1
# Copyright (C) 2006, 2007, 2008, 2010 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""On-disk mutex protecting a resource
19
bzr on-disk objects are locked by the existence of a directory with a
20
particular name within the control directory. We use this rather than OS
21
internal locks (such as flock etc) because they can be seen across all
22
transports, including http.
24
Objects can be read if there is only physical read access; therefore
25
readers can never be required to create a lock, though they will
26
check whether a writer is using the lock. Writers can't detect
27
whether anyone else is reading from the resource as they write.
28
This works because of ordering constraints that make sure readers
29
see a consistent view of existing data.
31
Waiting for a lock must be done by polling; this can be aborted after
34
Locks must always be explicitly released, typically from a try/finally
35
block -- they are not released from a finalizer or when Python
38
Locks may fail to be released if the process is abruptly terminated
39
(machine stop, SIGKILL) or if a remote transport becomes permanently
40
disconnected. There is therefore a method to break an existing lock.
41
This should rarely be used, and generally only with user approval.
42
Locks contain some information on when the lock was taken and by who
43
which may guide in deciding whether it can safely be broken. (This is
44
similar to the messages displayed by emacs and vim.) Note that if the
45
lock holder is still alive they will get no notification that the lock
46
has been broken and will continue their work -- so it is important to be
47
sure they are actually dead.
49
A lock is represented on disk by a directory of a particular name,
50
containing an information file. Taking a lock is done by renaming a
51
temporary directory into place. We use temporary directories because
52
for all known transports and filesystems we believe that exactly one
53
attempt to claim the lock will succeed and the others will fail. (Files
54
won't do because some filesystems or transports only have
55
rename-and-overwrite, making it hard to tell who won.)
57
The desired characteristics are:
59
* Locks are not reentrant. (That is, a client that tries to take a
60
lock it already holds may deadlock or fail.)
61
* Stale locks can be guessed at by a heuristic
62
* Lost locks can be broken by any client
63
* Failed lock operations leave little or no mess
64
* Deadlocks are avoided by having a timeout always in use, clients
65
desiring indefinite waits can retry or set a silly big timeout.
67
Storage formats use the locks, and also need to consider concurrency
68
issues underneath the lock. A format may choose not to use a lock
69
at all for some operations.
71
LockDirs always operate over a Transport. The transport may be readonly, in
72
which case the lock can be queried but not acquired.
74
Locks are identified by a path name, relative to a base transport.
76
Calling code will typically want to make sure there is exactly one LockDir
77
object per actual lock on disk. This module does nothing to prevent aliasing
78
and deadlocks will likely occur if the locks are aliased.
80
In the future we may add a "freshen" method which can be called
81
by a lock holder to check that their lock has not been broken, and to
82
update the timestamp within it.
86
>>> from bzrlib.transport.memory import MemoryTransport
87
>>> # typically will be obtained from a BzrDir, Branch, etc
88
>>> t = MemoryTransport()
89
>>> l = LockDir(t, 'sample-lock')
91
>>> token = l.wait_lock()
92
>>> # do something here
98
# TODO: We sometimes have the problem that our attempt to rename '1234' to
99
# 'held' fails because the transport server moves into an existing directory,
100
# rather than failing the rename. If we made the info file name the same as
101
# the locked directory name we would avoid this problem because moving into
102
# the held directory would implicitly clash. However this would not mesh with
103
# the existing locking code and needs a new format of the containing object.
104
# -- robertc, mbp 20070628
115
from bzrlib.errors import (
128
from bzrlib.trace import mutter, note
129
from bzrlib.osutils import format_delta, rand_chars, get_host_name
132
from bzrlib.lazy_import import lazy_import
133
lazy_import(globals(), """
134
from bzrlib import rio
137
# XXX: At the moment there is no consideration of thread safety on LockDir
138
# objects. This should perhaps be updated - e.g. if two threads try to take a
139
# lock at the same time they should *both* get it. But then that's unlikely
142
# TODO: Perhaps store some kind of note like the bzr command line in the lock
145
# TODO: Some kind of callback run while polling a lock to show progress
148
# TODO: Make sure to pass the right file and directory mode bits to all
149
# files/dirs created.
152
_DEFAULT_TIMEOUT_SECONDS = 300
153
_DEFAULT_POLL_SECONDS = 1.0
156
class LockDir(lock.Lock):
157
"""Write-lock guarding access to data.
160
__INFO_NAME = '/info'
162
def __init__(self, transport, path, file_modebits=0644, dir_modebits=0755):
163
"""Create a new LockDir object.
165
The LockDir is initially unlocked - this just creates the object.
