| 1 |
.\" Copyright 1993, Rickard E. Faith <[email protected]> |
| 2 |
.\" Copyright 2002, Michael Kerrisk <[email protected]> |
| 3 |
.\" Copyright, the authors of the Linux man-pages project |
| 4 |
.\" |
| 5 |
.\" SPDX-License-Identifier: Linux-man-pages-copyleft |
| 6 |
.\" |
| 7 |
.\" FIXME Maybe document LOCK_MAND, LOCK_RW, LOCK_READ, LOCK_WRITE |
| 8 |
.\" which only have effect for SAMBA. |
| 9 |
.\" |
| 10 |
.TH flock 2 2026-02-08 "Linux man-pages 6.18" |
| 11 |
.SH NAME |
| 12 |
flock \- apply or remove an advisory lock on an open file |
| 13 |
.SH LIBRARY |
| 14 |
Standard C library |
| 15 |
.RI ( libc ,\~ \-lc ) |
| 16 |
.SH SYNOPSIS |
| 17 |
.nf |
| 18 |
.B #include <sys/file.h> |
| 19 |
.P |
| 20 |
.BI "int flock(int " fd ", int " op ); |
| 21 |
.fi |
| 22 |
.SH DESCRIPTION |
| 23 |
Apply or remove an advisory lock on the open file specified by |
| 24 |
.IR fd . |
| 25 |
The argument |
| 26 |
.I op |
| 27 |
is one of the following: |
| 28 |
.RS 4 |
| 29 |
.TP 9 |
| 30 |
.B LOCK_SH |
| 31 |
Place a shared lock. |
| 32 |
More than one process may hold a shared lock for a given file |
| 33 |
at a given time. |
| 34 |
.TP |
| 35 |
.B LOCK_EX |
| 36 |
Place an exclusive lock. |
| 37 |
Only one process may hold an exclusive lock for a given |
| 38 |
file at a given time. |
| 39 |
.TP |
| 40 |
.B LOCK_UN |
| 41 |
Remove an existing lock held by this process. |
| 42 |
.RE |
| 43 |
.P |
| 44 |
A call to |
| 45 |
.BR flock () |
| 46 |
may block if an incompatible lock is held by another process. |
| 47 |
To make a nonblocking request, include |
| 48 |
.B LOCK_NB |
| 49 |
(by ORing) |
| 50 |
with any of the above operations. |
| 51 |
.P |
| 52 |
A single file may not simultaneously have both shared and exclusive locks. |
| 53 |
.P |
| 54 |
Locks created by |
| 55 |
.BR flock () |
| 56 |
are associated with an open file description (see |
| 57 |
.BR open (2)). |
| 58 |
This means that duplicate file descriptors (created by, for example, |
| 59 |
.BR fork (2) |
| 60 |
or |
| 61 |
.BR dup (2)) |
| 62 |
refer to the same lock, and this lock may be modified |
| 63 |
or released using any of these file descriptors. |
| 64 |
Furthermore, the lock is released either by an explicit |
| 65 |
.B LOCK_UN |
| 66 |
operation on any of these duplicate file descriptors, or when all |
| 67 |
such file descriptors have been closed. |
| 68 |
.P |
| 69 |
If a process uses |
| 70 |
.BR open (2) |
| 71 |
(or similar) to obtain more than one file descriptor for the same file, |
| 72 |
these file descriptors are treated independently by |
| 73 |
.BR flock (). |
| 74 |
An attempt to lock the file using one of these file descriptors |
| 75 |
may be denied by a lock that the calling process has |
| 76 |
already placed via another file descriptor. |
| 77 |
.P |
| 78 |
A process may hold only one type of lock (shared or exclusive) |
| 79 |
on a file. |
| 80 |
Subsequent |
| 81 |
.BR flock () |
| 82 |
calls on an already locked file will convert an existing lock to the new |
| 83 |
lock mode. |
| 84 |
.P |
| 85 |
Locks created by |
| 86 |
.BR flock () |
| 87 |
are preserved across an |
| 88 |
.BR execve (2). |
| 89 |
.P |
| 90 |
A shared or exclusive lock can be placed on a file regardless of the |
| 91 |
mode in which the file was opened. |
