blob: 35507af420fa20f02b0f1c51d7b4a491efdd6a6d [file] [log] [blame]
Aviv Keshet07f16242013-10-10 07:54:19 -07001import os, time, socket, shutil, glob, logging, traceback, tempfile
Cheng-Yi Chianga155e7e2015-08-20 20:42:04 +08002import subprocess
3
Simran Basi3b858a22015-03-17 16:23:24 -07004from multiprocessing import Lock
Aviv Keshet53a216a2013-08-27 13:58:46 -07005from autotest_lib.client.common_lib import autotemp, error
jadmanski31c49b72008-10-27 20:44:48 +00006from autotest_lib.server import utils, autotest
mblighe8b93af2009-01-30 00:45:53 +00007from autotest_lib.server.hosts import remote
mblighefccc1b2010-01-11 19:08:42 +00008from autotest_lib.client.common_lib.global_config import global_config
jadmanskica7da372008-10-21 16:26:52 +00009
Aviv Keshet53a216a2013-08-27 13:58:46 -070010# pylint: disable-msg=C0111
jadmanskica7da372008-10-21 16:26:52 +000011
mblighb86bfa12010-02-12 20:22:21 +000012get_value = global_config.get_config_value
13enable_master_ssh = get_value('AUTOSERV', 'enable_master_ssh', type=bool,
14 default=False)
mblighefccc1b2010-01-11 19:08:42 +000015
16
Fang Deng96667ca2013-08-01 17:46:18 -070017class AbstractSSHHost(remote.RemoteHost):
mblighbc9402b2009-12-29 01:15:34 +000018 """
19 This class represents a generic implementation of most of the
jadmanskica7da372008-10-21 16:26:52 +000020 framework necessary for controlling a host via ssh. It implements
21 almost all of the abstract Host methods, except for the core
mblighbc9402b2009-12-29 01:15:34 +000022 Host.run method.
23 """
jadmanskica7da372008-10-21 16:26:52 +000024
jadmanskif6562912008-10-21 17:59:01 +000025 def _initialize(self, hostname, user="root", port=22, password="",
Roshan Piusa58163a2015-10-14 13:36:29 -070026 is_client_install_supported=True, *args, **dargs):
jadmanskif6562912008-10-21 17:59:01 +000027 super(AbstractSSHHost, self)._initialize(hostname=hostname,
28 *args, **dargs)
Dan Shic07b8932014-12-11 15:22:30 -080029 # IP address is retrieved only on demand. Otherwise the host
30 # initialization will fail for host is not online.
31 self._ip = None
jadmanskica7da372008-10-21 16:26:52 +000032 self.user = user
33 self.port = port
34 self.password = password
Roshan Piusa58163a2015-10-14 13:36:29 -070035 self._is_client_install_supported = is_client_install_supported
showard6eafb492010-01-15 20:29:06 +000036 self._use_rsync = None
Fang Deng3af66202013-08-16 15:19:25 -070037 self.known_hosts_file = tempfile.mkstemp()[1]
jadmanskica7da372008-10-21 16:26:52 +000038
mblighefccc1b2010-01-11 19:08:42 +000039 """
40 Master SSH connection background job, socket temp directory and socket
41 control path option. If master-SSH is enabled, these fields will be
42 initialized by start_master_ssh when a new SSH connection is initiated.
43 """
44 self.master_ssh_job = None
45 self.master_ssh_tempdir = None
46 self.master_ssh_option = ''
47
Simran Basi3b858a22015-03-17 16:23:24 -070048 # Create a Lock to protect against race conditions.
49 self._lock = Lock()
50
showard6eafb492010-01-15 20:29:06 +000051
Dan Shic07b8932014-12-11 15:22:30 -080052 @property
53 def ip(self):
54 """@return IP address of the host.
55 """
56 if not self._ip:
57 self._ip = socket.getaddrinfo(self.hostname, None)[0][4][0]
58 return self._ip
59
60
Roshan Piusa58163a2015-10-14 13:36:29 -070061 @property
62 def is_client_install_supported(self):
63 """"
64 Returns True if the host supports autotest client installs, False
65 otherwise.
66 """
67 return self._is_client_install_supported
68
69
Fang Deng96667ca2013-08-01 17:46:18 -070070 def make_ssh_command(self, user="root", port=22, opts='',
71 hosts_file='/dev/null',
72 connect_timeout=30, alive_interval=300):
73 base_command = ("/usr/bin/ssh -a -x %s -o StrictHostKeyChecking=no "
74 "-o UserKnownHostsFile=%s -o BatchMode=yes "
75 "-o ConnectTimeout=%d -o ServerAliveInterval=%d "
76 "-l %s -p %d")
77 assert isinstance(connect_timeout, (int, long))
78 assert connect_timeout > 0 # can't disable the timeout
79 return base_command % (opts, hosts_file, connect_timeout,
80 alive_interval, user, port)
81
82
showard6eafb492010-01-15 20:29:06 +000083 def use_rsync(self):
84 if self._use_rsync is not None:
85 return self._use_rsync
86
mblighc9892c02010-01-06 19:02:16 +000087 # Check if rsync is available on the remote host. If it's not,
88 # don't try to use it for any future file transfers.
showard6eafb492010-01-15 20:29:06 +000089 self._use_rsync = self._check_rsync()
90 if not self._use_rsync:
Ilja H. Friedel04be2bd2014-05-07 21:29:59 -070091 logging.warning("rsync not available on remote host %s -- disabled",
mblighc9892c02010-01-06 19:02:16 +000092 self.hostname)
Eric Lie0493a42010-11-15 13:05:43 -080093 return self._use_rsync
mblighc9892c02010-01-06 19:02:16 +000094
95
96 def _check_rsync(self):
97 """
98 Check if rsync is available on the remote host.
99 """
100 try:
101 self.run("rsync --version", stdout_tee=None, stderr_tee=None)
102 except error.AutoservRunError:
103 return False
104 return True
105
jadmanskica7da372008-10-21 16:26:52 +0000106
showard56176ec2009-10-28 19:52:30 +0000107 def _encode_remote_paths(self, paths, escape=True):
mblighbc9402b2009-12-29 01:15:34 +0000108 """
109 Given a list of file paths, encodes it as a single remote path, in
110 the style used by rsync and scp.
