1WRITE(P) POSIX Programmer's Manual WRITE(P)
2
6 pwrite, write - write on a file
7
9 #include <unistd.h>
10 ssize_t pwrite(int fildes, const void *buf, size_t nbyte,
14 off_t offset);
15 ssize_t write(int fildes, const void *buf, size_t nbyte);
16
19 The write() function shall attempt to write nbyte bytes from the buffer
20 pointed to by buf to the file associated with the open file descriptor,
21 fildes.
22 Before any action described below is taken, and if nbyte is zero and
24 the file is a regular file, the write() function may detect and return
25 errors as described below. In the absence of errors, or if error detec‐
26 tion is not performed, the write() function shall return zero and have
27 no other results. If nbyte is zero and the file is not a regular file,
28 the results are unspecified.
29 On a regular file or other file capable of seeking, the actual writing
31 of data shall proceed from the position in the file indicated by the
32 file offset associated with fildes. Before successful return from
33 write(), the file offset shall be incremented by the number of bytes
34 actually written. On a regular file, if this incremented file offset is
35 greater than the length of the file, the length of the file shall be
36 set to this file offset.
37 On a file not capable of seeking, writing shall always take place
39 starting at the current position. The value of a file offset associated
40 with such a device is undefined.
41 If the O_APPEND flag of the file status flags is set, the file offset
43 shall be set to the end of the file prior to each write and no inter‐
44 vening file modification operation shall occur between changing the
45 file offset and the write operation.
46 If a write() requests that more bytes be written than there is room for
48 (for example, the process' file size limit or the physical end of a
49 medium), only as many bytes as there is room for shall be written. For
50 example, suppose there is space for 20 bytes more in a file before
51 reaching a limit. A write of 512 bytes will return 20. The next write
52 of a non-zero number of bytes would give a failure return (except as
53 noted below).
54 If the request would cause the file size to exceed the soft file size
56 limit for the process and there is no room for any bytes to be written,
57 the request shall fail and the implementation shall generate the
58 SIGXFSZ signal for the thread.
59 If write() is interrupted by a signal before it writes any data, it
61 shall return -1 with errno set to [EINTR].
62 If write() is interrupted by a signal after it successfully writes some
64 data, it shall return the number of bytes written.
65 If the value of nbyte is greater than {SSIZE_MAX}, the result is imple‐
67 mentation-defined.
68 After a write() to a regular file has successfully returned:
70 * Any successful read() from each byte position in the file that was
72 modified by that write shall return the data specified by the
73 write() for that position until such byte positions are again modi‐
74 fied.
75 * Any subsequent successful write() to the same byte position in the
77 file shall overwrite that file data.
78 Write requests to a pipe or FIFO shall be handled in the same way as a
80 regular file with the following exceptions:
81 * There is no file offset associated with a pipe, hence each write
83 request shall append to the end of the pipe.
84 * Write requests of {PIPE_BUF} bytes or less shall not be interleaved
86 with data from other processes doing writes on the same pipe. Writes
87 of greater than {PIPE_BUF} bytes may have data interleaved, on arbi‐
88 trary boundaries, with writes by other processes, whether or not the
89 O_NONBLOCK flag of the file status flags is set.
90 * If the O_NONBLOCK flag is clear, a write request may cause the
92 thread to block, but on normal completion it shall return nbyte.
93 * If the O_NONBLOCK flag is set, write() requests shall be handled
95 differently, in the following ways:
96 * The write() function shall not block the thread.
98 * A write request for {PIPE_BUF} or fewer bytes shall have the fol‐
100 lowing effect: if there is sufficient space available in the
101 pipe, write() shall transfer all the data and return the number
102 of bytes requested. Otherwise, write() shall transfer no data and
103 return -1 with errno set to [EAGAIN].
104 * A write request for more than {PIPE_BUF} bytes shall cause one of
106 the following:
107 * When at least one byte can be written, transfer what it can
109 and return the number of bytes written. When all data previ‐
110 ously written to the pipe is read, it shall transfer at least
111 {PIPE_BUF} bytes.
112 * When no data can be written, transfer no data, and return -1
114 with errno set to [EAGAIN].
115 When attempting to write to a file descriptor (other than a pipe or
117 FIFO) that supports non-blocking writes and cannot accept the data
118 immediately:
119 * If the O_NONBLOCK flag is clear, write() shall block the calling
121 thread until the data can be accepted.
122 * If the O_NONBLOCK flag is set, write() shall not block the thread.
124 If some data can be written without blocking the thread, write()
125 shall write what it can and return the number of bytes written. Oth‐
126 erwise, it shall return -1 and set errno to [EAGAIN].
127 Upon successful completion, where nbyte is greater than 0, write()
129 shall mark for update the st_ctime and st_mtime fields of the file, and
130 if the file is a regular file, the S_ISUID and S_ISGID bits of the file
131 mode may be cleared.
132 For regular files, no data transfer shall occur past the offset maximum
134 established in the open file description associated with fildes.
135 If fildes refers to a socket, write() shall be equivalent to send()
137 with no flags set.
138 If the O_DSYNC bit has been set, write I/O operations on the file
140 descriptor shall complete as defined by synchronized I/O data integrity
141 completion.
142 If the O_SYNC bit has been set, write I/O operations on the file
144 descriptor shall complete as defined by synchronized I/O file integrity
145 completion.
146 If fildes refers to a shared memory object, the result of the write()
148 function is unspecified.
149 If fildes refers to a typed memory object, the result of the write()
151 function is unspecified.