| 92 |
.SH RETURN VALUE |
| 93 |
On success, zero is returned. |
| 94 |
On error, \-1 is returned, and |
| 95 |
.I errno |
| 96 |
is set to indicate the error. |
| 97 |
.SH ERRORS |
| 98 |
.TP |
| 99 |
.B EBADF |
| 100 |
.I fd |
| 101 |
is not an open file descriptor. |
| 102 |
.TP |
| 103 |
.B EINTR |
| 104 |
While waiting to acquire a lock, the call was interrupted by |
| 105 |
delivery of a signal caught by a handler; |
| 106 |
see |
| 107 |
.BR signal (7). |
| 108 |
.TP |
| 109 |
.B EINVAL |
| 110 |
.I op |
| 111 |
is invalid. |
| 112 |
.TP |
| 113 |
.B ENOLCK |
| 114 |
The kernel ran out of memory for allocating lock records. |
| 115 |
.TP |
| 116 |
.B EWOULDBLOCK |
| 117 |
The file is locked and the |
| 118 |
.B LOCK_NB |
| 119 |
flag was selected. |
| 120 |
.SH VERSIONS |
| 121 |
Since Linux 2.0, |
| 122 |
.BR flock () |
| 123 |
is implemented as a system call in its own right rather |
| 124 |
than being emulated in the GNU C library as a call to |
| 125 |
.BR fcntl (2). |
| 126 |
With this implementation, |
| 127 |
there is no interaction between the types of lock |
| 128 |
placed by |
| 129 |
.BR flock () |
| 130 |
and |
| 131 |
.BR fcntl (2), |
| 132 |
and |
| 133 |
.BR flock () |
| 134 |
does not detect deadlock. |
| 135 |
(Note, however, that on some systems, such as the modern BSDs, |
| 136 |
.\" E.g., according to the flock(2) man page, FreeBSD since at least 5.3 |
| 137 |
.BR flock () |
| 138 |
and |
| 139 |
.BR fcntl (2) |
| 140 |
locks |
| 141 |
.I do |
| 142 |
interact with one another.) |
| 143 |
.SS CIFS details |
| 144 |
Up to Linux 5.4, |
| 145 |
.BR flock () |
| 146 |
is not propagated over SMB. |
| 147 |
A file with such locks will not appear locked for remote clients. |
| 148 |
.P |
| 149 |
Since Linux 5.5, |
| 150 |
.BR flock () |
| 151 |
locks are emulated with SMB byte-range locks on the entire file. |
| 152 |
Similarly to NFS, this means that |
| 153 |
.BR fcntl (2) |
| 154 |
and |
| 155 |
.BR flock () |
| 156 |
locks interact with one another. |
| 157 |
Another important side-effect is that the locks are not advisory anymore: |
| 158 |
any IO on a locked file will always fail with |
| 159 |
.B EACCES |
| 160 |
when done from a separate file descriptor. |
| 161 |
This difference originates from the design of locks in the SMB protocol, |
| 162 |
which provides mandatory locking semantics. |
| 163 |
.P |
| 164 |
Remote and mandatory locking semantics may vary with |
| 165 |
SMB protocol, mount options and server type. |
| 166 |
See |
| 167 |
.BR mount.cifs (8) |
| 168 |
for additional information. |
| 169 |
.SH STANDARDS |
| 170 |
BSD. |
| 171 |
.SH HISTORY |
| 172 |
4.4BSD (the |
| 173 |
.BR flock () |
| 174 |
call first appeared in 4.2BSD). |
| 175 |
A version of |
| 176 |
.BR flock (), |
| 177 |
possibly implemented in terms of |
| 178 |
.BR fcntl (2), |
| 179 |
appears on most UNIX systems. |
| 180 |
.SS NFS details |
| 181 |
Up to Linux 2.6.11, |
| 182 |
.BR flock () |
| 183 |
does not lock files over NFS |
| 184 |
(i.e., the scope of locks was limited to the local system). |
| 185 |
Instead, one could use |
| 186 |
.BR fcntl (2) |
| 187 |
byte-range locking, which does work over NFS, |
| 188 |
given a sufficiently recent version of |
| 189 |
