1
# Copyright (C) 2006-2010 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
116
from bzrlib.decorators import only_raises
117
from bzrlib.errors import (
131
from bzrlib.trace import mutter, note
132
from bzrlib.osutils import format_delta, rand_chars, get_host_name
135
from bzrlib.lazy_import import lazy_import
136
lazy_import(globals(), """
137
from bzrlib import rio
140
# XXX: At the moment there is no consideration of thread safety on LockDir
141
# objects. This should perhaps be updated - e.g. if two threads try to take a
142
# lock at the same time they should *both* get it. But then that's unlikely
145
# TODO: Perhaps store some kind of note like the bzr command line in the lock
148
# TODO: Some kind of callback run while polling a lock to show progress
151
# TODO: Make sure to pass the right file and directory mode bits to all
152
# files/dirs created.
155
_DEFAULT_TIMEOUT_SECONDS = 30
156
_DEFAULT_POLL_SECONDS = 1.0
159
class LockDir(lock.Lock):
160
"""Write-lock guarding access to data.
163
__INFO_NAME = '/info'
165
def __init__(self, transport, path, file_modebits=0644, dir_modebits=0755):
166
"""Create a new LockDir object.
168
The LockDir is initially unlocked - this just creates the object.
170
:param transport: Transport which will contain the lock
172
:param path: Path to the lock within the base directory of the
175
self.transport = transport
177
self._lock_held = False
178
self._locked_via_token = False
179
self._fake_read_lock = False
180
self._held_dir = path + '/held'
181
self._held_info_path = self._held_dir + self.__INFO_NAME
182
self._file_modebits = file_modebits
183
self._dir_modebits = dir_modebits
185
self._report_function = note
188
return '%s(%s%s)' % (self.__class__.__name__,
192
is_held = property(lambda self: self._lock_held)
194
def create(self, mode=None):
195
"""Create the on-disk lock.
197
This is typically only called when the object/directory containing the
198
directory is first created. The lock is not held when it's created.
200
self._trace("create lock directory")
202
self.transport.mkdir(self.path, mode=mode)
203
except (TransportError, PathError), e:
204
raise LockFailed(self, e)
207
def _attempt_lock(self):
208
"""Make the pending directory and attempt to rename into place.
210
If the rename succeeds, we read back the info file to check that we
213
If we fail to acquire the lock, this method is responsible for
214
cleaning up the pending directory if possible. (But it doesn't do
217
:returns: The nonce of the lock, if it was successfully acquired.
219
:raises LockContention: If the lock is held by someone else. The exception
220
contains the info of the current holder of the lock.
222
self._trace("lock_write...")
223
start_time = time.time()
225
tmpname = self._create_pending_dir()
226
except (errors.TransportError, PathError), e:
227
self._trace("... failed to create pending dir, %s", e)
228
raise LockFailed(self, e)
230
self.transport.rename(tmpname, self._held_dir)
231
except (errors.TransportError, PathError, DirectoryNotEmpty,
232
FileExists, ResourceBusy), e:
233
self._trace("... contention, %s", e)
234
self._remove_pending_dir(tmpname)
235
raise LockContention(self)
237
self._trace("... lock failed, %s", e)
238
self._remove_pending_dir(tmpname)
240
# We must check we really got the lock, because Launchpad's sftp
241
# server at one time had a bug were the rename would successfully
242
# move the new directory into the existing directory, which was
243
# incorrect. It's possible some other servers or filesystems will
244
# have a similar bug allowing someone to think they got the lock
245
# when it's already held.
247
# See <https://bugs.launchpad.net/bzr/+bug/498378> for one case.
249
# Strictly the check is unnecessary and a waste of time for most
250
# people, but probably worth trapping if something is wrong.
252
self._trace("after locking, info=%r", info)
254
raise LockFailed(self, "lock was renamed into place, but "
256
if info.get('nonce') != self.nonce:
257
self._trace("rename succeeded, "
258
"but lock is still held by someone else")
259
raise LockContention(self)
260
self._lock_held = True
261
self._trace("... lock succeeded after %dms",
262
(time.time() - start_time) * 1000)
265
def _remove_pending_dir(self, tmpname):
266
"""Remove the pending directory
268
This is called if we failed to rename into place, so that the pending
269
dirs don't clutter up the lockdir.