167
:param transport: Transport which will contain the lock
169
:param path: Path to the lock within the base directory of the
172
self.transport = transport
174
self._lock_held = False
175
self._locked_via_token = False
176
self._fake_read_lock = False
177
self._held_dir = path + '/held'
178
self._held_info_path = self._held_dir + self.__INFO_NAME
179
self._file_modebits = file_modebits
180
self._dir_modebits = dir_modebits
182
self._report_function = note
185
return '%s(%s%s)' % (self.__class__.__name__,
189
is_held = property(lambda self: self._lock_held)
191
def create(self, mode=None):
192
"""Create the on-disk lock.
194
This is typically only called when the object/directory containing the
195
directory is first created. The lock is not held when it's created.
197
self._trace("create lock directory")
199
self.transport.mkdir(self.path, mode=mode)
200
except (TransportError, PathError), e:
201
raise LockFailed(self, e)
204
def _attempt_lock(self):
205
"""Make the pending directory and attempt to rename into place.
207
If the rename succeeds, we read back the info file to check that we
210
If we fail to acquire the lock, this method is responsible for
211
cleaning up the pending directory if possible. (But it doesn't do
214
:returns: The nonce of the lock, if it was successfully acquired.
216
:raises LockContention: If the lock is held by someone else. The exception
217
contains the info of the current holder of the lock.
219
self._trace("lock_write...")
220
start_time = time.time()
222
tmpname = self._create_pending_dir()
223
except (errors.TransportError, PathError), e:
224
self._trace("... failed to create pending dir, %s", e)
225
raise LockFailed(self, e)
227
self.transport.rename(tmpname, self._held_dir)
228
except (errors.TransportError, PathError, DirectoryNotEmpty,
229
FileExists, ResourceBusy), e:
230
self._trace("... contention, %s", e)
231
self._remove_pending_dir(tmpname)
232
raise LockContention(self)
234
self._trace("... lock failed, %s", e)
235
self._remove_pending_dir(tmpname)
237
# We must check we really got the lock, because Launchpad's sftp
238
# server at one time had a bug were the rename would successfully
239
# move the new directory into the existing directory, which was
240
# incorrect. It's possible some other servers or filesystems will
241
# have a similar bug allowing someone to think they got the lock
242
# when it's already held.
244
# See <https://bugs.edge.launchpad.net/bzr/+bug/498378> for one case.
246
# Strictly the check is unnecessary and a waste of time for most
247
# people, but probably worth trapping if something is wrong.
249
self._trace("after locking, info=%r", info)
251
raise LockFailed(self, "lock was renamed into place, but "
253
if info.get('nonce') != self.nonce:
254
self._trace("rename succeeded, "
255
"but lock is still held by someone else")
256
raise LockContention(self)
257
self._lock_held = True
258
self._trace("... lock succeeded after %dms",
259
(time.time() - start_time) * 1000)
262
def _remove_pending_dir(self, tmpname):
263
"""Remove the pending directory
265
This is called if we failed to rename into place, so that the pending
266
dirs don't clutter up the lockdir.
268
self._trace("remove %s", tmpname)
270
self.transport.delete(tmpname + self.__INFO_NAME)
271
self.transport.rmdir(tmpname)
273
note("error removing pending lock: %s", e)
275
def _create_pending_dir(self):
276
tmpname = '%s/%s.tmp' % (self.path, rand_chars(10))
278
self.transport.mkdir(tmpname)
280
# This may raise a FileExists exception
281
# which is okay, it will be caught later and determined
282
# to be a LockContention.
283
self._trace("lock directory does not exist, creating it")
284
self.create(mode=self._dir_modebits)
285
# After creating the lock directory, try again
286
self.transport.mkdir(tmpname)
287
self.nonce = rand_chars(20)
288
info_bytes = self._prepare_info()
289
# We use put_file_non_atomic because we just created a new unique
290
# directory so we don't have to worry about files existing there.
291
# We'll rename the whole directory into place to get atomic
293
self.transport.put_bytes_non_atomic(tmpname + self.__INFO_NAME,
298
"""Release a held lock
300
if self._fake_read_lock:
301
self._fake_read_lock = False
303
if not self._lock_held:
304
return lock.cant_unlock_not_held(self)
305
if self._locked_via_token:
306
self._locked_via_token = False
307
self._lock_held = False
309
old_nonce = self.nonce
310
# rename before deleting, because we can't atomically remove the
312
start_time = time.time()
313
self._trace("unlocking")
314
tmpname = '%s/releasing.%s.tmp' % (self.path, rand_chars(20))
315
# gotta own it to unlock
317
self.transport.rename(self._held_dir, tmpname)
318
self._lock_held = False
319
self.transport.delete(tmpname + self.__INFO_NAME)
321
self.transport.rmdir(tmpname)
322
except DirectoryNotEmpty, e:
323
# There might have been junk left over by a rename that moved
324
# another locker within the 'held' directory. do a slower
325
# deletion where we list the directory and remove everything
328
# Maybe this should be broader to allow for ftp servers with
329
# non-specific error messages?
330
self._trace("doing recursive deletion of non-empty directory "
332
self.transport.delete_tree(tmpname)
333
self._trace("... unlock succeeded after %dms",
334
(time.time() - start_time) * 1000)
335
result = lock.LockResult(self.transport.abspath(self.path),
337
for hook in self.hooks['lock_released']:
340
def break_lock(self):
341
"""Break a lock not held by this instance of LockDir.
