(test)
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700415 control_type = control_data.CONTROL_TYPE_NAMES.SERVER
Kevin Cheng55265902016-10-19 12:46:50 -0700416 afe.create_job(control_file=control_file, name=test,
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700417 control_type=control_type, hosts=[dut])
418
419
420 def reboot(self, *args, **dargs):
421 """Reboot using special servo host reboot command."""
422 super(ServoHost, self).reboot(reboot_cmd=self.REBOOT_CMD,
423 *args, **dargs)
424
425
426 def _check_for_reboot(self, updater):
427 """Reboot this servo host if an upgrade is waiting.
Richard Barnette3a7697f2016-04-20 11:33:27 -0700428
429 If the host has successfully downloaded and finalized a new
430 build, reboot.
431
432 @param updater: a ChromiumOSUpdater instance for checking
433 whether reboot is needed.
434 @return Return a (status, build) tuple reflecting the
435 update_engine status and current build of the host
436 at the end of the call.
437 """
Richard Barnette9a26ad62016-06-10 12:03:08 -0700438 current_build_number = self._get_release_version()
Richard Barnette3a7697f2016-04-20 11:33:27 -0700439 status = updater.check_update_status()
440 if status == autoupdater.UPDATER_NEED_REBOOT:
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700441 # Check if we need to schedule an organized reboot.
442 afe = frontend_wrappers.RetryingAFE(timeout_min=5, delay_sec=10)
443 dut_list = self.get_attached_duts(afe)
444 logging.info('servo host has the following duts: %s', dut_list)
445 if len(dut_list) > 1:
446 logging.info('servo host has multiple duts, scheduling '
447 'synchronized reboot')
448 self.schedule_synchronized_reboot(dut_list, afe)
449 return status, current_build_number
450
451 logging.info('Rebooting servo host %s from build %s',
Richard Barnette3a7697f2016-04-20 11:33:27 -0700452 self.hostname, current_build_number)
453 # Tell the reboot() call not to wait for completion.
454 # Otherwise, the call will log reboot failure if servo does
455 # not come back. The logged reboot failure will lead to
456 # test job failure. If the test does not require servo, we
457 # don't want servo failure to fail the test with error:
458 # `Host did not return from reboot` in status.log.
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700459 self.reboot(fastsync=True, wait=False)
Richard Barnette3a7697f2016-04-20 11:33:27 -0700460
461 # We told the reboot() call not to wait, but we need to wait
462 # for the reboot before we continue. Alas. The code from
463 # here below is basically a copy of Host.wait_for_restart(),
464 # with the logging bits ripped out, so that they can't cause
465 # the failure logging problem described above.
466 #
467 # The black stain that this has left on my soul can never be
468 # erased.
469 old_boot_id = self.get_boot_id()
470 if not self.wait_down(timeout=self.WAIT_DOWN_REBOOT_TIMEOUT,
471 warning_timer=self.WAIT_DOWN_REBOOT_WARNING,
472 old_boot_id=old_boot_id):
473 raise error.AutoservHostError(
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700474 'servo host %s failed to shut down.' %
475 self.hostname)
Richard Barnette3a7697f2016-04-20 11:33:27 -0700476 if self.wait_up(timeout=120):
Richard Barnette9a26ad62016-06-10 12:03:08 -0700477 current_build_number = self._get_release_version()
Richard Barnette3a7697f2016-04-20 11:33:27 -0700478 status = updater.check_update_status()
479 logging.info('servo host %s back from reboot, with build %s',
480 self.hostname, current_build_number)
481 else:
482 raise error.AutoservHostError(
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700483 'servo host %s failed to come back from reboot.' %
484 self.hostname)
Richard Barnette3a7697f2016-04-20 11:33:27 -0700485 return status, current_build_number
486
487
beeps5e8c45a2013-12-17 22:05:11 -0800488 @_timer.decorate
Richard Barnette3a7697f2016-04-20 11:33:27 -0700489 def update_image(self, wait_for_update=False):
beeps5e8c45a2013-12-17 22:05:11 -0800490 """Update the image on the servo host, if needed.
491
J. Richard Barnette84895392015-04-30 12:31:01 -0700492 This method recognizes the following cases:
493 * If the Host is not running Chrome OS, do nothing.
494 * If a previously triggered update is now complete, reboot
495 to the new version.
496 * If the host is processing a previously triggered update,
497 do nothing.