111 """
showard56176ec2009-10-28 19:52:30 +0000112 if escape:
113 paths = [utils.scp_remote_escape(path) for path in paths]
114 return '%s@%s:"%s"' % (self.user, self.hostname, " ".join(paths))
jadmanskica7da372008-10-21 16:26:52 +0000115
jadmanskica7da372008-10-21 16:26:52 +0000116
mbligh45561782009-05-11 21:14:34 +0000117 def _make_rsync_cmd(self, sources, dest, delete_dest, preserve_symlinks):
mblighbc9402b2009-12-29 01:15:34 +0000118 """
119 Given a list of source paths and a destination path, produces the
jadmanskid7b79ed2009-01-07 17:19:48 +0000120 appropriate rsync command for copying them. Remote paths must be
mblighbc9402b2009-12-29 01:15:34 +0000121 pre-encoded.
122 """
Fang Deng96667ca2013-08-01 17:46:18 -0700123 ssh_cmd = self.make_ssh_command(user=self.user, port=self.port,
124 opts=self.master_ssh_option,
125 hosts_file=self.known_hosts_file)
jadmanskid7b79ed2009-01-07 17:19:48 +0000126 if delete_dest:
127 delete_flag = "--delete"
128 else:
129 delete_flag = ""
mbligh45561782009-05-11 21:14:34 +0000130 if preserve_symlinks:
131 symlink_flag = ""
132 else:
133 symlink_flag = "-L"
Dan Shi06d7fbf2014-02-12 12:34:41 -0800134 command = ("rsync %s %s --timeout=1800 --rsh='%s' -az --no-o --no-g "
David Hendricksb8904182014-06-02 15:22:49 -0700135 "%s \"%s\"")
mbligh45561782009-05-11 21:14:34 +0000136 return command % (symlink_flag, delete_flag, ssh_cmd,
David Hendricksb8904182014-06-02 15:22:49 -0700137 " ".join(['"%s"' % p for p in sources]), dest)
jadmanskid7b79ed2009-01-07 17:19:48 +0000138
139
Eric Li861b2d52011-02-04 14:50:35 -0800140 def _make_ssh_cmd(self, cmd):
141 """
142 Create a base ssh command string for the host which can be used
143 to run commands directly on the machine
144 """
Fang Deng96667ca2013-08-01 17:46:18 -0700145 base_cmd = self.make_ssh_command(user=self.user, port=self.port,
146 opts=self.master_ssh_option,
147 hosts_file=self.known_hosts_file)
Eric Li861b2d52011-02-04 14:50:35 -0800148
149 return '%s %s "%s"' % (base_cmd, self.hostname, utils.sh_escape(cmd))
150
jadmanskid7b79ed2009-01-07 17:19:48 +0000151 def _make_scp_cmd(self, sources, dest):
mblighbc9402b2009-12-29 01:15:34 +0000152 """
153 Given a list of source paths and a destination path, produces the
jadmanskid7b79ed2009-01-07 17:19:48 +0000154 appropriate scp command for encoding it. Remote paths must be
mblighbc9402b2009-12-29 01:15:34 +0000155 pre-encoded.
156 """
mblighc0649d62010-01-15 18:15:58 +0000157 command = ("scp -rq %s -o StrictHostKeyChecking=no "
lmraf676f32010-02-04 03:36:26 +0000158 "-o UserKnownHostsFile=%s -P %d %s '%s'")
Fang Deng3af66202013-08-16 15:19:25 -0700159 return command % (self.master_ssh_option, self.known_hosts_file,
mblighefccc1b2010-01-11 19:08:42 +0000160 self.port, " ".join(sources), dest)
jadmanskid7b79ed2009-01-07 17:19:48 +0000161
162
163 def _make_rsync_compatible_globs(self, path, is_local):
mblighbc9402b2009-12-29 01:15:34 +0000164 """
165 Given an rsync-style path, returns a list of globbed paths
jadmanskid7b79ed2009-01-07 17:19:48 +0000166 that will hopefully provide equivalent behaviour for scp. Does not
167 support the full range of rsync pattern matching behaviour, only that
168 exposed in the get/send_file interface (trailing slashes).
169
170 The is_local param is flag indicating if the paths should be
mblighbc9402b2009-12-29 01:15:34 +0000171 interpreted as local or remote paths.