271
self._trace("remove %s", tmpname)
273
self.transport.delete(tmpname + self.__INFO_NAME)
274
self.transport.rmdir(tmpname)
276
note("error removing pending lock: %s", e)
278
def _create_pending_dir(self):
279
tmpname = '%s/%s.tmp' % (self.path, rand_chars(10))
281
self.transport.mkdir(tmpname)
283
# This may raise a FileExists exception
284
# which is okay, it will be caught later and determined
285
# to be a LockContention.
286
self._trace("lock directory does not exist, creating it")
287
self.create(mode=self._dir_modebits)
288
# After creating the lock directory, try again
289
self.transport.mkdir(tmpname)
290
self.nonce = rand_chars(20)
291
info_bytes = self._prepare_info()
292
# We use put_file_non_atomic because we just created a new unique
293
# directory so we don't have to worry about files existing there.
294
# We'll rename the whole directory into place to get atomic
296
self.transport.put_bytes_non_atomic(tmpname + self.__INFO_NAME,
300
@only_raises(LockNotHeld, LockBroken)
302
"""Release a held lock
304
if self._fake_read_lock:
305
self._fake_read_lock = False
307
if not self._lock_held:
308
return lock.cant_unlock_not_held(self)
309
if self._locked_via_token:
310
self._locked_via_token = False
311
self._lock_held = False
313
old_nonce = self.nonce
314
# rename before deleting, because we can't atomically remove the
316
start_time = time.time()
317
self._trace("unlocking")
318
tmpname = '%s/releasing.%s.tmp' % (self.path, rand_chars(20))
319
# gotta own it to unlock
321
self.transport.rename(self._held_dir, tmpname)
322
self._lock_held = False
323
self.transport.delete(tmpname + self.__INFO_NAME)
325
self.transport.rmdir(tmpname)
326
except DirectoryNotEmpty, e:
327
# There might have been junk left over by a rename that moved
328
# another locker within the 'held' directory. do a slower
329
# deletion where we list the directory and remove everything
332
# Maybe this should be broader to allow for ftp servers with
333
# non-specific error messages?
334
self._trace("doing recursive deletion of non-empty directory "
336
self.transport.delete_tree(tmpname)
337
self._trace("... unlock succeeded after %dms",
338
(time.time() - start_time) * 1000)
339
result = lock.LockResult(self.transport.abspath(self.path),
341
for hook in self.hooks['lock_released']:
344
def break_lock(self):
345
"""Break a lock not held by this instance of LockDir.
347
This is a UI centric function: it uses the bzrlib.ui.ui_factory to
348
prompt for input if a lock is detected and there is any doubt about
349
it possibly being still active.
351
:returns: LockResult for the broken lock.
353
self._check_not_locked()
355
holder_info = self.peek()
356
except LockCorrupt, e:
357
# The lock info is corrupt.
358
if bzrlib.ui.ui_factory.get_boolean("Break (corrupt %r)" % (self,)):
359
self.force_break_corrupt(e.file_data)
361
if holder_info is not None:
362
lock_info = '\n'.join(self._format_lock_info(holder_info))
363
if bzrlib.ui.ui_factory.confirm_action(
364
"Break %(lock_info)s", 'bzrlib.lockdir.break',
365
dict(lock_info=lock_info)):
366
result = self.force_break(holder_info)
367
bzrlib.ui.ui_factory.show_message(
368
"Broke lock %s" % result.lock_url)
370
def force_break(self, dead_holder_info):
371
"""Release a lock held by another process.
373
WARNING: This should only be used when the other process is dead; if
374
it still thinks it has the lock there will be two concurrent writers.
375
In general the user's approval should be sought for lock breaks.
377
dead_holder_info must be the result of a previous LockDir.peek() call;
378
this is used to check that it's still held by the same process that
379
the user decided was dead. If this is not the current holder,
380
LockBreakMismatch is raised.
382
After the lock is broken it will not be held by any process.
383
It is possible that another process may sneak in and take the
384
lock before the breaking process acquires it.
386
:returns: LockResult for the broken lock.
388
if not isinstance(dead_holder_info, dict):
389
raise ValueError("dead_holder_info: %r" % dead_holder_info)
390
self._check_not_locked()
391
current_info = self.peek()
392
if current_info is None:
393
# must have been recently released
395
if current_info != dead_holder_info:
396
raise LockBreakMismatch(self, current_info, dead_holder_info)
397
tmpname = '%s/broken.%s.tmp' % (self.path, rand_chars(20))
398
self.transport.rename(self._held_dir, tmpname)
399
# check that we actually broke the right lock, not someone else;
400
# there's a small race window between checking it and doing the
402
broken_info_path = tmpname + self.__INFO_NAME
403
broken_info = self._read_info_file(broken_info_path)
404
if broken_info != dead_holder_info:
405
raise LockBreakMismatch(self, broken_info, dead_holder_info)
406
self.transport.delete(broken_info_path)
407
self.transport.rmdir(tmpname)
408
result = lock.LockResult(self.transport.abspath(self.path),
409
current_info.get('nonce'))
410
for hook in self.hooks['lock_broken']:
414
def force_break_corrupt(self, corrupt_info_lines):
415
"""Release a lock that has been corrupted.