498 * If the host is running a version of Chrome OS different
499 from the default for servo Hosts, trigger an update, but
500 don't wait for it to complete.
beeps5e8c45a2013-12-17 22:05:11 -0800501
Richard Barnette3a7697f2016-04-20 11:33:27 -0700502 @param wait_for_update If an update needs to be applied and
503 this is true, then don't return until the update is
504 downloaded and finalized, and the host rebooted.
beeps5e8c45a2013-12-17 22:05:11 -0800505 @raises dev_server.DevServerException: If all the devservers are down.
506 @raises site_utils.ParseBuildNameException: If the devserver returns
507 an invalid build name.
508 @raises autoupdater.ChromiumOSError: If something goes wrong in the
509 checking update engine client status or applying an update.
510 @raises AutoservRunError: If the update_engine_client isn't present on
511 the host, and the host is a cros_host.
J. Richard Barnette84895392015-04-30 12:31:01 -0700512
beeps5e8c45a2013-12-17 22:05:11 -0800513 """
Dan Shib795b5a2015-09-24 13:26:35 -0700514 # servod could be running in a Ubuntu workstation.
Richard Barnette9a26ad62016-06-10 12:03:08 -0700515 if not self.is_cros_host():
beeps5e8c45a2013-12-17 22:05:11 -0800516 logging.info('Not attempting an update, either %s is not running '
517 'chromeos or we cannot find enough information about '
518 'the host.', self.hostname)
519 return
520
Dan Shib795b5a2015-09-24 13:26:35 -0700521 if lsbrelease_utils.is_moblab():
522 logging.info('Not attempting an update, %s is running moblab.',
523 self.hostname)
524 return
525
Richard Barnette260cbd02016-10-06 12:23:28 -0700526 target_build = afe_utils.get_stable_cros_version(self.get_board())
J. Richard Barnette84895392015-04-30 12:31:01 -0700527 target_build_number = server_site_utils.ParseBuildName(
528 target_build)[3]
beeps5e8c45a2013-12-17 22:05:11 -0800529 ds = dev_server.ImageServer.resolve(self.hostname)
J. Richard Barnette84895392015-04-30 12:31:01 -0700530 url = ds.get_update_url(target_build)
beeps5e8c45a2013-12-17 22:05:11 -0800531
532 updater = autoupdater.ChromiumOSUpdater(update_url=url, host=self)
Richard Barnette3a7697f2016-04-20 11:33:27 -0700533 status, current_build_number = self._check_for_reboot(updater)
534 update_pending = True
beeps5e8c45a2013-12-17 22:05:11 -0800535 if status in autoupdater.UPDATER_PROCESSING_UPDATE:
536 logging.info('servo host %s already processing an update, update '
537 'engine client status=%s', self.hostname, status)
J. Richard Barnette84895392015-04-30 12:31:01 -0700538 elif current_build_number != target_build_number:
beeps5e8c45a2013-12-17 22:05:11 -0800539 logging.info('Using devserver url: %s to trigger update on '
540 'servo host %s, from %s to %s', url, self.hostname,
J. Richard Barnette84895392015-04-30 12:31:01 -0700541 current_build_number, target_build_number)
beeps5e8c45a2013-12-17 22:05:11 -0800542 try:
J. Richard Barnette84895392015-04-30 12:31:01 -0700543 ds.stage_artifacts(target_build,
544 artifacts=['full_payload'])
545 except Exception as e:
546 logging.error('Staging artifacts failed: %s', str(e))
547 logging.error('Abandoning update for this cycle.')
beeps5e8c45a2013-12-17 22:05:11 -0800548 else:
J. Richard Barnette84895392015-04-30 12:31:01 -0700549 try:
Richard Barnette7e53aa02016-05-20 10:49:40 -0700550 # TODO(jrbarnette): This 'touch' is a gross hack
551 # to get us past crbug.com/613603. Once that
552 # bug is resolved, we should remove this code.
553 self.run('touch /home/chronos/.oobe_completed')
J. Richard Barnette84895392015-04-30 12:31:01 -0700554 updater.trigger_update()
555 except autoupdater.RootFSUpdateError as e:
556 trigger_download_status = 'failed with %s' % str(e)
557 autotest_stats.Counter(
558 'servo_host.RootFSUpdateError').increment()
559 else:
560 trigger_download_status = 'passed'
561 logging.info('Triggered download and update %s for %s, '
562 'update engine currently in status %s',
563 trigger_download_status, self.hostname,
564 updater.check_update_status())
beeps5e8c45a2013-12-17 22:05:11 -0800565 else:
566 logging.info('servo host %s does not require an update.',
567 self.hostname)
Richard Barnette3a7697f2016-04-20 11:33:27 -0700568 update_pending = False
569
570 if update_pending and wait_for_update:
571 logging.info('Waiting for servo update to complete.')
