1
# Copyright (C) 2006, 2007 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 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
108
from cStringIO import StringIO
115
from bzrlib.errors import (
129
from bzrlib.trace import mutter, note
130
from bzrlib.transport import Transport
131
from bzrlib.osutils import rand_chars, format_delta
132
from bzrlib.rio import read_stanza, Stanza
136
# XXX: At the moment there is no consideration of thread safety on LockDir
137
# objects. This should perhaps be updated - e.g. if two threads try to take a
138
# lock at the same time they should *both* get it. But then that's unlikely
141
# TODO: Perhaps store some kind of note like the bzr command line in the lock
144
# TODO: Some kind of callback run while polling a lock to show progress
147
# TODO: Make sure to pass the right file and directory mode bits to all
148
# files/dirs created.
151
_DEFAULT_TIMEOUT_SECONDS = 300
152
_DEFAULT_POLL_SECONDS = 1.0
155
class LockDir(object):
156
"""Write-lock guarding access to data."""
158
__INFO_NAME = '/info'
160
def __init__(self, transport, path, file_modebits=0644, dir_modebits=0755):
161
"""Create a new LockDir object.
163
The LockDir is initially unlocked - this just creates the object.
165
:param transport: Transport which will contain the lock
167
:param path: Path to the lock within the base directory of the
170
self.transport = transport
172
self._lock_held = False
173
self._locked_via_token = False
174
self._fake_read_lock = False
175
self._held_dir = path + '/held'
176
self._held_info_path = self._held_dir + self.__INFO_NAME
177
self._file_modebits = file_modebits
178
self._dir_modebits = dir_modebits
180
self._report_function = note
183
return '%s(%s%s)' % (self.__class__.__name__,
187
is_held = property(lambda self: self._lock_held)
189
def create(self, mode=None):
190
"""Create the on-disk lock.
192
This is typically only called when the object/directory containing the
193
directory is first created. The lock is not held when it's created.
195
self._trace("create lock directory")
197
self.transport.mkdir(self.path, mode=mode)
198
except (TransportError, PathError), e:
199
raise LockFailed(self, e)
202
def _attempt_lock(self):
203
"""Make the pending directory and attempt to rename into place.
205
If the rename succeeds, we read back the info file to check that we
208
If we fail to acquire the lock, this method is responsible for
209
cleaning up the pending directory if possible. (But it doesn't do
212
:returns: The nonce of the lock, if it was successfully acquired.
214
:raises LockContention: If the lock is held by someone else. The exception
215
contains the info of the current holder of the lock.
217
self._trace("lock_write...")
218
start_time = time.time()
220
tmpname = self._create_pending_dir()
221
except (errors.TransportError, PathError), e:
222
self._trace("... failed to create pending dir, %s", e)
223
raise LockFailed(self, e)
225
self.transport.rename(tmpname, self._held_dir)
226
except (errors.TransportError, PathError, DirectoryNotEmpty,
227
FileExists, ResourceBusy), e:
228
self._trace("... contention, %s", e)
229
self._remove_pending_dir(tmpname)
230
raise LockContention(self)
232
self._trace("... lock failed, %s", e)
233
self._remove_pending_dir(tmpname)
235
# We must check we really got the lock, because Launchpad's sftp
236
# server at one time had a bug were the rename would successfully
237
# move the new directory into the existing directory, which was
238
# incorrect. It's possible some other servers or filesystems will
239
# have a similar bug allowing someone to think they got the lock
240
# when it's already held.
242
self._trace("after locking, info=%r", info)
243
if info['nonce'] != self.nonce:
244
self._trace("rename succeeded, "
245
"but lock is still held by someone else")
246
raise LockContention(self)
247
self._lock_held = True
248
self._trace("... lock succeeded after %dms",
249
(time.time() - start_time) * 1000)
252
def _remove_pending_dir(self, tmpname):
253
"""Remove the pending directory
255
This is called if we failed to rename into place, so that the pending
256
dirs don't clutter up the lockdir.
258
self._trace("remove %s", tmpname)
260
self.transport.delete(tmpname + self.__INFO_NAME)
261
self.transport.rmdir(tmpname)
263
note("error removing pending lock: %s", e)
265
def _create_pending_dir(self):
266
tmpname = '%s/%s.tmp' % (self.path, rand_chars(10))
268
self.transport.mkdir(tmpname)
270
# This may raise a FileExists exception
271
# which is okay, it will be caught later and determined
272
# to be a LockContention.
273
self._trace("lock directory does not exist, creating it")
274
self.create(mode=self._dir_modebits)
275
# After creating the lock directory, try again
276
self.transport.mkdir(tmpname)
277
self.nonce = rand_chars(20)
278
info_bytes = self._prepare_info()
