1
# Copyright (C) 2006, 2007, 2008, 2009 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
116
from bzrlib.decorators import only_raises
117
from bzrlib.errors import (
130
from bzrlib.trace import mutter, note
131
from bzrlib.osutils import format_delta, rand_chars, get_host_name
134
from bzrlib.lazy_import import lazy_import
135
lazy_import(globals(), """
136
from bzrlib import rio
139
# XXX: At the moment there is no consideration of thread safety on LockDir
140
# objects. This should perhaps be updated - e.g. if two threads try to take a
141
# lock at the same time they should *both* get it. But then that's unlikely
144
# TODO: Perhaps store some kind of note like the bzr command line in the lock
147
# TODO: Some kind of callback run while polling a lock to show progress
150
# TODO: Make sure to pass the right file and directory mode bits to all
151
# files/dirs created.
154
_DEFAULT_TIMEOUT_SECONDS = 300
155
_DEFAULT_POLL_SECONDS = 1.0
158
class LockDir(lock.Lock):
159
"""Write-lock guarding access to data.
162
__INFO_NAME = '/info'
164
def __init__(self, transport, path, file_modebits=0644, dir_modebits=0755):
165
"""Create a new LockDir object.
167
The LockDir is initially unlocked - this just creates the object.
169
:param transport: Transport which will contain the lock
171
:param path: Path to the lock within the base directory of the
174
self.transport = transport
176
self._lock_held = False
177
self._locked_via_token = False
178
self._fake_read_lock = False
179
self._held_dir = path + '/held'
180
self._held_info_path = self._held_dir + self.__INFO_NAME
181
self._file_modebits = file_modebits
182
self._dir_modebits = dir_modebits
184
self._report_function = note
187
return '%s(%s%s)' % (self.__class__.__name__,
191
is_held = property(lambda self: self._lock_held)
193
def create(self, mode=None):
194
"""Create the on-disk lock.
196
This is typically only called when the object/directory containing the
197
directory is first created. The lock is not held when it's created.
199
self._trace("create lock directory")
201
self.transport.mkdir(self.path, mode=mode)
202
except (TransportError, PathError), e:
203
raise LockFailed(self, e)
206
def _attempt_lock(self):
207
"""Make the pending directory and attempt to rename into place.
209
If the rename succeeds, we read back the info file to check that we
212
If we fail to acquire the lock, this method is responsible for
213
cleaning up the pending directory if possible. (But it doesn't do
216
:returns: The nonce of the lock, if it was successfully acquired.
218
:raises LockContention: If the lock is held by someone else. The exception
219
contains the info of the current holder of the lock.
221
self._trace("lock_write...")
222
start_time = time.time()
224
tmpname = self._create_pending_dir()
225
except (errors.TransportError, PathError), e:
226
self._trace("... failed to create pending dir, %s", e)
227
raise LockFailed(self, e)
229
self.transport.rename(tmpname, self._held_dir)
230
except (errors.TransportError, PathError, DirectoryNotEmpty,
231
FileExists, ResourceBusy), e:
232
self._trace("... contention, %s", e)
233
self._remove_pending_dir(tmpname)
234
raise LockContention(self)
236
self._trace("... lock failed, %s", e)
237
self._remove_pending_dir(tmpname)
239
# We must check we really got the lock, because Launchpad's sftp
240
# server at one time had a bug were the rename would successfully
241
# move the new directory into the existing directory, which was
242
# incorrect. It's possible some other servers or filesystems will
243
# have a similar bug allowing someone to think they got the lock
244
# when it's already held.
246
self._trace("after locking, info=%r", info)
247
if info['nonce'] != self.nonce:
248
self._trace("rename succeeded, "
249
"but lock is still held by someone else")
250
raise LockContention(self)
251
self._lock_held = True
252
self._trace("... lock succeeded after %dms",
253
(time.time() - start_time) * 1000)
256
def _remove_pending_dir(self, tmpname):
257
"""Remove the pending directory
259
This is called if we failed to rename into place, so that the pending
260
dirs don't clutter up the lockdir.
262
self._trace("remove %s", tmpname)
264
self.transport.delete(tmpname + self.__INFO_NAME)
265
self.transport.rmdir(tmpname)
267
note("error removing pending lock: %s", e)
269
def _create_pending_dir(self):
270
tmpname = '%s/%s.tmp' % (self.path, rand_chars(10))
272
self.transport.mkdir(tmpname)
274
# This may raise a FileExists exception
275
# which is okay, it will be caught later and determined
276
# to be a LockContention.
277
self._trace("lock directory does not exist, creating it")
278
self.create(mode=self._dir_modebits)
279
# After creating the lock directory, try again
280
self.transport.mkdir(tmpname)
281
self.nonce = rand_chars(20)
282
info_bytes = self._prepare_info()
