1
# Copyright (C) 2006, 2007, 2008 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.decorators import only_raises
116
from bzrlib.errors import (
129
from bzrlib.trace import mutter, note
130
from bzrlib.osutils import format_delta, rand_chars, get_host_name
133
from bzrlib.lazy_import import lazy_import
134
lazy_import(globals(), """
135
from bzrlib import rio
138
# XXX: At the moment there is no consideration of thread safety on LockDir
139
# objects. This should perhaps be updated - e.g. if two threads try to take a
140
# lock at the same time they should *both* get it. But then that's unlikely
143
# TODO: Perhaps store some kind of note like the bzr command line in the lock
146
# TODO: Some kind of callback run while polling a lock to show progress
149
# TODO: Make sure to pass the right file and directory mode bits to all
150
# files/dirs created.
153
_DEFAULT_TIMEOUT_SECONDS = 300
154
_DEFAULT_POLL_SECONDS = 1.0
157
class LockDir(lock.Lock):
158
"""Write-lock guarding access to data.
161
__INFO_NAME = '/info'
163
def __init__(self, transport, path, file_modebits=0644, dir_modebits=0755):
164
"""Create a new LockDir object.
166
The LockDir is initially unlocked - this just creates the object.
168
:param transport: Transport which will contain the lock
170
:param path: Path to the lock within the base directory of the
173
self.transport = transport
175
self._lock_held = False
176
self._locked_via_token = False
177
self._fake_read_lock = False
178
self._held_dir = path + '/held'
179
self._held_info_path = self._held_dir + self.__INFO_NAME
180
self._file_modebits = file_modebits
181
self._dir_modebits = dir_modebits
183
self._report_function = note
186
return '%s(%s%s)' % (self.__class__.__name__,
190
is_held = property(lambda self: self._lock_held)
192
def create(self, mode=None):
193
"""Create the on-disk lock.
195
This is typically only called when the object/directory containing the
196
directory is first created. The lock is not held when it's created.
198
self._trace("create lock directory")
200
self.transport.mkdir(self.path, mode=mode)
201
except (TransportError, PathError), e:
202
raise LockFailed(self, e)
205
def _attempt_lock(self):
206
"""Make the pending directory and attempt to rename into place.
208
If the rename succeeds, we read back the info file to check that we
211
If we fail to acquire the lock, this method is responsible for
212
cleaning up the pending directory if possible. (But it doesn't do
215
:returns: The nonce of the lock, if it was successfully acquired.
217
:raises LockContention: If the lock is held by someone else. The exception
218
contains the info of the current holder of the lock.
220
self._trace("lock_write...")
221
start_time = time.time()
223
tmpname = self._create_pending_dir()
224
except (errors.TransportError, PathError), e:
225
self._trace("... failed to create pending dir, %s", e)
226
raise LockFailed(self, e)
228
self.transport.rename(tmpname, self._held_dir)
229
except (errors.TransportError, PathError, DirectoryNotEmpty,
230
FileExists, ResourceBusy), e:
231
self._trace("... contention, %s", e)
232
self._remove_pending_dir(tmpname)
233
raise LockContention(self)
235
self._trace("... lock failed, %s", e)
236
self._remove_pending_dir(tmpname)
238
# We must check we really got the lock, because Launchpad's sftp
239
# server at one time had a bug were the rename would successfully
240
# move the new directory into the existing directory, which was
241
# incorrect. It's possible some other servers or filesystems will
242
# have a similar bug allowing someone to think they got the lock
243
# when it's already held.
245
self._trace("after locking, info=%r", info)
246
if info['nonce'] != self.nonce:
247
self._trace("rename succeeded, "
248
"but lock is still held by someone else")
249
raise LockContention(self)
250
self._lock_held = True
251
self._trace("... lock succeeded after %dms",
252
(time.time() - start_time) * 1000)
255
def _remove_pending_dir(self, tmpname):
256
"""Remove the pending directory
258
This is called if we failed to rename into place, so that the pending
259
dirs don't clutter up the lockdir.
261
self._trace("remove %s", tmpname)
263
self.transport.delete(tmpname + self.__INFO_NAME)
264
self.transport.rmdir(tmpname)
266
note("error removing pending lock: %s", e)
268
def _create_pending_dir(self):
269
tmpname = '%s/%s.tmp' % (self.path, rand_chars(10))
271
self.transport.mkdir(tmpname)
273
# This may raise a FileExists exception
274
# which is okay, it will be caught later and determined
275
# to be a LockContention.
276
self._trace("lock directory does not exist, creating it")
277
self.create(mode=self._dir_modebits)
278
# After creating the lock directory, try again
279
self.transport.mkdir(tmpname)
280
self.nonce = rand_chars(20)
281
info_bytes = self._prepare_info()