343
This is a UI centric function: it uses the bzrlib.ui.ui_factory to
344
prompt for input if a lock is detected and there is any doubt about
345
it possibly being still active.
347
self._check_not_locked()
348
holder_info = self.peek()
349
if holder_info is not None:
350
lock_info = '\n'.join(self._format_lock_info(holder_info))
351
if bzrlib.ui.ui_factory.get_boolean("Break %s" % lock_info):
352
self.force_break(holder_info)
354
def force_break(self, dead_holder_info):
355
"""Release a lock held by another process.
357
WARNING: This should only be used when the other process is dead; if
358
it still thinks it has the lock there will be two concurrent writers.
359
In general the user's approval should be sought for lock breaks.
361
dead_holder_info must be the result of a previous LockDir.peek() call;
362
this is used to check that it's still held by the same process that
363
the user decided was dead. If this is not the current holder,
364
LockBreakMismatch is raised.
366
After the lock is broken it will not be held by any process.
367
It is possible that another process may sneak in and take the
368
lock before the breaking process acquires it.
370
if not isinstance(dead_holder_info, dict):
371
raise ValueError("dead_holder_info: %r" % dead_holder_info)
372
self._check_not_locked()
373
current_info = self.peek()
374
if current_info is None:
375
# must have been recently released
377
if current_info != dead_holder_info:
378
raise LockBreakMismatch(self, current_info, dead_holder_info)
379
tmpname = '%s/broken.%s.tmp' % (self.path, rand_chars(20))
380
self.transport.rename(self._held_dir, tmpname)
381
# check that we actually broke the right lock, not someone else;
382
# there's a small race window between checking it and doing the
384
broken_info_path = tmpname + self.__INFO_NAME
385
broken_info = self._read_info_file(broken_info_path)
386
if broken_info != dead_holder_info:
387
raise LockBreakMismatch(self, broken_info, dead_holder_info)
388
self.transport.delete(broken_info_path)
389
self.transport.rmdir(tmpname)
390
result = lock.LockResult(self.transport.abspath(self.path),
391
current_info.get('nonce'))
392
for hook in self.hooks['lock_broken']:
395
def _check_not_locked(self):
396
"""If the lock is held by this instance, raise an error."""
398
raise AssertionError("can't break own lock: %r" % self)
401
"""Make sure that the lock is still held by this locker.
403
This should only fail if the lock was broken by user intervention,
404
or if the lock has been affected by a bug.
406
If the lock is not thought to be held, raises LockNotHeld. If
407
the lock is thought to be held but has been broken, raises
410
if not self._lock_held:
411
raise LockNotHeld(self)
414
# no lock there anymore!
415
raise LockBroken(self)
416
if info.get('nonce') != self.nonce:
417
# there is a lock, but not ours
418
raise LockBroken(self)
420
def _read_info_file(self, path):
421
"""Read one given info file.
423
peek() reads the info file of the lock holder, if any.
425
return self._parse_info(self.transport.get(path))
428
"""Check if the lock is held by anyone.
430
If it is held, this returns the lock info structure as a dict
431
which contains some information about the current lock holder.
432
Otherwise returns None.
435
info = self._read_info_file(self._held_info_path)
436
self._trace("peek -> held")
438
except NoSuchFile, e:
439
self._trace("peek -> not held")
441
def _prepare_info(self):
442
"""Write information about a pending lock to a temporary file.
444
# XXX: is creating this here inefficient?
445
config = bzrlib.config.GlobalConfig()
447
user = config.user_email()
448
except errors.NoEmailInUsername:
449
user = config.username()
450
s = rio.Stanza(hostname=get_host_name(),
451
pid=str(os.getpid()),
452
start_time=str(int(time.time())),
458
def _parse_info(self, info_file):
459
stanza = rio.read_stanza(info_file.readlines())
461
# see bug 185013; we fairly often end up with the info file being
462
# empty after an interruption; we could log a message here but
463
# there may not be much we can say
466
return stanza.as_dict()
468
def attempt_lock(self):
469
"""Take the lock; fail if it's already held.
471
If you wish to block until the lock can be obtained, call wait_lock()
474
:return: The lock token.
475
:raises LockContention: if the lock is held by someone else.