572 self.run('update_engine_client --follow', ignore_status=True)
beeps5e8c45a2013-12-17 22:05:11 -0800573
574
Richard Barnette9a26ad62016-06-10 12:03:08 -0700575 def verify(self):
576 """Update the servo host and verify it's in a good state."""
Richard Barnette79d78c42016-05-25 09:31:21 -0700577 # TODO(jrbarnette) Old versions of beaglebone_servo include
Richard Barnette9a26ad62016-06-10 12:03:08 -0700578 # the powerd package. If you touch the .oobe_completed file
579 # (as we do to work around an update_engine problem), then
580 # powerd will eventually shut down the beaglebone for lack
581 # of (apparent) activity. Current versions of
Richard Barnette79d78c42016-05-25 09:31:21 -0700582 # beaglebone_servo don't have powerd, but until we can purge
583 # the lab of the old images, we need to make sure powerd
584 # isn't running.
585 self.run('stop powerd', ignore_status=True)
Richard Barnette9a26ad62016-06-10 12:03:08 -0700586 try:
587 self._repair_strategy.verify(self)
588 except:
589 self.disconnect_servo()
590 raise
Fang Deng5d518f42013-08-02 14:04:32 -0700591
592
Richard Barnette9a26ad62016-06-10 12:03:08 -0700593 def repair(self):
594 """Attempt to repair servo host."""
595 try:
596 self._repair_strategy.repair(self)
597 except:
598 self.disconnect_servo()
599 raise
Fang Deng5d518f42013-08-02 14:04:32 -0700600
601
Fang Dengd4fe7392013-09-20 12:18:21 -0700602 def has_power(self):
603 """Return whether or not the servo host is powered by PoE."""
604 # TODO(fdeng): See crbug.com/302791
605 # For now, assume all servo hosts in the lab have power.
606 return self.is_in_lab()
607
608
609 def power_cycle(self):
610 """Cycle power to this host via PoE if it is a lab device.
611
Richard Barnette9a26ad62016-06-10 12:03:08 -0700612 @raises AutoservRepairError if it fails to power cycle the
Fang Dengd4fe7392013-09-20 12:18:21 -0700613 servo host.
614
615 """
616 if self.has_power():
617 try:
618 rpm_client.set_power(self.hostname, 'CYCLE')
619 except (socket.error, xmlrpclib.Error,
620 httplib.BadStatusLine,
621 rpm_client.RemotePowerException) as e:
Richard Barnette9a26ad62016-06-10 12:03:08 -0700622 raise hosts.AutoservRepairError(
Fang Dengd4fe7392013-09-20 12:18:21 -0700623 'Power cycling %s failed: %s' % (self.hostname, e))
624 else:
625 logging.info('Skipping power cycling, not a lab device.')
626
627
Dan Shi4d478522014-02-14 13:46:32 -0800628 def get_servo(self):
629 """Get the cached servo.Servo object.
Fang Deng5d518f42013-08-02 14:04:32 -0700630
Dan Shi4d478522014-02-14 13:46:32 -0800631 @return: a servo.Servo object.
Fang Deng5d518f42013-08-02 14:04:32 -0700632 """
Dan Shi4d478522014-02-14 13:46:32 -0800633 return self._servo
634
635
Richard Barnetteea3e4602016-06-10 12:36:41 -0700636def make_servo_hostname(dut_hostname):
637 """Given a DUT's hostname, return the hostname of its servo.
638
639 @param dut_hostname: hostname of a DUT.
640
641 @return hostname of the DUT's servo.
642
643 """
644 host_parts = dut_hostname.split('.')
645 host_parts[0] = host_parts[0] + '-servo'
646 return '.'.join(host_parts)
647
648
649def servo_host_is_up(servo_hostname):
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700650 """Given a servo host name, return if it's up or not.
Richard Barnetteea3e4602016-06-10 12:36:41 -0700651
652 @param servo_hostname: hostname of the servo host.
653
654 @return True if it's up, False otherwise
655 """
656 # Technically, this duplicates the SSH ping done early in the servo
657 # proxy initialization code. However, this ping ends in a couple
658 # seconds when if fails, rather than the 60 seconds it takes to decide
659 # that an SSH ping has timed out. Specifically, that timeout happens
660 # when our servo DNS name resolves, but there is no host at that IP.
