summaryrefslogtreecommitdiff
path: root/Documentation/technical/api-lockfile.txt
blob: 93b5f23e4c8a1bb45dc26d5f9cec402df7f57ee4 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
lockfile API
============

The lockfile API serves two purposes:

* Mutual exclusion and atomic file updates. When we want to change a
  file, we create a lockfile `<filename>.lock`, write the new file
  contents into it, and then rename the lockfile to its final
  destination `<filename>`. We create the `<filename>.lock` file with
  `O_CREAT|O_EXCL` so that we can notice and fail if somebody else has
  already locked the file, then atomically rename the lockfile to its
  final destination to commit the changes and unlock the file.

* Automatic cruft removal. If the program exits after we lock a file
  but before the changes have been committed, we want to make sure
  that we remove the lockfile. This is done by remembering the
  lockfiles we have created in a linked list and setting up an
  `atexit(3)` handler and a signal handler that clean up the
  lockfiles. This mechanism ensures that outstanding lockfiles are
  cleaned up if the program exits (including when `die()` is called)
  or if the program dies on a signal.

Please note that lockfiles only block other writers. Readers do not
block, but they are guaranteed to see either the old contents of the
file or the new contents of the file (assuming that the filesystem
implements `rename(2)` atomically).


Calling sequence
----------------

The caller:

* Allocates a `struct lock_file` either as a static variable or on the
  heap, initialized to zeros. Once you use the structure to call the
  `hold_lock_file_*` family of functions, it belongs to the lockfile
  subsystem and its storage must remain valid throughout the life of
  the program (i.e. you cannot use an on-stack variable to hold this
  structure).

* Attempts to create a lockfile by passing that variable and the path
  of the final destination (e.g. `$GIT_DIR/index`) to
  `hold_lock_file_for_update` or `hold_lock_file_for_append`.

* Writes new content for the destination file by either:

  * writing to the file descriptor returned by the `hold_lock_file_*`
    functions (also available via `lock->fd`).

  * calling `fdopen_lock_file` to get a `FILE` pointer for the open
    file and writing to the file using stdio.

When finished writing, the caller can:

* Close the file descriptor and rename the lockfile to its final
  destination by calling `commit_lock_file` or `commit_lock_file_to`.

* Close the file descriptor and remove the lockfile by calling
  `rollback_lock_file`.

* Close the file descriptor without removing or renaming the lockfile
  by calling `close_lock_file`, and later call `commit_lock_file`,
  `commit_lock_file_to`, `rollback_lock_file`, or `reopen_lock_file`.

Even after the lockfile is committed or rolled back, the `lock_file`
object must not be freed or altered by the caller. However, it may be
reused; just pass it to another call of `hold_lock_file_for_update` or
`hold_lock_file_for_append`.

If the program exits before you have called one of `commit_lock_file`,
`commit_lock_file_to`, `rollback_lock_file`, or `close_lock_file`, an
`atexit(3)` handler will close and remove the lockfile, rolling back
any uncommitted changes.

If you need to close the file descriptor you obtained from a
`hold_lock_file_*` function yourself, do so by calling
`close_lock_file`. You should never call `close(2)` or `fclose(3)`
yourself! Otherwise the `struct lock_file` structure would still think
that the file descriptor needs to be closed, and a commit or rollback
would result in duplicate calls to `close(2)`. Worse yet, if you close
and then later open another file descriptor for a completely different
purpose, then a commit or rollback might close that unrelated file
descriptor.


Error handling
--------------

The `hold_lock_file_*` functions return a file descriptor on success
or -1 on failure (unless `LOCK_DIE_ON_ERROR` is used; see below). On
errors, `errno` describes the reason for failure. Errors can be
reported by passing `errno` to one of the following helper functions:

unable_to_lock_message::

	Append an appropriate error message to a `strbuf`.

unable_to_lock_error::

	Emit an appropriate error message using `error()`.

unable_to_lock_die::

	Emit an appropriate error message and `die()`.