282
# We use put_file_non_atomic because we just created a new unique
283
# directory so we don't have to worry about files existing there.
284
# We'll rename the whole directory into place to get atomic
286
self.transport.put_bytes_non_atomic(tmpname + self.__INFO_NAME,
290
@only_raises(LockNotHeld, LockBroken)
292
"""Release a held lock
294
if self._fake_read_lock:
295
self._fake_read_lock = False
297
if not self._lock_held:
298
return lock.cant_unlock_not_held(self)
299
if self._locked_via_token:
300
self._locked_via_token = False
301
self._lock_held = False
303
old_nonce = self.nonce
304
# rename before deleting, because we can't atomically remove the
306
start_time = time.time()
307
self._trace("unlocking")
308
tmpname = '%s/releasing.%s.tmp' % (self.path, rand_chars(20))
309
# gotta own it to unlock
311
self.transport.rename(self._held_dir, tmpname)
312
self._lock_held = False
313
self.transport.delete(tmpname + self.__INFO_NAME)
315
self.transport.rmdir(tmpname)
316
except DirectoryNotEmpty, e:
317
# There might have been junk left over by a rename that moved
318
# another locker within the 'held' directory. do a slower
319
# deletion where we list the directory and remove everything
322
# Maybe this should be broader to allow for ftp servers with
323
# non-specific error messages?
324
self._trace("doing recursive deletion of non-empty directory "
326
self.transport.delete_tree(tmpname)
327
self._trace("... unlock succeeded after %dms",
328
(time.time() - start_time) * 1000)
329
result = lock.LockResult(self.transport.abspath(self.path),
331
for hook in self.hooks['lock_released']:
334
def break_lock(self):
335
"""Break a lock not held by this instance of LockDir.
337
This is a UI centric function: it uses the bzrlib.ui.ui_factory to
338
prompt for input if a lock is detected and there is any doubt about
339
it possibly being still active.
341
self._check_not_locked()
342
holder_info = self.peek()
343
if holder_info is not None:
344
lock_info = '\n'.join(self._format_lock_info(holder_info))
345
if bzrlib.ui.ui_factory.get_boolean("Break %s" % lock_info):
346
self.force_break(holder_info)
348
def force_break(self, dead_holder_info):
349
"""Release a lock held by another process.
351
WARNING: This should only be used when the other process is dead; if
352
it still thinks it has the lock there will be two concurrent writers.
353
In general the user's approval should be sought for lock breaks.
355
dead_holder_info must be the result of a previous LockDir.peek() call;
356
this is used to check that it's still held by the same process that
357
the user decided was dead. If this is not the current holder,
358
LockBreakMismatch is raised.
360
After the lock is broken it will not be held by any process.
361
It is possible that another process may sneak in and take the
362
lock before the breaking process acquires it.
364
if not isinstance(dead_holder_info, dict):
365
raise ValueError("dead_holder_info: %r" % dead_holder_info)
366
self._check_not_locked()
367
current_info = self.peek()
368
if current_info is None:
369
# must have been recently released
371
if current_info != dead_holder_info:
372
raise LockBreakMismatch(self, current_info, dead_holder_info)
373
tmpname = '%s/broken.%s.tmp' % (self.path, rand_chars(20))
374
self.transport.rename(self._held_dir, tmpname)
375
# check that we actually broke the right lock, not someone else;
376
# there's a small race window between checking it and doing the
378
broken_info_path = tmpname + self.__INFO_NAME
379
broken_info = self._read_info_file(broken_info_path)
380
if broken_info != dead_holder_info:
381
raise LockBreakMismatch(self, broken_info, dead_holder_info)
382
self.transport.delete(broken_info_path)
383
self.transport.rmdir(tmpname)
384
result = lock.LockResult(self.transport.abspath(self.path),
385
current_info.get('nonce'))
386
for hook in self.hooks['lock_broken']:
389
def _check_not_locked(self):
390
"""If the lock is held by this instance, raise an error."""
392
raise AssertionError("can't break own lock: %r" % self)
395
"""Make sure that the lock is still held by this locker.
397
This should only fail if the lock was broken by user intervention,
398
or if the lock has been affected by a bug.
400
If the lock is not thought to be held, raises LockNotHeld. If
401
the lock is thought to be held but has been broken, raises
404
if not self._lock_held:
405
raise LockNotHeld(self)
408
# no lock there anymore!
409
raise LockBroken(self)
410
if info.get('nonce') != self.nonce:
411
# there is a lock, but not ours
412
raise LockBroken(self)
414
def _read_info_file(self, path):
415
"""Read one given info file.
417
peek() reads the info file of the lock holder, if any.
419
return self._parse_info(self.transport.get(path))
422
"""Check if the lock is held by anyone.
424
If it is held, this returns the lock info structure as a rio Stanza,
425
which contains some information about the current lock holder.
426
Otherwise returns None.
429
info = self._read_info_file(self._held_info_path)
430
self._trace("peek -> held")
432
except NoSuchFile, e:
433
self._trace("peek -> not held")
435
def _prepare_info(self):
436
"""Write information about a pending lock to a temporary file.