283
# We use put_file_non_atomic because we just created a new unique
284
# directory so we don't have to worry about files existing there.
285
# We'll rename the whole directory into place to get atomic
287
self.transport.put_bytes_non_atomic(tmpname + self.__INFO_NAME,
291
@only_raises(LockNotHeld, LockBroken)
293
"""Release a held lock
295
if self._fake_read_lock:
296
self._fake_read_lock = False
298
if not self._lock_held:
299
return lock.cant_unlock_not_held(self)
300
if self._locked_via_token:
301
self._locked_via_token = False
302
self._lock_held = False
304
old_nonce = self.nonce
305
# rename before deleting, because we can't atomically remove the
307
start_time = time.time()
308
self._trace("unlocking")
309
tmpname = '%s/releasing.%s.tmp' % (self.path, rand_chars(20))
310
# gotta own it to unlock
312
self.transport.rename(self._held_dir, tmpname)
313
self._lock_held = False
314
self.transport.delete(tmpname + self.__INFO_NAME)
316
self.transport.rmdir(tmpname)
317
except DirectoryNotEmpty, e:
318
# There might have been junk left over by a rename that moved
319
# another locker within the 'held' directory. do a slower
320
# deletion where we list the directory and remove everything
323
# Maybe this should be broader to allow for ftp servers with
324
# non-specific error messages?
325
self._trace("doing recursive deletion of non-empty directory "
327
self.transport.delete_tree(tmpname)
328
self._trace("... unlock succeeded after %dms",
329
(time.time() - start_time) * 1000)
330
result = lock.LockResult(self.transport.abspath(self.path),
332
for hook in self.hooks['lock_released']:
335
def break_lock(self):
336
"""Break a lock not held by this instance of LockDir.
338
This is a UI centric function: it uses the bzrlib.ui.ui_factory to
339
prompt for input if a lock is detected and there is any doubt about
340
it possibly being still active.
342
self._check_not_locked()
343
holder_info = self.peek()
344
if holder_info is not None:
345
lock_info = '\n'.join(self._format_lock_info(holder_info))
346
if bzrlib.ui.ui_factory.get_boolean("Break %s" % lock_info):
347
self.force_break(holder_info)
349
def force_break(self, dead_holder_info):
350
"""Release a lock held by another process.
352
WARNING: This should only be used when the other process is dead; if
353
it still thinks it has the lock there will be two concurrent writers.
354
In general the user's approval should be sought for lock breaks.
356
dead_holder_info must be the result of a previous LockDir.peek() call;
357
this is used to check that it's still held by the same process that
358
the user decided was dead. If this is not the current holder,
359
LockBreakMismatch is raised.
361
After the lock is broken it will not be held by any process.
362
It is possible that another process may sneak in and take the
363
lock before the breaking process acquires it.
365
if not isinstance(dead_holder_info, dict):
366
raise ValueError("dead_holder_info: %r" % dead_holder_info)
367
self._check_not_locked()
368
current_info = self.peek()
369
if current_info is None:
370
# must have been recently released
372
if current_info != dead_holder_info:
373
raise LockBreakMismatch(self, current_info, dead_holder_info)
374
tmpname = '%s/broken.%s.tmp' % (self.path, rand_chars(20))
375
self.transport.rename(self._held_dir, tmpname)
376
# check that we actually broke the right lock, not someone else;
377
# there's a small race window between checking it and doing the
379
broken_info_path = tmpname + self.__INFO_NAME
380
broken_info = self._read_info_file(broken_info_path)
381
if broken_info != dead_holder_info:
382
raise LockBreakMismatch(self, broken_info, dead_holder_info)
383
self.transport.delete(broken_info_path)
384
self.transport.rmdir(tmpname)
385
result = lock.LockResult(self.transport.abspath(self.path),
386
current_info.get('nonce'))
387
for hook in self.hooks['lock_broken']:
390
def _check_not_locked(self):
391
"""If the lock is held by this instance, raise an error."""
393
raise AssertionError("can't break own lock: %r" % self)
396
"""Make sure that the lock is still held by this locker.
398
This should only fail if the lock was broken by user intervention,
399
or if the lock has been affected by a bug.
401
If the lock is not thought to be held, raises LockNotHeld. If
402
the lock is thought to be held but has been broken, raises
405
if not self._lock_held:
406
raise LockNotHeld(self)
409
# no lock there anymore!
410
raise LockBroken(self)
411
if info.get('nonce') != self.nonce:
412
# there is a lock, but not ours
413
raise LockBroken(self)
415
def _read_info_file(self, path):
416
"""Read one given info file.
418
peek() reads the info file of the lock holder, if any.
420
return self._parse_info(self.transport.get_bytes(path))
423
"""Check if the lock is held by anyone.
425
If it is held, this returns the lock info structure as a rio Stanza,
426
which contains some information about the current lock holder.
427
Otherwise returns None.
430
info = self._read_info_file(self._held_info_path)
431
self._trace("peek -> held")
433
except NoSuchFile, e:
434
self._trace("peek -> not held")
436
def _prepare_info(self):
437
"""Write information about a pending lock to a temporary file.