661 logging.info('Pinging servo host at %s', servo_hostname)
662 ping_config = ping_runner.PingConfig(
663 servo_hostname, count=3,
664 ignore_result=True, ignore_status=True)
665 return ping_runner.PingRunner().ping(ping_config).received > 0
666
667
Richard Barnettee519dcd2016-08-15 17:37:17 -0700668def _map_afe_board_to_servo_board(afe_board):
669 """Map a board we get from the AFE to a servo appropriate value.
670
671 Many boards are identical to other boards for servo's purposes.
672 This function makes that mapping.
673
674 @param afe_board string board name received from AFE.
675 @return board we expect servo to have.
676
677 """
678 KNOWN_SUFFIXES = ['-freon', '_freon', '_moblab', '-cheets']
679 BOARD_MAP = {'gizmo': 'panther'}
680 mapped_board = afe_board
681 if afe_board in BOARD_MAP:
682 mapped_board = BOARD_MAP[afe_board]
683 else:
684 for suffix in KNOWN_SUFFIXES:
685 if afe_board.endswith(suffix):
686 mapped_board = afe_board[0:-len(suffix)]
687 break
688 if mapped_board != afe_board:
689 logging.info('Mapping AFE board=%s to %s', afe_board, mapped_board)
690 return mapped_board
691
692
Richard Barnetteea3e4602016-06-10 12:36:41 -0700693def _get_standard_servo_args(dut_host):
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700694 """Return servo data associated with a given DUT.
Richard Barnetteea3e4602016-06-10 12:36:41 -0700695
696 This checks for the presence of servo host and port attached to the
697 given `dut_host`. This data should be stored in the
Kevin Cheng05ae2a42016-06-06 10:12:48 -0700698 `_afe_host.attributes` field in the provided `dut_host` parameter.
Richard Barnetteea3e4602016-06-10 12:36:41 -0700699
700 @param dut_host Instance of `Host` on which to find the servo
701 attributes.
702 @return A tuple of `servo_args` dict with host and an option port,
703 plus an `is_in_lab` flag indicating whether this in the CrOS
704 test lab, or some different environment.
705 """
706 servo_args = None
707 is_in_lab = False
708 is_ssp_moblab = False
709 if utils.is_in_container():
710 is_moblab = _CONFIG.get_config_value(
711 'SSP', 'is_moblab', type=bool, default=False)
712 is_ssp_moblab = is_moblab
713 else:
714 is_moblab = utils.is_moblab()
Kevin Cheng05ae2a42016-06-06 10:12:48 -0700715 attrs = dut_host._afe_host.attributes
Richard Barnetteea3e4602016-06-10 12:36:41 -0700716 if attrs and SERVO_HOST_ATTR in attrs:
717 servo_host = attrs[SERVO_HOST_ATTR]
718 if (is_ssp_moblab and servo_host in ['localhost', '127.0.0.1']):
719 servo_host = _CONFIG.get_config_value(
720 'SSP', 'host_container_ip', type=str, default=None)
721 servo_args = {SERVO_HOST_ATTR: servo_host}
722 if SERVO_PORT_ATTR in attrs:
Kevin Cheng692e5292016-08-14 00:23:24 -0700723 try:
724 servo_port = attrs[SERVO_PORT_ATTR]
725 servo_args[SERVO_PORT_ATTR] = int(servo_port)
726 except ValueError:
727 logging.error('servo port is not an int: %s', servo_port)
728 # Let's set the servo args to None since we're not creating
729 # the ServoHost object with the proper port now.
730 servo_args = None
Kevin Cheng643ce8a2016-09-15 15:42:12 -0700731 if SERVO_SERIAL_ATTR in attrs:
732 servo_args[SERVO_SERIAL_ATTR] = attrs[SERVO_SERIAL_ATTR]
Richard Barnetteea3e4602016-06-10 12:36:41 -0700733 is_in_lab = (not is_moblab
734 and utils.host_is_in_lab_zone(servo_host))
735
736 # TODO(jrbarnette): This test to use the default lab servo hostname
737 # is a legacy that we need only until every host in the DB has
738 # proper attributes.
739 elif (not is_moblab and
740 not dnsname_mangler.is_ip_address(dut_host.hostname)):
741 servo_host = make_servo_hostname(dut_host.hostname)
742 is_in_lab = utils.host_is_in_lab_zone(servo_host)
743 if is_in_lab:
744 servo_args = {SERVO_HOST_ATTR: servo_host}
Richard Barnette9a26ad62016-06-10 12:03:08 -0700745 if servo_args is not None:
746 servo_board = afe_utils.get_board(dut_host)
747 if servo_board is not None:
748 servo_board = _map_afe_board_to_servo_board(servo_board)
749 servo_args[SERVO_BOARD_ATTR] = servo_board
Richard Barnetteea3e4602016-06-10 12:36:41 -0700750 return servo_args, is_in_lab
751
752
Dan Shi023aae32016-05-25 11:13:01 -0700753def create_servo_host(dut, servo_args, try_lab_servo=False,
Richard Barnette9a26ad62016-06-10 12:03:08 -0700754 try_servo_repair=False):
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700755 """Create a ServoHost object for a given DUT, if appropriate.