172 """
jadmanskid7b79ed2009-01-07 17:19:48 +0000173
174 # non-trailing slash paths should just work
175 if len(path) == 0 or path[-1] != "/":
176 return [path]
177
178 # make a function to test if a pattern matches any files
179 if is_local:
showard56176ec2009-10-28 19:52:30 +0000180 def glob_matches_files(path, pattern):
181 return len(glob.glob(path + pattern)) > 0
jadmanskid7b79ed2009-01-07 17:19:48 +0000182 else:
showard56176ec2009-10-28 19:52:30 +0000183 def glob_matches_files(path, pattern):
184 result = self.run("ls \"%s\"%s" % (utils.sh_escape(path),
185 pattern),
186 stdout_tee=None, ignore_status=True)
jadmanskid7b79ed2009-01-07 17:19:48 +0000187 return result.exit_status == 0
188
189 # take a set of globs that cover all files, and see which are needed
190 patterns = ["*", ".[!.]*"]
showard56176ec2009-10-28 19:52:30 +0000191 patterns = [p for p in patterns if glob_matches_files(path, p)]
jadmanskid7b79ed2009-01-07 17:19:48 +0000192
193 # convert them into a set of paths suitable for the commandline
jadmanskid7b79ed2009-01-07 17:19:48 +0000194 if is_local:
showard56176ec2009-10-28 19:52:30 +0000195 return ["\"%s\"%s" % (utils.sh_escape(path), pattern)
196 for pattern in patterns]
jadmanskid7b79ed2009-01-07 17:19:48 +0000197 else:
showard56176ec2009-10-28 19:52:30 +0000198 return [utils.scp_remote_escape(path) + pattern
199 for pattern in patterns]
jadmanskid7b79ed2009-01-07 17:19:48 +0000200
201
202 def _make_rsync_compatible_source(self, source, is_local):
mblighbc9402b2009-12-29 01:15:34 +0000203 """
204 Applies the same logic as _make_rsync_compatible_globs, but
jadmanskid7b79ed2009-01-07 17:19:48 +0000205 applies it to an entire list of sources, producing a new list of
mblighbc9402b2009-12-29 01:15:34 +0000206 sources, properly quoted.
207 """
jadmanskid7b79ed2009-01-07 17:19:48 +0000208 return sum((self._make_rsync_compatible_globs(path, is_local)
209 for path in source), [])
jadmanskica7da372008-10-21 16:26:52 +0000210
211
mblighfeac0102009-04-28 18:31:12 +0000212 def _set_umask_perms(self, dest):
mblighbc9402b2009-12-29 01:15:34 +0000213 """
214 Given a destination file/dir (recursively) set the permissions on
215 all the files and directories to the max allowed by running umask.
216 """
mblighfeac0102009-04-28 18:31:12 +0000217
218 # now this looks strange but I haven't found a way in Python to _just_
219 # get the umask, apparently the only option is to try to set it
220 umask = os.umask(0)
221 os.umask(umask)
222
223 max_privs = 0777 & ~umask
224
225 def set_file_privs(filename):
Chris Masone567d0d92011-12-19 09:38:30 -0800226 """Sets mode of |filename|. Assumes |filename| exists."""
227 file_stat = os.stat(filename)
mblighfeac0102009-04-28 18:31:12 +0000228
229 file_privs = max_privs
230 # if the original file permissions do not have at least one
231 # executable bit then do not set it anywhere
232 if not file_stat.st_mode & 0111:
233 file_privs &= ~0111
234
235 os.chmod(filename, file_privs)
236
237 # try a bottom-up walk so changes on directory permissions won't cut
238 # our access to the files/directories inside it
239 for root, dirs, files in os.walk(dest, topdown=False):
240 # when setting the privileges we emulate the chmod "X" behaviour
241 # that sets to execute only if it is a directory or any of the
242 # owner/group/other already has execute right
243 for dirname in dirs:
244 os.chmod(os.path.join(root, dirname), max_privs)
245
Chris Masone567d0d92011-12-19 09:38:30 -0800246 # Filter out broken symlinks as we go.
247 for filename in filter(os.path.exists, files):
mblighfeac0102009-04-28 18:31:12 +0000248 set_file_privs(os.path.join(root, filename))
249
250
251 # now set privs for the dest itself
252 if os.path.isdir(dest):
253 os.chmod(dest, max_privs)
254 else:
255 set_file_privs(dest)
256
257
mbligh45561782009-05-11 21:14:34 +0000258 def get_file(self, source, dest, delete_dest=False, preserve_perm=True,
259 preserve_symlinks=False):
jadmanskica7da372008-10-21 16:26:52 +0000260 """
261 Copy files from the remote host to a local path.
262
263 Directories will be copied recursively.
264 If a source component is a directory with a trailing slash,
265 the content of the directory will be copied, otherwise, the
266 directory itself and its content will be copied. This
267 behavior is similar to that of the program 'rsync'.
268
269 Args:
270 source: either
271 1) a single file or directory, as a string
272 2) a list of one or more (possibly mixed)
273 files or directories
274 dest: a file or a directory (if source contains a
275 directory or more than one element, you must
276 supply a directory dest)
mbligh89e258d2008-10-24 13:58:08 +0000277 delete_dest: if this is true, the command will also clear
278 out any old files at dest that are not in the
279 source
mblighfeac0102009-04-28 18:31:12 +0000280 preserve_perm: tells get_file() to try to preserve the sources
281 permissions on files and dirs
mbligh45561782009-05-11 21:14:34 +0000282 preserve_symlinks: try to preserve symlinks instead of
283 transforming them into files/dirs on copy
jadmanskica7da372008-10-21 16:26:52 +0000284
285 Raises:
286 AutoservRunError: the scp command failed
287 """
Simran Basi882f15b2013-10-29 14:59:34 -0700288 logging.debug('get_file. source: %s, dest: %s, delete_dest: %s,'
289 'preserve_perm: %s, preserve_symlinks:%s', source, dest,
290 delete_dest, preserve_perm, preserve_symlinks)
mblighefccc1b2010-01-11 19:08:42 +0000291 # Start a master SSH connection if necessary.