417
This is very similar to force_break, it except it doesn't assume that
418
self.peek() can work.
420
:param corrupt_info_lines: the lines of the corrupted info file, used
421
to check that the lock hasn't changed between reading the (corrupt)
422
info file and calling force_break_corrupt.
424
# XXX: this copes with unparseable info files, but what about missing
425
# info files? Or missing lock dirs?
426
self._check_not_locked()
427
tmpname = '%s/broken.%s.tmp' % (self.path, rand_chars(20))
428
self.transport.rename(self._held_dir, tmpname)
429
# check that we actually broke the right lock, not someone else;
430
# there's a small race window between checking it and doing the
432
broken_info_path = tmpname + self.__INFO_NAME
433
f = self.transport.get(broken_info_path)
434
broken_lines = f.readlines()
435
if broken_lines != corrupt_info_lines:
436
raise LockBreakMismatch(self, broken_lines, corrupt_info_lines)
437
self.transport.delete(broken_info_path)
438
self.transport.rmdir(tmpname)
439
result = lock.LockResult(self.transport.abspath(self.path))
440
for hook in self.hooks['lock_broken']:
443
def _check_not_locked(self):
444
"""If the lock is held by this instance, raise an error."""
446
raise AssertionError("can't break own lock: %r" % self)
449
"""Make sure that the lock is still held by this locker.
451
This should only fail if the lock was broken by user intervention,
452
or if the lock has been affected by a bug.
454
If the lock is not thought to be held, raises LockNotHeld. If
455
the lock is thought to be held but has been broken, raises
458
if not self._lock_held:
459
raise LockNotHeld(self)
462
# no lock there anymore!
463
raise LockBroken(self)
464
if info.get('nonce') != self.nonce:
465
# there is a lock, but not ours
466
raise LockBroken(self)
468
def _read_info_file(self, path):
469
"""Read one given info file.
471
peek() reads the info file of the lock holder, if any.
473
return self._parse_info(self.transport.get_bytes(path))
476
"""Check if the lock is held by anyone.
478
If it is held, this returns the lock info structure as a dict
479
which contains some information about the current lock holder.
480
Otherwise returns None.
483
info = self._read_info_file(self._held_info_path)
484
self._trace("peek -> held")
486
except NoSuchFile, e:
487
self._trace("peek -> not held")
489
def _prepare_info(self):
490
"""Write information about a pending lock to a temporary file.
492
# XXX: is creating this here inefficient?
493
config = bzrlib.config.GlobalConfig()
495
user = config.username()
496
except errors.NoWhoami:
497
user = osutils.getuser_unicode()
498
s = rio.Stanza(hostname=get_host_name(),
499
pid=str(os.getpid()),
500
start_time=str(int(time.time())),
506
def _parse_info(self, info_bytes):
507
lines = osutils.split_lines(info_bytes)
509
stanza = rio.read_stanza(lines)
510
except ValueError, e:
511
mutter('Corrupt lock info file: %r', lines)
512
raise LockCorrupt("could not parse lock info file: " + str(e),
515
# see bug 185013; we fairly often end up with the info file being
516
# empty after an interruption; we could log a message here but
517
# there may not be much we can say
520
return stanza.as_dict()
522
def attempt_lock(self):
523
"""Take the lock; fail if it's already held.
525
If you wish to block until the lock can be obtained, call wait_lock()
528
:return: The lock token.
529
:raises LockContention: if the lock is held by someone else.
531
if self._fake_read_lock:
532
raise LockContention(self)
533
result = self._attempt_lock()
534
hook_result = lock.LockResult(self.transport.abspath(self.path),
536
for hook in self.hooks['lock_acquired']:
540
def wait_lock(self, timeout=None, poll=None, max_attempts=None):
541
"""Wait a certain period for a lock.
543
If the lock can be acquired within the bounded time, it
544
is taken and this returns. Otherwise, LockContention
545
is raised. Either way, this function should return within
546
approximately `timeout` seconds. (It may be a bit more if
547
a transport operation takes a long time to complete.)