Similarly, `commit_lock_file`, `commit_lock_file_to`, and
`close_lock_file` return 0 on success. On failure they set `errno`
appropriately, do their best to roll back the lockfile, and return -1.


Flags
-----

The following flags can be passed to `hold_lock_file_for_update` or
`hold_lock_file_for_append`:

LOCK_NO_DEREF::

	Usually symbolic links in the destination path are resolved
	and the lockfile is created by adding ".lock" to the resolved
	path. If `LOCK_NO_DEREF` is set, then the lockfile is created
	by adding ".lock" to the path argument itself. This option is
	used, for example, when locking a symbolic reference, which
	for backwards-compatibility reasons can be a symbolic link
	containing the name of the referred-to-reference.

LOCK_DIE_ON_ERROR::

	If a lock is already taken for the file, `die()` with an error
	message. If this option is not specified, trying to lock a
	file that is already locked returns -1 to the caller.


The functions
-------------

hold_lock_file_for_update::

	Take a pointer to `struct lock_file`, the path of the file to
	be locked (e.g. `$GIT_DIR/index`) and a flags argument (see
	above). Attempt to create a lockfile for the destination and
	return the file descriptor for writing to the file.

hold_lock_file_for_append::

	Like `hold_lock_file_for_update`, but before returning copy
	the existing contents of the file (if any) to the lockfile and
	position its write pointer at the end of the file.

fdopen_lock_file::

	Associate a stdio stream with the lockfile. Return NULL
	(*without* rolling back the lockfile) on error. The stream is
	closed automatically when `close_lock_file` is called or when
	the file is committed or rolled back.

get_locked_file_path::

	Return the path of the file that is locked by the specified
	lock_file object. The caller must free the memory.

commit_lock_file::

	Take a pointer to the `struct lock_file` initialized with an
	earlier call to `hold_lock_file_for_update` or
	`hold_lock_file_for_append`, close the file descriptor, and
	rename the lockfile to its final destination. Return 0 upon
	success. On failure, roll back the lock file and return -1,
	with `errno` set to the value from the failing call to
	`close(2)` or `rename(2)`. It is a bug to call
	`commit_lock_file` for a `lock_file` object that is not
	currently locked.

commit_lock_file_to::

	Like `commit_lock_file()`, except that it takes an explicit
	`path` argument to which the lockfile should be renamed. The
	`path` must be on the same filesystem as the lock file.

rollback_lock_file::

	Take a pointer to the `struct lock_file` initialized with an
	earlier call to `hold_lock_file_for_update` or
	`hold_lock_file_for_append`, close the file descriptor and
	remove the lockfile. It is a NOOP to call
	`rollback_lock_file()` for a `lock_file` object that has
	already been committed or rolled back.

close_lock_file::

	Take a pointer to the `struct lock_file` initialized with an
	earlier call to `hold_lock_file_for_update` or
	`hold_lock_file_for_append`. Close the file descriptor (and
	the file pointer if it has been opened using
	`fdopen_lock_file`). Return 0 upon success. On failure to
	`close(2)`, return a negative value and roll back the lock
	file. Usually `commit_lock_file`, `commit_lock_file_to`, or
	`rollback_lock_file` should eventually be called if
	`close_lock_file` succeeds.

reopen_lock_file::

	Re-open a lockfile that has been closed (using
	`close_lock_file`) but not yet committed or rolled back. This
	can be used to implement a sequence of operations like the
	following:

	* Lock file.

	* Write new contents to lockfile, then `close_lock_file` to
	  cause the contents to be written to disk.

	* Pass the name of the lockfile to another program to allow it
	  (and nobody else) to inspect the contents you wrote, while
	  still holding the lock yourself.

	* `reopen_lock_file` to reopen the lockfile. Make further
	  updates to the contents.

	* `commit_lock_file` to make the final version permanent.