279
# We use put_file_non_atomic because we just created a new unique
280
# directory so we don't have to worry about files existing there.
281
# We'll rename the whole directory into place to get atomic
283
self.transport.put_bytes_non_atomic(tmpname + self.__INFO_NAME,
288
"""Release a held lock
290
if self._fake_read_lock:
291
self._fake_read_lock = False
293
if not self._lock_held:
294
raise LockNotHeld(self)
295
if self._locked_via_token:
296
self._locked_via_token = False
297
self._lock_held = False
299
# rename before deleting, because we can't atomically remove the
301
start_time = time.time()
302
self._trace("unlocking")
303
tmpname = '%s/releasing.%s.tmp' % (self.path, rand_chars(20))
304
# gotta own it to unlock
306
self.transport.rename(self._held_dir, tmpname)
307
self._lock_held = False
308
self.transport.delete(tmpname + self.__INFO_NAME)
310
self.transport.rmdir(tmpname)
311
except DirectoryNotEmpty, e:
312
# There might have been junk left over by a rename that moved
313
# another locker within the 'held' directory. do a slower
314
# deletion where we list the directory and remove everything
317
# Maybe this should be broader to allow for ftp servers with
318
# non-specific error messages?
319
self._trace("doing recursive deletion of non-empty directory "
321
self.transport.delete_tree(tmpname)
322
self._trace("... unlock succeeded after %dms",
323
(time.time() - start_time) * 1000)
325
def break_lock(self):
326
"""Break a lock not held by this instance of LockDir.
328
This is a UI centric function: it uses the bzrlib.ui.ui_factory to
329
prompt for input if a lock is detected and there is any doubt about
330
it possibly being still active.
332
self._check_not_locked()
333
holder_info = self.peek()
334
if holder_info is not None:
335
lock_info = '\n'.join(self._format_lock_info(holder_info))
336
if bzrlib.ui.ui_factory.get_boolean("Break %s" % lock_info):
337
self.force_break(holder_info)
339
def force_break(self, dead_holder_info):
340
"""Release a lock held by another process.
342
WARNING: This should only be used when the other process is dead; if
343
it still thinks it has the lock there will be two concurrent writers.
344
In general the user's approval should be sought for lock breaks.
346
dead_holder_info must be the result of a previous LockDir.peek() call;
347
this is used to check that it's still held by the same process that
348
the user decided was dead. If this is not the current holder,
349
LockBreakMismatch is raised.
351
After the lock is broken it will not be held by any process.
352
It is possible that another process may sneak in and take the
353
lock before the breaking process acquires it.
355
if not isinstance(dead_holder_info, dict):
356
raise ValueError("dead_holder_info: %r" % dead_holder_info)
357
self._check_not_locked()
358
current_info = self.peek()
359
if current_info is None:
360
# must have been recently released
362
if current_info != dead_holder_info:
363
raise LockBreakMismatch(self, current_info, dead_holder_info)
364
tmpname = '%s/broken.%s.tmp' % (self.path, rand_chars(20))
365
self.transport.rename(self._held_dir, tmpname)
366
# check that we actually broke the right lock, not someone else;
367
# there's a small race window between checking it and doing the
369
broken_info_path = tmpname + self.__INFO_NAME
370
broken_info = self._read_info_file(broken_info_path)
371
if broken_info != dead_holder_info:
372
raise LockBreakMismatch(self, broken_info, dead_holder_info)
373
self.transport.delete(broken_info_path)
374
self.transport.rmdir(tmpname)
376
def _check_not_locked(self):
377
"""If the lock is held by this instance, raise an error."""
379
raise AssertionError("can't break own lock: %r" % self)
382
"""Make sure that the lock is still held by this locker.
384
This should only fail if the lock was broken by user intervention,
385
or if the lock has been affected by a bug.
387
If the lock is not thought to be held, raises LockNotHeld. If
388
the lock is thought to be held but has been broken, raises
391
if not self._lock_held:
392
raise LockNotHeld(self)
395
# no lock there anymore!
396
raise LockBroken(self)
397
if info.get('nonce') != self.nonce:
398
# there is a lock, but not ours
399
raise LockBroken(self)
401
def _read_info_file(self, path):
402
"""Read one given info file.
404
peek() reads the info file of the lock holder, if any.
406
return self._parse_info(self.transport.get(path))
409
"""Check if the lock is held by anyone.
411
If it is held, this returns the lock info structure as a rio Stanza,
412
which contains some information about the current lock holder.
413
Otherwise returns None.
416
info = self._read_info_file(self._held_info_path)
417
self._trace("peek -> held")
419
except NoSuchFile, e:
420
self._trace("peek -> not held")
422
def _prepare_info(self):
423
"""Write information about a pending lock to a temporary file.
426
# XXX: is creating this here inefficient?
427
config = bzrlib.config.GlobalConfig()
429
user = config.user_email()
430
except errors.NoEmailInUsername:
431
user = config.username()
432
s = Stanza(hostname=socket.gethostname(),
433
pid=str(os.getpid()),
434
start_time=str(int(time.time())),
440
def _parse_info(self, info_file):
441
return read_stanza(info_file.readlines()).as_dict()
443
def attempt_lock(self):
444
"""Take the lock; fail if it's already held.
