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
assert isinstance(transport, Transport), \
171
("not a transport: %r" % transport)
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
self._trace("after locking, info=%r", info)
245
if info['nonce'] != self.nonce:
246
self._trace("rename succeeded, "
247
"but lock is still held by someone else")
248
raise LockContention(self)
249
self._lock_held = True
250
self._trace("... lock succeeded after %dms",
251
(time.time() - start_time) * 1000)
254
def _remove_pending_dir(self, tmpname):
255
"""Remove the pending directory
257
This is called if we failed to rename into place, so that the pending
258
dirs don't clutter up the lockdir.
260
self._trace("remove %s", tmpname)
262
self.transport.delete(tmpname + self.__INFO_NAME)
263
self.transport.rmdir(tmpname)
265
note("error removing pending lock: %s", e)
267
def _create_pending_dir(self):
268
tmpname = '%s/%s.tmp' % (self.path, rand_chars(10))
270
self.transport.mkdir(tmpname)
272
# This may raise a FileExists exception
273
# which is okay, it will be caught later and determined
274
# to be a LockContention.
275
self._trace("lock directory does not exist, creating it")
276
self.create(mode=self._dir_modebits)
277
# After creating the lock directory, try again
278
self.transport.mkdir(tmpname)
279
self.nonce = rand_chars(20)
280
info_bytes = self._prepare_info()
281
# We use put_file_non_atomic because we just created a new unique
282
# directory so we don't have to worry about files existing there.
283
# We'll rename the whole directory into place to get atomic
285
self.transport.put_bytes_non_atomic(tmpname + self.__INFO_NAME,
290
"""Release a held lock
292
if self._fake_read_lock:
293
self._fake_read_lock = False
295
if not self._lock_held:
296
raise LockNotHeld(self)
297
if self._locked_via_token:
298
self._locked_via_token = False
299
self._lock_held = False
301
# rename before deleting, because we can't atomically remove the
303
start_time = time.time()
304
self._trace("unlocking")
305
tmpname = '%s/releasing.%s.tmp' % (self.path, rand_chars(20))
306
# gotta own it to unlock
308
self.transport.rename(self._held_dir, tmpname)
309
self._lock_held = False
310
self.transport.delete(tmpname + self.__INFO_NAME)
312
self.transport.rmdir(tmpname)
313
except DirectoryNotEmpty, e:
314
# There might have been junk left over by a rename that moved
315
# another locker within the 'held' directory. do a slower
316
# deletion where we list the directory and remove everything
319
# Maybe this should be broader to allow for ftp servers with
320
# non-specific error messages?
321
self._trace("doing recursive deletion of non-empty directory "
323
self.transport.delete_tree(tmpname)
324
self._trace("... unlock succeeded after %dms",
325
(time.time() - start_time) * 1000)
327
def break_lock(self):
328
"""Break a lock not held by this instance of LockDir.
330
This is a UI centric function: it uses the bzrlib.ui.ui_factory to
331
prompt for input if a lock is detected and there is any doubt about
332
it possibly being still active.
334
self._check_not_locked()
335
holder_info = self.peek()
336
if holder_info is not None:
337
lock_info = '\n'.join(self._format_lock_info(holder_info))
338
if bzrlib.ui.ui_factory.get_boolean("Break %s" % lock_info):
339
self.force_break(holder_info)
341
def force_break(self, dead_holder_info):
342
"""Release a lock held by another process.
344
WARNING: This should only be used when the other process is dead; if
345
it still thinks it has the lock there will be two concurrent writers.
346
In general the user's approval should be sought for lock breaks.
348
dead_holder_info must be the result of a previous LockDir.peek() call;
349
this is used to check that it's still held by the same process that
350
the user decided was dead. If this is not the current holder,
351
LockBreakMismatch is raised.
353
After the lock is broken it will not be held by any process.
354
It is possible that another process may sneak in and take the
355
lock before the breaking process acquires it.
357
if not isinstance(dead_holder_info, dict):
358
raise ValueError("dead_holder_info: %r" % dead_holder_info)
359
self._check_not_locked()
360
current_info = self.peek()
361
if current_info is None:
362
# must have been recently released
364
if current_info != dead_holder_info:
365
raise LockBreakMismatch(self, current_info, dead_holder_info)
366
tmpname = '%s/broken.%s.tmp' % (self.path, rand_chars(20))
367
self.transport.rename(self._held_dir, tmpname)
368
# check that we actually broke the right lock, not someone else;
369
# there's a small race window between checking it and doing the
371
broken_info_path = tmpname + self.__INFO_NAME
372
broken_info = self._read_info_file(broken_info_path)
373
if broken_info != dead_holder_info:
374
raise LockBreakMismatch(self, broken_info, dead_holder_info)
375
self.transport.delete(broken_info_path)
376
self.transport.rmdir(tmpname)
378
def _check_not_locked(self):
379
"""If the lock is held by this instance, raise an error."""
381
raise AssertionError("can't break own lock: %r" % self)
384
"""Make sure that the lock is still held by this locker.
386
This should only fail if the lock was broken by user intervention,
387
or if the lock has been affected by a bug.
389
If the lock is not thought to be held, raises LockNotHeld. If
390
the lock is thought to be held but has been broken, raises
393
if not self._lock_held:
394
raise LockNotHeld(self)
397
# no lock there anymore!
398
raise LockBroken(self)
399
if info.get('nonce') != self.nonce:
400
# there is a lock, but not ours
401
raise LockBroken(self)
403
def _read_info_file(self, path):
404
"""Read one given info file.