477
if self._fake_read_lock:
478
raise LockContention(self)
479
result = self._attempt_lock()
480
hook_result = lock.LockResult(self.transport.abspath(self.path),
482
for hook in self.hooks['lock_acquired']:
486
def wait_lock(self, timeout=None, poll=None, max_attempts=None):
487
"""Wait a certain period for a lock.
489
If the lock can be acquired within the bounded time, it
490
is taken and this returns. Otherwise, LockContention
491
is raised. Either way, this function should return within
492
approximately `timeout` seconds. (It may be a bit more if
493
a transport operation takes a long time to complete.)
495
:param timeout: Approximate maximum amount of time to wait for the
498
:param poll: Delay in seconds between retrying the lock.
500
:param max_attempts: Maximum number of times to try to lock.
502
:return: The lock token.
505
timeout = _DEFAULT_TIMEOUT_SECONDS
507
poll = _DEFAULT_POLL_SECONDS
508
# XXX: the transport interface doesn't let us guard against operations
509
# there taking a long time, so the total elapsed time or poll interval
510
# may be more than was requested.
511
deadline = time.time() + timeout
518
return self.attempt_lock()
519
except LockContention:
520
# possibly report the blockage, then try again
522
# TODO: In a few cases, we find out that there's contention by
523
# reading the held info and observing that it's not ours. In
524
# those cases it's a bit redundant to read it again. However,
525
# the normal case (??) is that the rename fails and so we
526
# don't know who holds the lock. For simplicity we peek
528
new_info = self.peek()
529
if new_info is not None and new_info != last_info:
530
if last_info is None:
531
start = 'Unable to obtain'
533
start = 'Lock owner changed for'
535
formatted_info = self._format_lock_info(new_info)
536
if deadline_str is None:
537
deadline_str = time.strftime('%H:%M:%S',
538
time.localtime(deadline))
539
lock_url = self.transport.abspath(self.path)
540
self._report_function('%s %s\n'
542
'%s\n' # locked ... ago
543
'Will continue to try until %s, unless '
545
'If you\'re sure that it\'s not being '
546
'modified, use bzr break-lock %s',
554
if (max_attempts is not None) and (attempt_count >= max_attempts):
555
self._trace("exceeded %d attempts")
556
raise LockContention(self)
557
if time.time() + poll < deadline:
558
self._trace("waiting %ss", poll)
561
self._trace("timeout after waiting %ss", timeout)
562
raise LockContention(self)
564
def leave_in_place(self):
565
self._locked_via_token = True
567
def dont_leave_in_place(self):
568
self._locked_via_token = False
570
def lock_write(self, token=None):
571
"""Wait for and acquire the lock.
573
:param token: if this is already locked, then lock_write will fail
574
unless the token matches the existing lock.
575
:returns: a token if this instance supports tokens, otherwise None.
576
:raises TokenLockingNotSupported: when a token is given but this
577
instance doesn't support using token locks.
578
:raises MismatchedToken: if the specified token doesn't match the token
579
of the existing lock.
581
A token should be passed in if you know that you have locked the object
582
some other way, and need to synchronise this object's state with that
585
XXX: docstring duplicated from LockableFiles.lock_write.
587
if token is not None:
588
self.validate_token(token)
590
self._lock_held = True
591
self._locked_via_token = True
594
return self.wait_lock()
597
"""Compatibility-mode shared lock.
599
LockDir doesn't support shared read-only locks, so this
600
just pretends that the lock is taken but really does nothing.
602
# At the moment Branches are commonly locked for read, but
603
# we can't rely on that remotely. Once this is cleaned up,
604
# reenable this warning to prevent it coming back in
606
## warn("LockDir.lock_read falls back to write lock")
607
if self._lock_held or self._fake_read_lock:
608
raise LockContention(self)
609
self._fake_read_lock = True
611
def _format_lock_info(self, info):
612
"""Turn the contents of peek() into something for the user"""
613
lock_url = self.transport.abspath(self.path)
614
start_time = info.get('start_time')
615
if start_time is None:
616
time_ago = '(unknown)'
618
time_ago = format_delta(time.time() - int(info['start_time']))
620
'lock %s' % (lock_url,),
621
'held by %s on host %s [process #%s]' %
622
tuple([info.get(x, '<unknown>') for x in ['user', 'hostname', 'pid']]),
623
'locked %s' % (time_ago,),
626
def validate_token(self, token):
627
if token is not None:
633
lock_token = info.get('nonce')
634
if token != lock_token:
635
raise errors.TokenMismatch(token, lock_token)
637
self._trace("revalidated by token %r", token)
639
def _trace(self, format, *args):
640
if 'lock' not in debug.debug_flags:
642
mutter(str(self) + ": " + (format % args))
added
removed
bzrlib/lockdir.py
