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 (
127
from bzrlib.trace import mutter, note
128
from bzrlib.transport import Transport
129
from bzrlib.osutils import rand_chars, format_delta
130
from bzrlib.rio import read_stanza, Stanza
134
# XXX: At the moment there is no consideration of thread safety on LockDir
135
# objects. This should perhaps be updated - e.g. if two threads try to take a
136
# lock at the same time they should *both* get it. But then that's unlikely
139
# TODO: Perhaps store some kind of note like the bzr command line in the lock
142
# TODO: Some kind of callback run while polling a lock to show progress
145
# TODO: Make sure to pass the right file and directory mode bits to all
146
# files/dirs created.
149
_DEFAULT_TIMEOUT_SECONDS = 300
150
_DEFAULT_POLL_SECONDS = 1.0
153
class LockDir(object):
154
"""Write-lock guarding access to data."""
156
__INFO_NAME = '/info'
158
def __init__(self, transport, path, file_modebits=0644, dir_modebits=0755):
159
"""Create a new LockDir object.
161
The LockDir is initially unlocked - this just creates the object.
163
:param transport: Transport which will contain the lock
165
:param path: Path to the lock within the base directory of the
168
assert isinstance(transport, Transport), \
169
("not a transport: %r" % transport)
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
if self.transport.is_readonly():
196
raise UnlockableTransport(self.transport)
197
self._trace("create lock directory")
198
self.transport.mkdir(self.path, mode=mode)
200
def _attempt_lock(self):
201
"""Make the pending directory and attempt to rename into place.
203
If the rename succeeds, we read back the info file to check that we
206
If we fail to acquire the lock, this method is responsible for
207
cleaning up the pending directory if possible. (But it doesn't do
210
:returns: The nonce of the lock, if it was successfully acquired.
212
:raises LockContention: If the lock is held by someone else. The exception
213
contains the info of the current holder of the lock.
215
self._trace("lock_write...")
216
start_time = time.time()
217
tmpname = self._create_pending_dir()
219
self.transport.rename(tmpname, self._held_dir)
220
except (PathError, DirectoryNotEmpty, FileExists, ResourceBusy), e:
221
self._trace("... contention, %s", e)
222
self._remove_pending_dir(tmpname)
223
raise LockContention(self)
225
self._trace("... lock failed, %s", e)
226
self._remove_pending_dir(tmpname)
228
# We must check we really got the lock, because Launchpad's sftp
229
# server at one time had a bug were the rename would successfully
230
# move the new directory into the existing directory, which was
231
# incorrect. It's possible some other servers or filesystems will
232
# have a similar bug allowing someone to think they got the lock
233
# when it's already held.
235
self._trace("after locking, info=%r", info)
236
if info['nonce'] != self.nonce:
237
self._trace("rename succeeded, "
238
"but lock is still held by someone else")
239
raise LockContention(self)
240
self._lock_held = True
241
self._trace("... lock succeeded after %dms",
242
(time.time() - start_time) * 1000)
245
def _remove_pending_dir(self, tmpname):
246
"""Remove the pending directory
248
This is called if we failed to rename into place, so that the pending
249
dirs don't clutter up the lockdir.
251
self._trace("remove %s", tmpname)
253
self.transport.delete(tmpname + self.__INFO_NAME)
254
self.transport.rmdir(tmpname)
256
note("error removing pending lock: %s", e)
258
def _create_pending_dir(self):
259
tmpname = '%s/%s.tmp' % (self.path, rand_chars(10))
261
self.transport.mkdir(tmpname)
263
# This may raise a FileExists exception
264
# which is okay, it will be caught later and determined
265
# to be a LockContention.
266
self._trace("lock directory does not exist, creating it")
267
self.create(mode=self._dir_modebits)
268
# After creating the lock directory, try again
269
self.transport.mkdir(tmpname)
270
self.nonce = rand_chars(20)
271
info_bytes = self._prepare_info()
272
# We use put_file_non_atomic because we just created a new unique
273
# directory so we don't have to worry about files existing there.
274
# We'll rename the whole directory into place to get atomic
276
self.transport.put_bytes_non_atomic(tmpname + self.__INFO_NAME,
281
"""Release a held lock
283
if self._fake_read_lock:
284
self._fake_read_lock = False
286
if not self._lock_held:
287
raise LockNotHeld(self)
288
if self._locked_via_token:
289
self._locked_via_token = False
290
self._lock_held = False
292
# rename before deleting, because we can't atomically remove the
294
start_time = time.time()
295
self._trace("unlocking")
296
tmpname = '%s/releasing.%s.tmp' % (self.path, rand_chars(20))
297
# gotta own it to unlock
299
self.transport.rename(self._held_dir, tmpname)
300
self._lock_held = False
301
self.transport.delete(tmpname + self.__INFO_NAME)
303
self.transport.rmdir(tmpname)
304
except DirectoryNotEmpty, e:
305
# There might have been junk left over by a rename that moved
306
# another locker within the 'held' directory. do a slower
307
# deletion where we list the directory and remove everything
310
# Maybe this should be broader to allow for ftp servers with
311
# non-specific error messages?
312
self._trace("doing recursive deletion of non-empty directory "
314
self.transport.delete_tree(tmpname)
315
self._trace("... unlock succeeded after %dms",
316
(time.time() - start_time) * 1000)
318
def break_lock(self):
319
"""Break a lock not held by this instance of LockDir.
321
This is a UI centric function: it uses the bzrlib.ui.ui_factory to
322
prompt for input if a lock is detected and there is any doubt about
323
it possibly being still active.
325
self._check_not_locked()
326
holder_info = self.peek()
327
if holder_info is not None:
328
lock_info = '\n'.join(self._format_lock_info(holder_info))
329
if bzrlib.ui.ui_factory.get_boolean("Break %s" % lock_info):
330
self.force_break(holder_info)
332
def force_break(self, dead_holder_info):
333
"""Release a lock held by another process.
335
WARNING: This should only be used when the other process is dead; if
336
it still thinks it has the lock there will be two concurrent writers.
337
In general the user's approval should be sought for lock breaks.
339
dead_holder_info must be the result of a previous LockDir.peek() call;
340
this is used to check that it's still held by the same process that
341
the user decided was dead. If this is not the current holder,
342
LockBreakMismatch is raised.
344
After the lock is broken it will not be held by any process.
345
It is possible that another process may sneak in and take the
346
lock before the breaking process acquires it.
348
if not isinstance(dead_holder_info, dict):
349
raise ValueError("dead_holder_info: %r" % dead_holder_info)
350
self._check_not_locked()
351
current_info = self.peek()
352
if current_info is None:
353
# must have been recently released
355
if current_info != dead_holder_info:
356
raise LockBreakMismatch(self, current_info, dead_holder_info)
357
tmpname = '%s/broken.%s.tmp' % (self.path, rand_chars(20))
358
self.transport.rename(self._held_dir, tmpname)
359
# check that we actually broke the right lock, not someone else;
360
# there's a small race window between checking it and doing the
362
broken_info_path = tmpname + self.__INFO_NAME
363
broken_info = self._read_info_file(broken_info_path)
364
if broken_info != dead_holder_info:
365
raise LockBreakMismatch(self, broken_info, dead_holder_info)
366
self.transport.delete(broken_info_path)
367
self.transport.rmdir(tmpname)
369
def _check_not_locked(self):
370
"""If the lock is held by this instance, raise an error."""
372
raise AssertionError("can't break own lock: %r" % self)
375
"""Make sure that the lock is still held by this locker.
377
This should only fail if the lock was broken by user intervention,
378
or if the lock has been affected by a bug.
380
If the lock is not thought to be held, raises LockNotHeld. If
381
the lock is thought to be held but has been broken, raises
384
if not self._lock_held:
385
raise LockNotHeld(self)
388
# no lock there anymore!
389
raise LockBroken(self)
390
if info.get('nonce') != self.nonce:
391
# there is a lock, but not ours
392
raise LockBroken(self)
394
def _read_info_file(self, path):
395
"""Read one given info file.
397
peek() reads the info file of the lock holder, if any.
399
return self._parse_info(self.transport.get(path))
402
"""Check if the lock is held by anyone.
404
If it is held, this returns the lock info structure as a rio Stanza,
405
which contains some information about the current lock holder.
406
Otherwise returns None.
409
info = self._read_info_file(self._held_info_path)
410
self._trace("peek -> held")
411
assert isinstance(info, dict), \
412
"bad parse result %r" % info
414
except NoSuchFile, e:
415
self._trace("peek -> not held")
417
def _prepare_info(self):
418
"""Write information about a pending lock to a temporary file.
421
# XXX: is creating this here inefficient?
422
config = bzrlib.config.GlobalConfig()
424
user = config.user_email()
425
except errors.NoEmailInUsername:
426
user = config.username()
427
s = Stanza(hostname=socket.gethostname(),
428
pid=str(os.getpid()),
429
start_time=str(int(time.time())),
435
def _parse_info(self, info_file):
436
return read_stanza(info_file.readlines()).as_dict()
438
def attempt_lock(self):
439
"""Take the lock; fail if it's already held.