Dan Shi4d478522014-02-14 13:46:32 -0800756
Richard Barnette9a26ad62016-06-10 12:03:08 -0700757 This function attempts to create and verify or repair a `ServoHost`
758 object for a servo connected to the given `dut`, subject to various
759 constraints imposed by the parameters:
760 * When the `servo_args` parameter is not `None`, a servo
761 host must be created, and must be checked with `repair()`.
762 * Otherwise, if a servo exists in the lab and `try_lab_servo` is
763 true:
764 * If `try_servo_repair` is true, then create a servo host and
765 check it with `repair()`.
766 * Otherwise, if the servo responds to `ping` then create a
767 servo host and check it with `verify()`.
Fang Denge545abb2014-12-30 18:43:47 -0800768
Richard Barnette9a26ad62016-06-10 12:03:08 -0700769 In cases where `servo_args` was not `None`, repair failure
770 exceptions are passed back to the caller; otherwise, exceptions
771 are logged and then discarded.
772
773 Parameters for a servo host consist of a host name, port number, and
774 DUT board, and are determined from one of these sources, in order of
775 priority:
Richard Barnetteea3e4602016-06-10 12:36:41 -0700776 * Servo attributes from the `dut` parameter take precedence over
777 all other sources of information.
778 * If a DNS entry for the servo based on the DUT hostname exists in
779 the CrOS lab network, that hostname is used with the default
Richard Barnette9a26ad62016-06-10 12:03:08 -0700780 port and the DUT's board.
Richard Barnetteea3e4602016-06-10 12:36:41 -0700781 * If no other options are found, the parameters will be taken
Richard Barnette9a26ad62016-06-10 12:03:08 -0700782 from the `servo_args` dict passed in from the caller.
Richard Barnetteea3e4602016-06-10 12:36:41 -0700783
784 @param dut An instance of `Host` from which to take
785 servo parameters (if available).
786 @param servo_args A dictionary with servo parameters to use if
787 they can't be found from `dut`. If this
788 argument is supplied, unrepaired exceptions
789 from `verify()` will be passed back to the
790 caller.
791 @param try_lab_servo If not true, servo host creation will be
792 skipped unless otherwise required by the
793 caller.
Richard Barnette9a26ad62016-06-10 12:03:08 -0700794 @param try_servo_repair If true, check a servo host with
795 `repair()` instead of `verify()`.
Dan Shi4d478522014-02-14 13:46:32 -0800796
797 @returns: A ServoHost object or None. See comments above.
798
799 """
Richard Barnette9a26ad62016-06-10 12:03:08 -0700800 require_repair = servo_args is not None
Richard Barnetteea3e4602016-06-10 12:36:41 -0700801 is_in_lab = False
Richard Barnette9a26ad62016-06-10 12:03:08 -0700802 if dut is not None and (try_lab_servo or require_repair):
Richard Barnetteea3e4602016-06-10 12:36:41 -0700803 servo_args_override, is_in_lab = _get_standard_servo_args(dut)
804 if servo_args_override is not None:
805 servo_args = servo_args_override
806 if servo_args is None:
807 return None
Richard Barnette9a26ad62016-06-10 12:03:08 -0700808 if (not require_repair and not try_servo_repair and
809 not servo_host_is_up(servo_args[SERVO_HOST_ATTR])):
Dan Shibbb0cb62014-03-24 17:50:57 -0700810 return None
Richard Barnette9a26ad62016-06-10 12:03:08 -0700811 newhost = ServoHost(is_in_lab=is_in_lab, **servo_args)
812 # Note that the logic of repair() includes everything done
813 # by verify(). It's sufficient to call one or the other;
814 # we don't need both.
815 if require_repair:
816 newhost.repair()
817 else:
818 try:
819 if try_servo_repair:
820 newhost.repair()
821 else:
822 newhost.verify()
Kevin Cheng5f2ba6c2016-09-28 10:20:05 -0700823 except Exception:
Richard Barnette9a26ad62016-06-10 12:03:08 -0700824 operation = 'repair' if try_servo_repair else 'verification'
825 logging.exception('Servo %s failed for %s',
826 operation, newhost.hostname)
827 return newhost