446
If you wish to block until the lock can be obtained, call wait_lock()
449
:return: The lock token.
450
:raises LockContention: if the lock is held by someone else.
452
if self._fake_read_lock:
453
raise LockContention(self)
454
return self._attempt_lock()
456
def wait_lock(self, timeout=None, poll=None, max_attempts=None):
457
"""Wait a certain period for a lock.
459
If the lock can be acquired within the bounded time, it
460
is taken and this returns. Otherwise, LockContention
461
is raised. Either way, this function should return within
462
approximately `timeout` seconds. (It may be a bit more if
463
a transport operation takes a long time to complete.)
465
:param timeout: Approximate maximum amount of time to wait for the
468
:param poll: Delay in seconds between retrying the lock.
470
:param max_attempts: Maximum number of times to try to lock.
472
:return: The lock token.
475
timeout = _DEFAULT_TIMEOUT_SECONDS
477
poll = _DEFAULT_POLL_SECONDS
478
# XXX: the transport interface doesn't let us guard against operations
479
# there taking a long time, so the total elapsed time or poll interval
480
# may be more than was requested.
481
deadline = time.time() + timeout
488
return self.attempt_lock()
489
except LockContention:
490
# possibly report the blockage, then try again
492
# TODO: In a few cases, we find out that there's contention by
493
# reading the held info and observing that it's not ours. In
494
# those cases it's a bit redundant to read it again. However,
495
# the normal case (??) is that the rename fails and so we
496
# don't know who holds the lock. For simplicity we peek
498
new_info = self.peek()
499
if new_info is not None and new_info != last_info:
500
if last_info is None:
501
start = 'Unable to obtain'
503
start = 'Lock owner changed for'
505
formatted_info = self._format_lock_info(new_info)
506
if deadline_str is None:
507
deadline_str = time.strftime('%H:%M:%S',
508
time.localtime(deadline))
509
lock_url = self.transport.abspath(self.path)
510
self._report_function('%s %s\n'
512
'%s\n' # locked ... ago
513
'Will continue to try until %s, unless '
515
'If you\'re sure that it\'s not being '
516
'modified, use bzr break-lock %s',
524
if (max_attempts is not None) and (attempt_count >= max_attempts):
525
self._trace("exceeded %d attempts")
526
raise LockContention(self)
527
if time.time() + poll < deadline:
528
self._trace("waiting %ss", poll)
531
self._trace("timeout after waiting %ss", timeout)
532
raise LockContention(self)
534
def leave_in_place(self):
535
self._locked_via_token = True
537
def dont_leave_in_place(self):
538
self._locked_via_token = False
540
def lock_write(self, token=None):
541
"""Wait for and acquire the lock.
543
:param token: if this is already locked, then lock_write will fail
544
unless the token matches the existing lock.
545
:returns: a token if this instance supports tokens, otherwise None.
546
:raises TokenLockingNotSupported: when a token is given but this
547
instance doesn't support using token locks.
548
:raises MismatchedToken: if the specified token doesn't match the token
549
of the existing lock.
551
A token should be passed in if you know that you have locked the object
552
some other way, and need to synchronise this object's state with that
555
XXX: docstring duplicated from LockableFiles.lock_write.
557
if token is not None:
558
self.validate_token(token)
560
self._lock_held = True
561
self._locked_via_token = True
564
return self.wait_lock()
567
"""Compatibility-mode shared lock.
569
LockDir doesn't support shared read-only locks, so this
570
just pretends that the lock is taken but really does nothing.
572
# At the moment Branches are commonly locked for read, but
573
# we can't rely on that remotely. Once this is cleaned up,
574
# reenable this warning to prevent it coming back in
576
## warn("LockDir.lock_read falls back to write lock")
577
if self._lock_held or self._fake_read_lock:
578
raise LockContention(self)
579
self._fake_read_lock = True
581
def _format_lock_info(self, info):
582
"""Turn the contents of peek() into something for the user"""
583
lock_url = self.transport.abspath(self.path)
584
delta = time.time() - int(info['start_time'])
586
'lock %s' % (lock_url,),
587
'held by %(user)s on host %(hostname)s [process #%(pid)s]' % info,
588
'locked %s' % (format_delta(delta),),
591
def validate_token(self, token):
592
if token is not None:
598
lock_token = info.get('nonce')
599
if token != lock_token:
600
raise errors.TokenMismatch(token, lock_token)
602
self._trace("revalidated by token %r", token)
604
def _trace(self, format, *args):
605
if 'lock' not in debug.debug_flags:
607
mutter(str(self) + ": " + (format % args))
added
removed
bzrlib/lockdir.py