406
peek() reads the info file of the lock holder, if any.
408
return self._parse_info(self.transport.get(path))
411
"""Check if the lock is held by anyone.
413
If it is held, this returns the lock info structure as a rio Stanza,
414
which contains some information about the current lock holder.
415
Otherwise returns None.
418
info = self._read_info_file(self._held_info_path)
419
self._trace("peek -> held")
420
assert isinstance(info, dict), \
421
"bad parse result %r" % info
423
except NoSuchFile, e:
424
self._trace("peek -> not held")
426
def _prepare_info(self):
427
"""Write information about a pending lock to a temporary file.
430
# XXX: is creating this here inefficient?
431
config = bzrlib.config.GlobalConfig()
433
user = config.user_email()
434
except errors.NoEmailInUsername:
435
user = config.username()
436
s = Stanza(hostname=socket.gethostname(),
437
pid=str(os.getpid()),
438
start_time=str(int(time.time())),
444
def _parse_info(self, info_file):
445
return read_stanza(info_file.readlines()).as_dict()
447
def attempt_lock(self):
448
"""Take the lock; fail if it's already held.
450
If you wish to block until the lock can be obtained, call wait_lock()
453
:return: The lock token.
454
:raises LockContention: if the lock is held by someone else.
456
if self._fake_read_lock:
457
raise LockContention(self)
458
return self._attempt_lock()
460
def wait_lock(self, timeout=None, poll=None, max_attempts=None):
461
"""Wait a certain period for a lock.
463
If the lock can be acquired within the bounded time, it
464
is taken and this returns. Otherwise, LockContention
465
is raised. Either way, this function should return within
466
approximately `timeout` seconds. (It may be a bit more if
467
a transport operation takes a long time to complete.)
469
:param timeout: Approximate maximum amount of time to wait for the
472
:param poll: Delay in seconds between retrying the lock.
474
:param max_attempts: Maximum number of times to try to lock.
476
:return: The lock token.
479
timeout = _DEFAULT_TIMEOUT_SECONDS
481
poll = _DEFAULT_POLL_SECONDS
482
# XXX: the transport interface doesn't let us guard against operations
483
# there taking a long time, so the total elapsed time or poll interval
484
# may be more than was requested.
485
deadline = time.time() + timeout
492
return self.attempt_lock()
493
except LockContention:
494
# possibly report the blockage, then try again
496
# TODO: In a few cases, we find out that there's contention by
497
# reading the held info and observing that it's not ours. In
498
# those cases it's a bit redundant to read it again. However,
499
# the normal case (??) is that the rename fails and so we
500
# don't know who holds the lock. For simplicity we peek
502
new_info = self.peek()
503
if new_info is not None and new_info != last_info:
504
if last_info is None:
505
start = 'Unable to obtain'
507
start = 'Lock owner changed for'
509
formatted_info = self._format_lock_info(new_info)
510
if deadline_str is None:
511
deadline_str = time.strftime('%H:%M:%S',
512
time.localtime(deadline))
513
self._report_function('%s %s\n'
515
'%s\n' # locked ... ago
516
'Will continue to try until %s\n',
523
if (max_attempts is not None) and (attempt_count >= max_attempts):
524
self._trace("exceeded %d attempts")
525
raise LockContention(self)
526
if time.time() + poll < deadline:
527
self._trace("waiting %ss", poll)
530
self._trace("timeout after waiting %ss", timeout)
531
raise LockContention(self)
533
def leave_in_place(self):
534
self._locked_via_token = True
536
def dont_leave_in_place(self):
537
self._locked_via_token = False
539
def lock_write(self, token=None):
540
"""Wait for and acquire the lock.
542
:param token: if this is already locked, then lock_write will fail
543
unless the token matches the existing lock.
544
:returns: a token if this instance supports tokens, otherwise None.
545
:raises TokenLockingNotSupported: when a token is given but this
546
instance doesn't support using token locks.
547
:raises MismatchedToken: if the specified token doesn't match the token
548
of the existing lock.
550
A token should be passed in if you know that you have locked the object
551
some other way, and need to synchronise this object's state with that
554
XXX: docstring duplicated from LockableFiles.lock_write.
556
if token is not None:
557
self.validate_token(token)
559
self._lock_held = True
560
self._locked_via_token = True
563
return self.wait_lock()
566
"""Compatibility-mode shared lock.
568
LockDir doesn't support shared read-only locks, so this
569
just pretends that the lock is taken but really does nothing.
571
# At the moment Branches are commonly locked for read, but
572
# we can't rely on that remotely. Once this is cleaned up,
573
# reenable this warning to prevent it coming back in
575
## warn("LockDir.lock_read falls back to write lock")
576
if self._lock_held or self._fake_read_lock:
577
raise LockContention(self)
578
self._fake_read_lock = True
580
def _format_lock_info(self, info):
581
"""Turn the contents of peek() into something for the user"""
582
lock_url = self.transport.abspath(self.path)
583
delta = time.time() - int(info['start_time'])
585
'lock %s' % (lock_url,),
586
'held by %(user)s on host %(hostname)s [process #%(pid)s]' % info,
587
'locked %s' % (format_delta(delta),),
590
def validate_token(self, token):
591
if token is not None:
597
lock_token = info.get('nonce')
598
if token != lock_token:
599
raise errors.TokenMismatch(token, lock_token)
601
self._trace("revalidated by token %r", token)
603
def _trace(self, format, *args):
604
if 'lock' not in debug.debug_flags:
606
mutter(str(self) + ": " + (format % args))
added
removed
bzrlib/lockdir.py