292 self.start_master_ssh()
293
jadmanskica7da372008-10-21 16:26:52 +0000294 if isinstance(source, basestring):
295 source = [source]
jadmanskid7b79ed2009-01-07 17:19:48 +0000296 dest = os.path.abspath(dest)
jadmanskica7da372008-10-21 16:26:52 +0000297
mblighc9892c02010-01-06 19:02:16 +0000298 # If rsync is disabled or fails, try scp.
showard6eafb492010-01-15 20:29:06 +0000299 try_scp = True
300 if self.use_rsync():
Simran Basi882f15b2013-10-29 14:59:34 -0700301 logging.debug('Using Rsync.')
mblighc9892c02010-01-06 19:02:16 +0000302 try:
303 remote_source = self._encode_remote_paths(source)
304 local_dest = utils.sh_escape(dest)
305 rsync = self._make_rsync_cmd([remote_source], local_dest,
306 delete_dest, preserve_symlinks)
307 utils.run(rsync)
showard6eafb492010-01-15 20:29:06 +0000308 try_scp = False
mblighc9892c02010-01-06 19:02:16 +0000309 except error.CmdError, e:
Ilja H. Friedel04be2bd2014-05-07 21:29:59 -0700310 logging.warning("trying scp, rsync failed: %s", e)
mblighc9892c02010-01-06 19:02:16 +0000311
312 if try_scp:
Simran Basi882f15b2013-10-29 14:59:34 -0700313 logging.debug('Trying scp.')
jadmanskid7b79ed2009-01-07 17:19:48 +0000314 # scp has no equivalent to --delete, just drop the entire dest dir
315 if delete_dest and os.path.isdir(dest):
316 shutil.rmtree(dest)
317 os.mkdir(dest)
jadmanskica7da372008-10-21 16:26:52 +0000318
jadmanskid7b79ed2009-01-07 17:19:48 +0000319 remote_source = self._make_rsync_compatible_source(source, False)
320 if remote_source:
showard56176ec2009-10-28 19:52:30 +0000321 # _make_rsync_compatible_source() already did the escaping
322 remote_source = self._encode_remote_paths(remote_source,
323 escape=False)
jadmanskid7b79ed2009-01-07 17:19:48 +0000324 local_dest = utils.sh_escape(dest)
jadmanski2583a432009-02-10 23:59:11 +0000325 scp = self._make_scp_cmd([remote_source], local_dest)
jadmanskid7b79ed2009-01-07 17:19:48 +0000326 try:
327 utils.run(scp)
328 except error.CmdError, e:
Simran Basi882f15b2013-10-29 14:59:34 -0700329 logging.debug('scp failed: %s', e)
jadmanskid7b79ed2009-01-07 17:19:48 +0000330 raise error.AutoservRunError(e.args[0], e.args[1])
jadmanskica7da372008-10-21 16:26:52 +0000331
mblighfeac0102009-04-28 18:31:12 +0000332 if not preserve_perm:
333 # we have no way to tell scp to not try to preserve the
334 # permissions so set them after copy instead.
335 # for rsync we could use "--no-p --chmod=ugo=rwX" but those
336 # options are only in very recent rsync versions
337 self._set_umask_perms(dest)
338
jadmanskica7da372008-10-21 16:26:52 +0000339
mbligh45561782009-05-11 21:14:34 +0000340 def send_file(self, source, dest, delete_dest=False,
341 preserve_symlinks=False):
jadmanskica7da372008-10-21 16:26:52 +0000342 """
343 Copy files from a local path to the remote host.
344
345 Directories will be copied recursively.
346 If a source component is a directory with a trailing slash,
347 the content of the directory will be copied, otherwise, the
348 directory itself and its content will be copied. This
349 behavior is similar to that of the program 'rsync'.
350
351 Args:
352 source: either
353 1) a single file or directory, as a string
354 2) a list of one or more (possibly mixed)
355 files or directories
356 dest: a file or a directory (if source contains a
357 directory or more than one element, you must
358 supply a directory dest)
mbligh89e258d2008-10-24 13:58:08 +0000359 delete_dest: if this is true, the command will also clear
360 out any old files at dest that are not in the
361 source
mbligh45561782009-05-11 21:14:34 +0000362 preserve_symlinks: controls if symlinks on the source will be
363 copied as such on the destination or transformed into the
364 referenced file/directory
jadmanskica7da372008-10-21 16:26:52 +0000365
366 Raises:
367 AutoservRunError: the scp command failed
368 """
Simran Basi882f15b2013-10-29 14:59:34 -0700369 logging.debug('send_file. source: %s, dest: %s, delete_dest: %s,'
370 'preserve_symlinks:%s', source, dest,
371 delete_dest, preserve_symlinks)
mblighefccc1b2010-01-11 19:08:42 +0000372 # Start a master SSH connection if necessary.
373 self.start_master_ssh()
374
jadmanskica7da372008-10-21 16:26:52 +0000375 if isinstance(source, basestring):
376 source = [source]
jadmanski2583a432009-02-10 23:59:11 +0000377 remote_dest = self._encode_remote_paths([dest])
jadmanskica7da372008-10-21 16:26:52 +0000378
mukesh agrawal0d3616c2015-07-17 15:47:36 -0700379 local_sources = [utils.sh_escape(path) for path in source]
380 if not local_sources:
381 raise error.TestError('source |%s| yielded an empty list' % (
382 source))
383 if any([local_source.find('\x00') != -1 for
384 local_source in local_sources]):
385 raise error.TestError('one or more sources include NUL char')
386
mblighc9892c02010-01-06 19:02:16 +0000387 # If rsync is disabled or fails, try scp.
showard6eafb492010-01-15 20:29:06 +0000388 try_scp = True
389 if self.use_rsync():
Simran Basi882f15b2013-10-29 14:59:34 -0700390 logging.debug('Using Rsync.')
mblighc9892c02010-01-06 19:02:16 +0000391 try:
mblighc9892c02010-01-06 19:02:16 +0000392 rsync = self._make_rsync_cmd(local_sources, remote_dest,
393 delete_dest, preserve_symlinks)
394 utils.run(rsync)
showard6eafb492010-01-15 20:29:06 +0000395 try_scp = False
mblighc9892c02010-01-06 19:02:16 +0000396 except error.CmdError, e:
Ilja H. Friedel04be2bd2014-05-07 21:29:59 -0700397 logging.warning("trying scp, rsync failed: %s", e)
mblighc9892c02010-01-06 19:02:16 +0000398
399 if try_scp:
Simran Basi882f15b2013-10-29 14:59:34 -0700400 logging.debug('Trying scp.')