438
# XXX: is creating this here inefficient?
439
config = bzrlib.config.GlobalConfig()
441
user = config.user_email()
442
except errors.NoEmailInUsername:
443
user = config.username()
444
s = rio.Stanza(hostname=get_host_name(),
445
pid=str(os.getpid()),
446
start_time=str(int(time.time())),
452
def _parse_info(self, info_file):
453
return rio.read_stanza(info_file.readlines()).as_dict()
455
def attempt_lock(self):
456
"""Take the lock; fail if it's already held.
458
If you wish to block until the lock can be obtained, call wait_lock()
461
:return: The lock token.
462
:raises LockContention: if the lock is held by someone else.
464
if self._fake_read_lock:
465
raise LockContention(self)
466
result = self._attempt_lock()
467
hook_result = lock.LockResult(self.transport.abspath(self.path),
469
for hook in self.hooks['lock_acquired']:
473
def wait_lock(self, timeout=None, poll=None, max_attempts=None):
474
"""Wait a certain period for a lock.
476
If the lock can be acquired within the bounded time, it
477
is taken and this returns. Otherwise, LockContention
478
is raised. Either way, this function should return within
479
approximately `timeout` seconds. (It may be a bit more if
480
a transport operation takes a long time to complete.)
482
:param timeout: Approximate maximum amount of time to wait for the
485
:param poll: Delay in seconds between retrying the lock.
487
:param max_attempts: Maximum number of times to try to lock.
489
:return: The lock token.
492
timeout = _DEFAULT_TIMEOUT_SECONDS
494
poll = _DEFAULT_POLL_SECONDS
495
# XXX: the transport interface doesn't let us guard against operations
496
# there taking a long time, so the total elapsed time or poll interval
497
# may be more than was requested.
498
deadline = time.time() + timeout
505
return self.attempt_lock()
506
except LockContention:
507
# possibly report the blockage, then try again
509
# TODO: In a few cases, we find out that there's contention by
510
# reading the held info and observing that it's not ours. In
511
# those cases it's a bit redundant to read it again. However,
512
# the normal case (??) is that the rename fails and so we
513
# don't know who holds the lock. For simplicity we peek
515
new_info = self.peek()
516
if new_info is not None and new_info != last_info:
517
if last_info is None:
518
start = 'Unable to obtain'
520
start = 'Lock owner changed for'
522
formatted_info = self._format_lock_info(new_info)
523
if deadline_str is None:
524
deadline_str = time.strftime('%H:%M:%S',
525
time.localtime(deadline))
526
lock_url = self.transport.abspath(self.path)
527
self._report_function('%s %s\n'
529
'%s\n' # locked ... ago
530
'Will continue to try until %s, unless '
532
'If you\'re sure that it\'s not being '
533
'modified, use bzr break-lock %s',
541
if (max_attempts is not None) and (attempt_count >= max_attempts):
542
self._trace("exceeded %d attempts")
543
raise LockContention(self)
544
if time.time() + poll < deadline:
545
self._trace("waiting %ss", poll)
548
self._trace("timeout after waiting %ss", timeout)
549
raise LockContention(self)
551
def leave_in_place(self):
552
self._locked_via_token = True
554
def dont_leave_in_place(self):
555
self._locked_via_token = False
557
def lock_write(self, token=None):
558
"""Wait for and acquire the lock.
560
:param token: if this is already locked, then lock_write will fail
561
unless the token matches the existing lock.
562
:returns: a token if this instance supports tokens, otherwise None.
563
:raises TokenLockingNotSupported: when a token is given but this
564
instance doesn't support using token locks.
565
:raises MismatchedToken: if the specified token doesn't match the token
566
of the existing lock.
568
A token should be passed in if you know that you have locked the object
569
some other way, and need to synchronise this object's state with that
572
XXX: docstring duplicated from LockableFiles.lock_write.
574
if token is not None:
575
self.validate_token(token)
577
self._lock_held = True
578
self._locked_via_token = True
581
return self.wait_lock()
584
"""Compatibility-mode shared lock.
586
LockDir doesn't support shared read-only locks, so this
587
just pretends that the lock is taken but really does nothing.
589
# At the moment Branches are commonly locked for read, but
590
# we can't rely on that remotely. Once this is cleaned up,
591
# reenable this warning to prevent it coming back in
593
## warn("LockDir.lock_read falls back to write lock")
594
if self._lock_held or self._fake_read_lock:
595
raise LockContention(self)
596
self._fake_read_lock = True
598
def _format_lock_info(self, info):
599
"""Turn the contents of peek() into something for the user"""
600
lock_url = self.transport.abspath(self.path)
601
delta = time.time() - int(info['start_time'])
603
'lock %s' % (lock_url,),
604
'held by %(user)s on host %(hostname)s [process #%(pid)s]' % info,
605
'locked %s' % (format_delta(delta),),
608
def validate_token(self, token):
609
if token is not None:
615
lock_token = info.get('nonce')
616
if token != lock_token:
617
raise errors.TokenMismatch(token, lock_token)
619
self._trace("revalidated by token %r", token)
621
def _trace(self, format, *args):
622
if 'lock' not in debug.debug_flags:
624
mutter(str(self) + ": " + (format % args))
added
removed
bzrlib/lockdir.py