439
# XXX: is creating this here inefficient?
440
config = bzrlib.config.GlobalConfig()
442
user = config.user_email()
443
except errors.NoEmailInUsername:
444
user = config.username()
445
s = rio.Stanza(hostname=get_host_name(),
446
pid=str(os.getpid()),
447
start_time=str(int(time.time())),
453
def _parse_info(self, info_bytes):
454
# TODO: Handle if info_bytes is empty
455
return rio.read_stanza(osutils.split_lines(info_bytes)).as_dict()
457
def attempt_lock(self):
458
"""Take the lock; fail if it's already held.
460
If you wish to block until the lock can be obtained, call wait_lock()
463
:return: The lock token.
464
:raises LockContention: if the lock is held by someone else.
466
if self._fake_read_lock:
467
raise LockContention(self)
468
result = self._attempt_lock()
469
hook_result = lock.LockResult(self.transport.abspath(self.path),
471
for hook in self.hooks['lock_acquired']:
475
def wait_lock(self, timeout=None, poll=None, max_attempts=None):
476
"""Wait a certain period for a lock.
478
If the lock can be acquired within the bounded time, it
479
is taken and this returns. Otherwise, LockContention
480
is raised. Either way, this function should return within
481
approximately `timeout` seconds. (It may be a bit more if
482
a transport operation takes a long time to complete.)
484
:param timeout: Approximate maximum amount of time to wait for the
487
:param poll: Delay in seconds between retrying the lock.
489
:param max_attempts: Maximum number of times to try to lock.
491
:return: The lock token.
494
timeout = _DEFAULT_TIMEOUT_SECONDS
496
poll = _DEFAULT_POLL_SECONDS
497
# XXX: the transport interface doesn't let us guard against operations
498
# there taking a long time, so the total elapsed time or poll interval
499
# may be more than was requested.
500
deadline = time.time() + timeout
507
return self.attempt_lock()
508
except LockContention:
509
# possibly report the blockage, then try again
511
# TODO: In a few cases, we find out that there's contention by
512
# reading the held info and observing that it's not ours. In
513
# those cases it's a bit redundant to read it again. However,
514
# the normal case (??) is that the rename fails and so we
515
# don't know who holds the lock. For simplicity we peek
517
new_info = self.peek()
518
if new_info is not None and new_info != last_info:
519
if last_info is None:
520
start = 'Unable to obtain'
522
start = 'Lock owner changed for'
524
formatted_info = self._format_lock_info(new_info)
525
if deadline_str is None:
526
deadline_str = time.strftime('%H:%M:%S',
527
time.localtime(deadline))
528
lock_url = self.transport.abspath(self.path)
529
self._report_function('%s %s\n'
531
'%s\n' # locked ... ago
532
'Will continue to try until %s, unless '
534
'If you\'re sure that it\'s not being '
535
'modified, use bzr break-lock %s',
543
if (max_attempts is not None) and (attempt_count >= max_attempts):
544
self._trace("exceeded %d attempts")
545
raise LockContention(self)
546
if time.time() + poll < deadline:
547
self._trace("waiting %ss", poll)
550
self._trace("timeout after waiting %ss", timeout)
551
raise LockContention(self)
553
def leave_in_place(self):
554
self._locked_via_token = True
556
def dont_leave_in_place(self):
557
self._locked_via_token = False
559
def lock_write(self, token=None):
560
"""Wait for and acquire the lock.
562
:param token: if this is already locked, then lock_write will fail
563
unless the token matches the existing lock.
564
:returns: a token if this instance supports tokens, otherwise None.
565
:raises TokenLockingNotSupported: when a token is given but this
566
instance doesn't support using token locks.
567
:raises MismatchedToken: if the specified token doesn't match the token
568
of the existing lock.
570
A token should be passed in if you know that you have locked the object
571
some other way, and need to synchronise this object's state with that
574
XXX: docstring duplicated from LockableFiles.lock_write.
576
if token is not None:
577
self.validate_token(token)
579
self._lock_held = True
580
self._locked_via_token = True
583
return self.wait_lock()
586
"""Compatibility-mode shared lock.
588
LockDir doesn't support shared read-only locks, so this
589
just pretends that the lock is taken but really does nothing.
591
# At the moment Branches are commonly locked for read, but
592
# we can't rely on that remotely. Once this is cleaned up,
593
# reenable this warning to prevent it coming back in
595
## warn("LockDir.lock_read falls back to write lock")
596
if self._lock_held or self._fake_read_lock:
597
raise LockContention(self)
598
self._fake_read_lock = True
600
def _format_lock_info(self, info):
601
"""Turn the contents of peek() into something for the user"""
602
lock_url = self.transport.abspath(self.path)
603
delta = time.time() - int(info['start_time'])
605
'lock %s' % (lock_url,),
606
'held by %(user)s on host %(hostname)s [process #%(pid)s]' % info,
607
'locked %s' % (format_delta(delta),),
610
def validate_token(self, token):
611
if token is not None:
617
lock_token = info.get('nonce')
618
if token != lock_token:
619
raise errors.TokenMismatch(token, lock_token)
621
self._trace("revalidated by token %r", token)
623
def _trace(self, format, *args):
624
if 'lock' not in debug.debug_flags:
626
mutter(str(self) + ": " + (format % args))
added
removed
bzrlib/lockdir.py