jadmanskid7b79ed2009-01-07 17:19:48 +0000401 # scp has no equivalent to --delete, just drop the entire dest dir
402 if delete_dest:
showard27160152009-07-15 14:28:42 +0000403 is_dir = self.run("ls -d %s/" % dest,
jadmanskid7b79ed2009-01-07 17:19:48 +0000404 ignore_status=True).exit_status == 0
405 if is_dir:
406 cmd = "rm -rf %s && mkdir %s"
mbligh5a0ca532009-08-03 16:44:34 +0000407 cmd %= (dest, dest)
jadmanskid7b79ed2009-01-07 17:19:48 +0000408 self.run(cmd)
jadmanskica7da372008-10-21 16:26:52 +0000409
jadmanski2583a432009-02-10 23:59:11 +0000410 local_sources = self._make_rsync_compatible_source(source, True)
411 if local_sources:
412 scp = self._make_scp_cmd(local_sources, remote_dest)
jadmanskid7b79ed2009-01-07 17:19:48 +0000413 try:
414 utils.run(scp)
415 except error.CmdError, e:
Simran Basi882f15b2013-10-29 14:59:34 -0700416 logging.debug('scp failed: %s', e)
jadmanskid7b79ed2009-01-07 17:19:48 +0000417 raise error.AutoservRunError(e.args[0], e.args[1])
mukesh agrawal0d3616c2015-07-17 15:47:36 -0700418 else:
419 logging.debug('skipping scp for empty source list')
jadmanskid7b79ed2009-01-07 17:19:48 +0000420
jadmanskica7da372008-10-21 16:26:52 +0000421
Simran Basi1621c632015-10-14 12:22:23 -0700422 def verify_ssh_user_access(self):
423 """Verify ssh access to this host.
424
425 @returns False if ssh_ping fails due to Permissions error, True
426 otherwise.
427 """
428 try:
429 self.ssh_ping()
430 except (error.AutoservSshPermissionDeniedError,
431 error.AutoservSshPingHostError):
432 return False
433 return True
434
435
beeps46dadc92013-11-07 14:07:10 -0800436 def ssh_ping(self, timeout=60, base_cmd='true'):
beepsadd66d32013-03-04 17:21:51 -0800437 """
438 Pings remote host via ssh.
439
440 @param timeout: Time in seconds before giving up.
441 Defaults to 60 seconds.
beeps46dadc92013-11-07 14:07:10 -0800442 @param base_cmd: The base command to run with the ssh ping.
443 Defaults to true.
beepsadd66d32013-03-04 17:21:51 -0800444 @raise AutoservSSHTimeout: If the ssh ping times out.
445 @raise AutoservSshPermissionDeniedError: If ssh ping fails due to
446 permissions.
447 @raise AutoservSshPingHostError: For other AutoservRunErrors.
448 """
jadmanskica7da372008-10-21 16:26:52 +0000449 try:
beeps46dadc92013-11-07 14:07:10 -0800450 self.run(base_cmd, timeout=timeout, connect_timeout=timeout)
jadmanskica7da372008-10-21 16:26:52 +0000451 except error.AutoservSSHTimeout:
mblighd0e94982009-07-11 00:15:18 +0000452 msg = "Host (ssh) verify timed out (timeout = %d)" % timeout
jadmanskica7da372008-10-21 16:26:52 +0000453 raise error.AutoservSSHTimeout(msg)
mbligh9d738d62009-03-09 21:17:10 +0000454 except error.AutoservSshPermissionDeniedError:
455 #let AutoservSshPermissionDeniedError be visible to the callers
456 raise
jadmanskica7da372008-10-21 16:26:52 +0000457 except error.AutoservRunError, e:
mblighc971c5f2009-06-08 16:48:54 +0000458 # convert the generic AutoservRunError into something more
459 # specific for this context
460 raise error.AutoservSshPingHostError(e.description + '\n' +
461 repr(e.result_obj))
jadmanskica7da372008-10-21 16:26:52 +0000462
463
beeps46dadc92013-11-07 14:07:10 -0800464 def is_up(self, timeout=60, base_cmd='true'):
jadmanskica7da372008-10-21 16:26:52 +0000465 """
beeps46dadc92013-11-07 14:07:10 -0800466 Check if the remote host is up by ssh-ing and running a base command.
jadmanskica7da372008-10-21 16:26:52 +0000467
beepsadd66d32013-03-04 17:21:51 -0800468 @param timeout: timeout in seconds.
beeps46dadc92013-11-07 14:07:10 -0800469 @param base_cmd: a base command to run with ssh. The default is 'true'.
beepsadd66d32013-03-04 17:21:51 -0800470 @returns True if the remote host is up before the timeout expires,
471 False otherwise.
jadmanskica7da372008-10-21 16:26:52 +0000472 """
473 try:
beeps46dadc92013-11-07 14:07:10 -0800474 self.ssh_ping(timeout=timeout, base_cmd=base_cmd)
jadmanskica7da372008-10-21 16:26:52 +0000475 except error.AutoservError:
476 return False
477 else:
478 return True
479
480
481 def wait_up(self, timeout=None):
482 """
483 Wait until the remote host is up or the timeout expires.
484
485 In fact, it will wait until an ssh connection to the remote
486 host can be established, and getty is running.