441
If you wish to block until the lock can be obtained, call wait_lock()
444
:return: The lock token.
445
:raises LockContention: if the lock is held by someone else.
447
if self._fake_read_lock:
448
raise LockContention(self)
449
if self.transport.is_readonly():
450
raise UnlockableTransport(self.transport)
451
return self._attempt_lock()
453
def wait_lock(self, timeout=None, poll=None, max_attempts=None):
454
"""Wait a certain period for a lock.
456
If the lock can be acquired within the bounded time, it
457
is taken and this returns. Otherwise, LockContention
458
is raised. Either way, this function should return within
459
approximately `timeout` seconds. (It may be a bit more if
460
a transport operation takes a long time to complete.)
462
:param timeout: Approximate maximum amount of time to wait for the
465
:param poll: Delay in seconds between retrying the lock.
467
:param max_attempts: Maximum number of times to try to lock.
469
:return: The lock token.
472
timeout = _DEFAULT_TIMEOUT_SECONDS
474
poll = _DEFAULT_POLL_SECONDS
475
# XXX: the transport interface doesn't let us guard against operations
476
# there taking a long time, so the total elapsed time or poll interval
477
# may be more than was requested.
478
deadline = time.time() + timeout
485
return self.attempt_lock()
486
except LockContention:
487
# possibly report the blockage, then try again
489
# TODO: In a few cases, we find out that there's contention by
490
# reading the held info and observing that it's not ours. In
491
# those cases it's a bit redundant to read it again. However,
492
# the normal case (??) is that the rename fails and so we
493
# don't know who holds the lock. For simplicity we peek
495
new_info = self.peek()
496
if new_info is not None and new_info != last_info:
497
if last_info is None:
498
start = 'Unable to obtain'
500
start = 'Lock owner changed for'
502
formatted_info = self._format_lock_info(new_info)
503
if deadline_str is None:
504
deadline_str = time.strftime('%H:%M:%S',
505
time.localtime(deadline))
506
self._report_function('%s %s\n'
508
'%s\n' # locked ... ago
509
'Will continue to try until %s\n',
516
if (max_attempts is not None) and (attempt_count >= max_attempts):
517
self._trace("exceeded %d attempts")
518
raise LockContention(self)
519
if time.time() + poll < deadline:
520
self._trace("waiting %ss", poll)
523
self._trace("timeout after waiting %ss", timeout)
524
raise LockContention(self)
526
def leave_in_place(self):
527
self._locked_via_token = True
529
def dont_leave_in_place(self):
530
self._locked_via_token = False
532
def lock_write(self, token=None):
533
"""Wait for and acquire the lock.
535
:param token: if this is already locked, then lock_write will fail
536
unless the token matches the existing lock.
537
:returns: a token if this instance supports tokens, otherwise None.
538
:raises TokenLockingNotSupported: when a token is given but this
539
instance doesn't support using token locks.
540
:raises MismatchedToken: if the specified token doesn't match the token
541
of the existing lock.
543
A token should be passed in if you know that you have locked the object
544
some other way, and need to synchronise this object's state with that
547
XXX: docstring duplicated from LockableFiles.lock_write.
549
if token is not None:
550
self.validate_token(token)
552
self._lock_held = True
553
self._locked_via_token = True
556
return self.wait_lock()
559
"""Compatibility-mode shared lock.
561
LockDir doesn't support shared read-only locks, so this
562
just pretends that the lock is taken but really does nothing.
564
# At the moment Branches are commonly locked for read, but
565
# we can't rely on that remotely. Once this is cleaned up,
566
# reenable this warning to prevent it coming back in
568
## warn("LockDir.lock_read falls back to write lock")
569
if self._lock_held or self._fake_read_lock:
570
raise LockContention(self)
571
self._fake_read_lock = True
573
def _format_lock_info(self, info):
574
"""Turn the contents of peek() into something for the user"""
575
lock_url = self.transport.abspath(self.path)
576
delta = time.time() - int(info['start_time'])
578
'lock %s' % (lock_url,),
579
'held by %(user)s on host %(hostname)s [process #%(pid)s]' % info,
580
'locked %s' % (format_delta(delta),),
583
def validate_token(self, token):
584
if token is not None:
590
lock_token = info.get('nonce')
591
if token != lock_token:
592
raise errors.TokenMismatch(token, lock_token)
594
self._trace("revalidated by token %r", token)
596
def _trace(self, format, *args):
597
if 'lock' not in debug.debug_flags:
599
mutter(str(self) + ": " + (format % args))
added
removed
bzrlib/lockdir.py