549
:param timeout: Approximate maximum amount of time to wait for the
552
:param poll: Delay in seconds between retrying the lock.
554
:param max_attempts: Maximum number of times to try to lock.
556
:return: The lock token.
559
timeout = _DEFAULT_TIMEOUT_SECONDS
561
poll = _DEFAULT_POLL_SECONDS
562
# XXX: the transport interface doesn't let us guard against operations
563
# there taking a long time, so the total elapsed time or poll interval
564
# may be more than was requested.
565
deadline = time.time() + timeout
572
return self.attempt_lock()
573
except LockContention:
574
# possibly report the blockage, then try again
576
# TODO: In a few cases, we find out that there's contention by
577
# reading the held info and observing that it's not ours. In
578
# those cases it's a bit redundant to read it again. However,
579
# the normal case (??) is that the rename fails and so we
580
# don't know who holds the lock. For simplicity we peek
582
new_info = self.peek()
583
if new_info is not None and new_info != last_info:
584
if last_info is None:
585
start = 'Unable to obtain'
587
start = 'Lock owner changed for'
589
formatted_info = self._format_lock_info(new_info)
590
if deadline_str is None:
591
deadline_str = time.strftime('%H:%M:%S',
592
time.localtime(deadline))
593
# As local lock urls are correct we display them.
594
# We avoid displaying remote lock urls.
595
lock_url = self.transport.abspath(self.path)
596
if lock_url.startswith('file://'):
597
lock_url = lock_url.split('.bzr/')[0]
600
user, hostname, pid, time_ago = formatted_info
601
msg = ('%s lock %s ' # lock_url
605
'[process #%s], ' # pid
606
'acquired %s.') # time ago
607
msg_args = [start, lock_url, user, hostname, pid, time_ago]
609
msg += ('\nWill continue to try until %s, unless '
611
msg_args.append(deadline_str)
612
msg += '\nSee "bzr help break-lock" for more.'
613
self._report_function(msg, *msg_args)
614
if (max_attempts is not None) and (attempt_count >= max_attempts):
615
self._trace("exceeded %d attempts")
616
raise LockContention(self)
617
if time.time() + poll < deadline:
618
self._trace("waiting %ss", poll)
621
# As timeout is always 0 for remote locks
622
# this block is applicable only for local
624
self._trace("timeout after waiting %ss", timeout)
625
raise LockContention('(local)', lock_url)
627
def leave_in_place(self):
628
self._locked_via_token = True
630
def dont_leave_in_place(self):
631
self._locked_via_token = False
633
def lock_write(self, token=None):
634
"""Wait for and acquire the lock.
636
:param token: if this is already locked, then lock_write will fail
637
unless the token matches the existing lock.
638
:returns: a token if this instance supports tokens, otherwise None.
639
:raises TokenLockingNotSupported: when a token is given but this
640
instance doesn't support using token locks.
641
:raises MismatchedToken: if the specified token doesn't match the token
642
of the existing lock.
644
A token should be passed in if you know that you have locked the object
645
some other way, and need to synchronise this object's state with that
648
XXX: docstring duplicated from LockableFiles.lock_write.
650
if token is not None:
651
self.validate_token(token)
653
self._lock_held = True
654
self._locked_via_token = True
657
return self.wait_lock()
660
"""Compatibility-mode shared lock.
662
LockDir doesn't support shared read-only locks, so this
663
just pretends that the lock is taken but really does nothing.
665
# At the moment Branches are commonly locked for read, but
666
# we can't rely on that remotely. Once this is cleaned up,
667
# reenable this warning to prevent it coming back in
669
## warn("LockDir.lock_read falls back to write lock")
670
if self._lock_held or self._fake_read_lock:
671
raise LockContention(self)
672
self._fake_read_lock = True
674
def _format_lock_info(self, info):
675
"""Turn the contents of peek() into something for the user"""
676
start_time = info.get('start_time')
677
if start_time is None:
678
time_ago = '(unknown)'
680
time_ago = format_delta(time.time() - int(info['start_time']))
681
user = info.get('user', '<unknown>')
682
hostname = info.get('hostname', '<unknown>')
683
pid = info.get('pid', '<unknown>')
691
def validate_token(self, token):
692
if token is not None:
698
lock_token = info.get('nonce')
699
if token != lock_token:
700
raise errors.TokenMismatch(token, lock_token)
702
self._trace("revalidated by token %r", token)
704
def _trace(self, format, *args):
705
if 'lock' not in debug.debug_flags:
707
mutter(str(self) + ": " + (format % args))
added
removed
bzrlib/lockdir.py