487
jadmanskic0354912010-01-12 15:57:29 +0000488 @param timeout time limit in seconds before returning even
489 if the host is not up.
jadmanskica7da372008-10-21 16:26:52 +0000490
beepsadd66d32013-03-04 17:21:51 -0800491 @returns True if the host was found to be up before the timeout expires,
492 False otherwise
jadmanskica7da372008-10-21 16:26:52 +0000493 """
494 if timeout:
beeps46dadc92013-11-07 14:07:10 -0800495 current_time = int(time.time())
496 end_time = current_time + timeout
jadmanskica7da372008-10-21 16:26:52 +0000497
beepsadd66d32013-03-04 17:21:51 -0800498 while not timeout or current_time < end_time:
499 if self.is_up(timeout=end_time - current_time):
jadmanskica7da372008-10-21 16:26:52 +0000500 try:
501 if self.are_wait_up_processes_up():
jadmanski7ebac3d2010-06-17 16:06:31 +0000502 logging.debug('Host %s is now up', self.hostname)
jadmanskica7da372008-10-21 16:26:52 +0000503 return True
504 except error.AutoservError:
505 pass
506 time.sleep(1)
beeps46dadc92013-11-07 14:07:10 -0800507 current_time = int(time.time())
jadmanskica7da372008-10-21 16:26:52 +0000508
jadmanski7ebac3d2010-06-17 16:06:31 +0000509 logging.debug('Host %s is still down after waiting %d seconds',
510 self.hostname, int(timeout + time.time() - end_time))
jadmanskica7da372008-10-21 16:26:52 +0000511 return False
512
513
jadmanskic0354912010-01-12 15:57:29 +0000514 def wait_down(self, timeout=None, warning_timer=None, old_boot_id=None):
jadmanskica7da372008-10-21 16:26:52 +0000515 """
516 Wait until the remote host is down or the timeout expires.
517
jadmanskic0354912010-01-12 15:57:29 +0000518 If old_boot_id is provided, this will wait until either the machine
519 is unpingable or self.get_boot_id() returns a value different from
520 old_boot_id. If the boot_id value has changed then the function
521 returns true under the assumption that the machine has shut down
522 and has now already come back up.
jadmanskica7da372008-10-21 16:26:52 +0000523
jadmanskic0354912010-01-12 15:57:29 +0000524 If old_boot_id is None then until the machine becomes unreachable the
525 method assumes the machine has not yet shut down.
jadmanskica7da372008-10-21 16:26:52 +0000526
beepsadd66d32013-03-04 17:21:51 -0800527 Based on this definition, the 4 possible permutations of timeout
528 and old_boot_id are:
529 1. timeout and old_boot_id: wait timeout seconds for either the
530 host to become unpingable, or the boot id
531 to change. In the latter case we've rebooted
532 and in the former case we've only shutdown,
533 but both cases return True.
534 2. only timeout: wait timeout seconds for the host to become unpingable.
535 If the host remains pingable throughout timeout seconds
536 we return False.
537 3. only old_boot_id: wait forever until either the host becomes
538 unpingable or the boot_id changes. Return true
539 when either of those conditions are met.
540 4. not timeout, not old_boot_id: wait forever till the host becomes
541 unpingable.
542
jadmanskic0354912010-01-12 15:57:29 +0000543 @param timeout Time limit in seconds before returning even
544 if the host is still up.
545 @param warning_timer Time limit in seconds that will generate
546 a warning if the host is not down yet.
547 @param old_boot_id A string containing the result of self.get_boot_id()
548 prior to the host being told to shut down. Can be None if this is
549 not available.
550
551 @returns True if the host was found to be down, False otherwise
jadmanskica7da372008-10-21 16:26:52 +0000552 """
mblighe5e3cf22010-05-27 23:33:14 +0000553 #TODO: there is currently no way to distinguish between knowing
554 #TODO: boot_id was unsupported and not knowing the boot_id.
beeps46dadc92013-11-07 14:07:10 -0800555 current_time = int(time.time())
jadmanskica7da372008-10-21 16:26:52 +0000556 if timeout:
mbligh2ed998f2009-04-08 21:03:47 +0000557 end_time = current_time + timeout
jadmanskica7da372008-10-21 16:26:52 +0000558
mbligh2ed998f2009-04-08 21:03:47 +0000559 if warning_timer:
560 warn_time = current_time + warning_timer
561
jadmanskic0354912010-01-12 15:57:29 +0000562 if old_boot_id is not None:
563 logging.debug('Host %s pre-shutdown boot_id is %s',
564 self.hostname, old_boot_id)
565
beepsadd66d32013-03-04 17:21:51 -0800566 # Impose semi real-time deadline constraints, since some clients
567 # (eg: watchdog timer tests) expect strict checking of time elapsed.
568 # Each iteration of this loop is treated as though it atomically
569 # completes within current_time, this is needed because if we used
570 # inline time.time() calls instead then the following could happen:
571 #
572 # while not timeout or time.time() < end_time: [23 < 30]
573 # some code. [takes 10 secs]
574 # try:
575 # new_boot_id = self.get_boot_id(timeout=end_time - time.time())
576 # [30 - 33]
577 # The last step will lead to a return True, when in fact the machine
578 # went down at 32 seconds (>30). Hence we need to pass get_boot_id
579 # the same time that allowed us into that iteration of the loop.
mbligh2ed998f2009-04-08 21:03:47 +0000580 while not timeout or current_time < end_time:
jadmanskic0354912010-01-12 15:57:29 +0000581 try:
beeps46dadc92013-11-07 14:07:10 -0800582 new_boot_id = self.get_boot_id(timeout=end_time-current_time)
mblighdbc7e4a2010-01-15 20:34:20 +0000583 except error.AutoservError:
jadmanskic0354912010-01-12 15:57:29 +0000584 logging.debug('Host %s is now unreachable over ssh, is down',
585 self.hostname)
jadmanskica7da372008-10-21 16:26:52 +0000586 return True
jadmanskic0354912010-01-12 15:57:29 +0000587 else:
588 # if the machine is up but the boot_id value has changed from
589 # old boot id, then we can assume the machine has gone down
590 # and then already come back up
591 if old_boot_id is not None and old_boot_id != new_boot_id:
592 logging.debug('Host %s now has boot_id %s and so must '
593 'have rebooted', self.hostname, new_boot_id)
594 return True
mbligh2ed998f2009-04-08 21:03:47 +0000595
596 if warning_timer and current_time > warn_time:
Scott Zawalskic86fdeb2013-10-23 10:24:04 -0400597 self.record("INFO", None, "shutdown",
mbligh2ed998f2009-04-08 21:03:47 +0000598 "Shutdown took longer than %ds" % warning_timer)
599 # Print the warning only once.