Linux and a server which supports locking. |
| 190 |
.P |
| 191 |
Since Linux 2.6.12, NFS clients support |
| 192 |
.BR flock () |
| 193 |
locks by emulating them as |
| 194 |
.BR fcntl (2) |
| 195 |
byte-range locks on the entire file. |
| 196 |
This means that |
| 197 |
.BR fcntl (2) |
| 198 |
and |
| 199 |
.BR flock () |
| 200 |
locks |
| 201 |
.I do |
| 202 |
interact with one another over NFS. |
| 203 |
It also means that in order to place an exclusive lock, |
| 204 |
the file must be opened for writing. |
| 205 |
.P |
| 206 |
Since Linux 2.6.37, |
| 207 |
.\" commit 5eebde23223aeb0ad2d9e3be6590ff8bbfab0fc2 |
| 208 |
the kernel supports a compatibility mode that allows |
| 209 |
.BR flock () |
| 210 |
locks (and also |
| 211 |
.BR fcntl (2) |
| 212 |
byte region locks) to be treated as local; |
| 213 |
see the discussion of the |
| 214 |
.I "local_lock" |
| 215 |
option in |
| 216 |
.BR nfs (5). |
| 217 |
.SH NOTES |
| 218 |
.BR flock () |
| 219 |
places advisory locks only; |
| 220 |
given suitable permissions on a file, |
| 221 |
a process is free to ignore the use of |
| 222 |
.BR flock () |
| 223 |
and perform I/O on the file. |
| 224 |
.P |
| 225 |
.BR flock () |
| 226 |
and |
| 227 |
.BR fcntl (2) |
| 228 |
locks have different semantics with respect to forked processes and |
| 229 |
.BR dup (2). |
| 230 |
On systems that implement |
| 231 |
.BR flock () |
| 232 |
using |
| 233 |
.BR fcntl (2), |
| 234 |
the semantics of |
| 235 |
.BR flock () |
| 236 |
will be different from those described in this manual page. |
| 237 |
.P |
| 238 |
Converting a lock |
| 239 |
(shared to exclusive, or vice versa) is not guaranteed to be atomic: |
| 240 |
the existing lock is first removed, and then a new lock is established. |
| 241 |
Between these two steps, |
| 242 |
a pending lock request by another process may be granted, |
| 243 |
with the result that the conversion either blocks, or fails if |
| 244 |
.B LOCK_NB |
| 245 |
was specified. |
| 246 |
(This is the original BSD behavior, |
| 247 |
and occurs on many other implementations.) |
| 248 |
.\" Kernel 2.5.21 changed things a little: during lock conversion |
| 249 |
.\" it is now the highest priority process that will get the lock -- mtk |
| 250 |
.P |
| 251 |
Release of a lock when a file descriptor is closed |
| 252 |
is not sequenced after all observable effects of |
| 253 |
.BR close (2). |
| 254 |
For example, |
| 255 |
if one process writes a file while holding an exclusive lock, |
| 256 |
then closes that file, |
| 257 |
and another process blocks placing a shared lock on that file |
| 258 |
to wait until it is closed, |
| 259 |
it may observe that a subsequent |
| 260 |
.BR execve (2) |
| 261 |
of that file fails with |
| 262 |
.BR ETXTBSY , |
| 263 |
and |
| 264 |
.BR umount (2) |
| 265 |
of its underlying filesystem fails with |
| 266 |
.BR EBUSY , |
| 267 |
as if the file were still open in the first process. |
| 268 |
.SH SEE ALSO |
| 269 |
.BR flock (1), |
| 270 |
.BR close (2), |
| 271 |
.BR dup (2), |
| 272 |
.BR execve (2), |
| 273 |
.BR fcntl (2), |
| 274 |
.BR fork (2), |
| 275 |
.BR open (2), |
| 276 |
.BR lockf (3), |
| 277 |
.BR lslocks (8) |
| 278 |
.P |
| 279 |
.I Documentation/filesystems/locks.rst |
| 280 |
in the Linux kernel source tree |