600 warning_timer = None
mbligha4464402009-04-17 20:13:41 +0000601 # If a machine is stuck switching runlevels
602 # This may cause the machine to reboot.
603 self.run('kill -HUP 1', ignore_status=True)
mbligh2ed998f2009-04-08 21:03:47 +0000604
jadmanskica7da372008-10-21 16:26:52 +0000605 time.sleep(1)
beeps46dadc92013-11-07 14:07:10 -0800606 current_time = int(time.time())
jadmanskica7da372008-10-21 16:26:52 +0000607
608 return False
jadmanskif6562912008-10-21 17:59:01 +0000609
mbligha0a27592009-01-24 01:41:36 +0000610
jadmanskif6562912008-10-21 17:59:01 +0000611 # tunable constants for the verify & repair code
mblighb86bfa12010-02-12 20:22:21 +0000612 AUTOTEST_GB_DISKSPACE_REQUIRED = get_value("SERVER",
613 "gb_diskspace_required",
Fang Deng6b05f5b2013-03-20 13:42:11 -0700614 type=float,
615 default=20.0)
mbligha0a27592009-01-24 01:41:36 +0000616
jadmanskif6562912008-10-21 17:59:01 +0000617
showardca572982009-09-18 21:20:01 +0000618 def verify_connectivity(self):
619 super(AbstractSSHHost, self).verify_connectivity()
jadmanskif6562912008-10-21 17:59:01 +0000620
showardb18134f2009-03-20 20:52:18 +0000621 logging.info('Pinging host ' + self.hostname)
jadmanskif6562912008-10-21 17:59:01 +0000622 self.ssh_ping()
mbligh2ba7ab02009-08-24 22:09:26 +0000623 logging.info("Host (ssh) %s is alive", self.hostname)
jadmanskif6562912008-10-21 17:59:01 +0000624
jadmanski80deb752009-01-21 17:14:16 +0000625 if self.is_shutting_down():
mblighc971c5f2009-06-08 16:48:54 +0000626 raise error.AutoservHostIsShuttingDownError("Host is shutting down")
jadmanski80deb752009-01-21 17:14:16 +0000627
mblighb49b5232009-02-12 21:54:49 +0000628
showardca572982009-09-18 21:20:01 +0000629 def verify_software(self):
630 super(AbstractSSHHost, self).verify_software()
jadmanskif6562912008-10-21 17:59:01 +0000631 try:
showardad812bf2009-10-20 23:49:56 +0000632 self.check_diskspace(autotest.Autotest.get_install_dir(self),
633 self.AUTOTEST_GB_DISKSPACE_REQUIRED)
jadmanskif6562912008-10-21 17:59:01 +0000634 except error.AutoservHostError:
635 raise # only want to raise if it's a space issue
showardad812bf2009-10-20 23:49:56 +0000636 except autotest.AutodirNotFoundError:
showardca572982009-09-18 21:20:01 +0000637 # autotest dir may not exist, etc. ignore
638 logging.debug('autodir space check exception, this is probably '
639 'safe to ignore\n' + traceback.format_exc())
mblighefccc1b2010-01-11 19:08:42 +0000640
641
642 def close(self):
643 super(AbstractSSHHost, self).close()
644 self._cleanup_master_ssh()
Fang Deng3af66202013-08-16 15:19:25 -0700645 os.remove(self.known_hosts_file)
mblighefccc1b2010-01-11 19:08:42 +0000646
647
648 def _cleanup_master_ssh(self):
649 """
650 Release all resources (process, temporary directory) used by an active
651 master SSH connection.
652 """
653 # If a master SSH connection is running, kill it.
654 if self.master_ssh_job is not None:
Aviv Keshet46250752013-08-27 15:52:06 -0700655 logging.debug('Nuking master_ssh_job.')
mblighefccc1b2010-01-11 19:08:42 +0000656 utils.nuke_subprocess(self.master_ssh_job.sp)
657 self.master_ssh_job = None
658
659 # Remove the temporary directory for the master SSH socket.
660 if self.master_ssh_tempdir is not None:
Aviv Keshet46250752013-08-27 15:52:06 -0700661 logging.debug('Cleaning master_ssh_tempdir.')
mblighefccc1b2010-01-11 19:08:42 +0000662 self.master_ssh_tempdir.clean()
663 self.master_ssh_tempdir = None
664 self.master_ssh_option = ''
665
666
Aviv Keshet0749a822013-10-17 09:53:26 -0700667 def start_master_ssh(self, timeout=5):
mblighefccc1b2010-01-11 19:08:42 +0000668 """
669 Called whenever a slave SSH connection needs to be initiated (e.g., by
670 run, rsync, scp). If master SSH support is enabled and a master SSH
671 connection is not active already, start a new one in the background.
672 Also, cleanup any zombie master SSH connections (e.g., dead due to
673 reboot).
Aviv Keshet0749a822013-10-17 09:53:26 -0700674
675 timeout: timeout in seconds (default 5) to wait for master ssh
676 connection to be established. If timeout is reached, a
677 warning message is logged, but no other action is taken.
mblighefccc1b2010-01-11 19:08:42 +0000678 """
679 if not enable_master_ssh:
680 return
681
Simran Basi3b858a22015-03-17 16:23:24 -0700682 # Multiple processes might try in parallel to clean up the old master
683 # ssh connection and create a new one, therefore use a lock to protect
684 # against race conditions.
685 with self._lock:
686 # If a previously started master SSH connection is not running
687 # anymore, it needs to be cleaned up and then restarted.
688 if self.master_ssh_job is not None:
689 socket_path = os.path.join(self.master_ssh_tempdir.name,
690 'socket')
691 if (not os.path.exists(socket_path) or
692 self.master_ssh_job.sp.poll() is not None):
693 logging.info("Master ssh connection to %s is down.",
694 self.hostname)
695 self._cleanup_master_ssh()
mblighefccc1b2010-01-11 19:08:42 +0000696
Simran Basi3b858a22015-03-17 16:23:24 -0700697 # Start a new master SSH connection.
698 if self.master_ssh_job is None:
699 # Create a shared socket in a temp location.
700 self.master_ssh_tempdir = autotemp.tempdir(
701 unique_id='ssh-master')
702 self.master_ssh_option = ("-o ControlPath=%s/socket" %
703 self.master_ssh_tempdir.name)
mblighefccc1b2010-01-11 19:08:42 +0000704
Simran Basi3b858a22015-03-17 16:23:24 -0700705 # Start the master SSH connection in the background.
706 master_cmd = self.ssh_command(
707 options="-N -o ControlMaster=yes")
708 logging.info("Starting master ssh connection '%s'", master_cmd)
709 self.master_ssh_job = utils.BgJob(master_cmd,
710 nickname='master-ssh',
711 no_pipes=True)
712 # To prevent a race between the the master ssh connection
713 # startup and its first attempted use, wait for socket file to
714 # exist before returning.
715 end_time = time.time() + timeout
716 socket_file_path = os.path.join(self.master_ssh_tempdir.name,
717 'socket')
718 while time.time() < end_time:
719 if os.path.exists(socket_file_path):
720 break
721 time.sleep(.2)
722 else:
723 logging.info('Timed out waiting for master-ssh connection '
724 'to be established.')
mbligh0a883702010-04-21 01:58:34 +0000725
726
727 def clear_known_hosts(self):
728 """Clears out the temporary ssh known_hosts file.
729
730 This is useful if the test SSHes to the machine, then reinstalls it,
731 then SSHes to it again. It can be called after the reinstall to
732 reduce the spam in the logs.
733 """
734 logging.info("Clearing known hosts for host '%s', file '%s'.",
Fang Deng3af66202013-08-16 15:19:25 -0700735 self.hostname, self.known_hosts_file)
mbligh0a883702010-04-21 01:58:34 +0000736 # Clear out the file by opening it for writing and then closing.
Fang Deng3af66202013-08-16 15:19:25 -0700737 fh = open(self.known_hosts_file, "w")
mbligh0a883702010-04-21 01:58:34 +0000738 fh.close()
Prashanth B98509c72014-04-04 16:01:34 -0700739
740
741 def collect_logs(self, remote_src_dir, local_dest_dir, ignore_errors=True):
742 """Copy log directories from a host to a local directory.
743
744 @param remote_src_dir: A destination directory on the host.
745 @param local_dest_dir: A path to a local destination directory.
746 If it doesn't exist it will be created.
747 @param ignore_errors: If True, ignore exceptions.
748
749 @raises OSError: If there were problems creating the local_dest_dir and
750 ignore_errors is False.
751 @raises AutoservRunError, AutotestRunError: If something goes wrong
752 while copying the directories and ignore_errors is False.
753 """
754 locally_created_dest = False
755 if (not os.path.exists(local_dest_dir)
756 or not os.path.isdir(local_dest_dir)):
757 try:
758 os.makedirs(local_dest_dir)
759 locally_created_dest = True
760 except OSError as e:
761 logging.warning('Unable to collect logs from host '
762 '%s: %s', self.hostname, e)
763 if not ignore_errors:
764 raise
765 return
766 try:
767 self.get_file(
768 remote_src_dir, local_dest_dir, preserve_symlinks=True)
769 except (error.AutotestRunError, error.AutoservRunError,
770 error.AutoservSSHTimeout) as e:
771 logging.warning('Collection of %s to local dir %s from host %s '
772 'failed: %s', remote_src_dir, local_dest_dir,
773 self.hostname, e)
774 if locally_created_dest:
775 shutil.rmtree(local_dest_dir, ignore_errors=ignore_errors)
776 if not ignore_errors:
777 raise
Cheng-Yi Chianga155e7e2015-08-20 20:42:04 +0800778
779
780 def _create_ssh_tunnel(self, port, local_port):
781 """Create an ssh tunnel from local_port to port.
782
783 @param port: remote port on the host.
784 @param local_port: local forwarding port.
785
786 @return: the tunnel process.
787 """
788 tunnel_options = '-n -N -q -L %d:localhost:%d' % (local_port, port)
789 ssh_cmd = self.make_ssh_command(opts=tunnel_options)
790 tunnel_cmd = '%s %s' % (ssh_cmd, self.hostname)
791 logging.debug('Full tunnel command: %s', tunnel_cmd)
792 tunnel_proc = subprocess.Popen(tunnel_cmd, shell=True, close_fds=True)
793 logging.debug('Started ssh tunnel, local = %d'
794 ' remote = %d, pid = %d',
795 local_port, port, tunnel_proc.pid)
796 return tunnel_proc
Gilad Arnolda76bef02015-09-29 13:55:15 -0700797
798
799 def get_os_type(self):
800 """Returns the host OS descriptor (to be implemented in subclasses).
801
802 @return A string describing the OS type.
803 """
804 raise NotImplementedError