152 If fildes refers to a STREAM, the operation of write() shall be deter‐
154 mined by the values of the minimum and maximum nbyte range (packet
155 size) accepted by the STREAM. These values are determined by the top‐
156 most STREAM module. If nbyte falls within the packet size range, nbyte
157 bytes shall be written. If nbyte does not fall within the range and
158 the minimum packet size value is 0, write() shall break the buffer into
159 maximum packet size segments prior to sending the data downstream (the
160 last segment may contain less than the maximum packet size). If nbyte
161 does not fall within the range and the minimum value is non-zero,
162 write() shall fail with errno set to [ERANGE]. Writing a zero-length
163 buffer ( nbyte is 0) to a STREAMS device sends 0 bytes with 0 returned.
164 However, writing a zero-length buffer to a STREAMS-based pipe or FIFO
165 sends no message and 0 is returned. The process may issue I_SWROPT
166 ioctl() to enable zero-length messages to be sent across the pipe or
167 FIFO.
168 When writing to a STREAM, data messages are created with a priority
170 band of 0. When writing to a STREAM that is not a pipe or FIFO:
171 * If O_NONBLOCK is clear, and the STREAM cannot accept data (the
173 STREAM write queue is full due to internal flow control conditions),
174 write() shall block until data can be accepted.
175 * If O_NONBLOCK is set and the STREAM cannot accept data, write()
177 shall return -1 and set errno to [EAGAIN].
178 * If O_NONBLOCK is set and part of the buffer has been written while a
180 condition in which the STREAM cannot accept additional data occurs,
181 write() shall terminate and return the number of bytes written.
182 In addition, write() shall fail if the STREAM head has processed an
184 asynchronous error before the call. In this case, the value of errno
185 does not reflect the result of write(), but reflects the prior error.
186 The pwrite() function shall be equivalent to write(), except that it
188 writes into a given position without changing the file pointer. The
189 first three arguments to pwrite() are the same as write() with the
190 addition of a fourth argument offset for the desired position inside
191 the file.
192
194 Upon successful completion, write() and pwrite() shall return the
195 number of bytes actually written to the file associated with fildes.
196 This number shall never be greater than nbyte. Otherwise, -1 shall be
197 returned and errno set to indicate the error.
198
200 The write() and pwrite() functions shall fail if:
201 EAGAIN The O_NONBLOCK flag is set for the file descriptor and the
203 thread would be delayed in the write() operation.
204 EBADF The fildes argument is not a valid file descriptor open for
206 writing.
207 EFBIG An attempt was made to write a file that exceeds the implementa‐
209 tion-defined maximum file size or the process' file size
210 limit, and there was no room for any bytes to be written.
211 EFBIG The file is a regular file, nbyte is greater than 0, and the
213 starting position is greater than or equal to the offset maximum
214 established in the open file description associated with fildes.
215 EINTR The write operation was terminated due to the receipt of a sig‐
217 nal, and no data was transferred.
218 EIO The process is a member of a background process group attempting
220 to write to its controlling terminal, TOSTOP is set, the process
221 is neither ignoring nor blocking SIGTTOU, and the process group
222 of the process is orphaned. This error may also be returned
223 under implementation-defined conditions.
224 ENOSPC There was no free space remaining on the device containing the
226 file.
227 EPIPE An attempt is made to write to a pipe or FIFO that is not open
229 for reading by any process, or that only has one end open. A
230 SIGPIPE signal shall also be sent to the thread.
231 ERANGE The transfer request size was outside the range supported by the
233 STREAMS file associated with fildes.
234 The write() function shall fail if:
237 EAGAIN or EWOULDBLOCK
239 The file descriptor is for a socket, is marked O_NONBLOCK, and
241 write would block.
242 ECONNRESET
244 A write was attempted on a socket that is not connected.
245 EPIPE A write was attempted on a socket that is shut down for writing,
247 or is no longer connected. In the latter case, if the socket is
248 of type SOCK_STREAM, the SIGPIPE signal is generated to the
249 calling process.
250 The write() and pwrite() functions may fail if:
253 EINVAL The STREAM or multiplexer referenced by fildes is linked
255 (directly or indirectly) downstream from a multiplexer.
256 EIO A physical I/O error has occurred.
258 ENOBUFS
260 Insufficient resources were available in the system to perform
261 the operation.
262 ENXIO A request was made of a nonexistent device, or the request was
264 outside the capabilities of the device.
265 ENXIO A hangup occurred on the STREAM being written to.
267 A write to a STREAMS file may fail if an error message has been
270 received at the STREAM head. In this case, errno is set to the value
271 included in the error message.
272 The write() function may fail if:
274 EACCES A write was attempted on a socket and the calling process does
276 not have appropriate privileges.
277 ENETDOWN
279 A write was attempted on a socket and the local network inter‐
280 face used to reach the destination is down.
281 ENETUNREACH
283 A write was attempted on a socket and no route to the network is
285 present.
286 The pwrite() function shall fail and the file pointer remain unchanged
289 if:
290 EINVAL The offset argument is invalid. The value is negative.
292 ESPIPE fildes is associated with a pipe or FIFO.
294 The following sections are informative.
297
299 Writing from a Buffer
300 The following example writes data from the buffer pointed to by buf to
301 the file associated with the file descriptor fd.
302 #include <sys/types.h>
305 #include <string.h>
306 ...
307 char buf[20];
308 size_t nbytes;
309 ssize_t bytes_written;
310 int fd;
311 ...
312 strcpy(buf, "This is a test\n");
313 nbytes = strlen(buf);
314 bytes_written = write(fd, buf, nbytes);
317 ...
318
320 None.
321
323 See also the RATIONALE section in read() .
324 An attempt to write to a pipe or FIFO has several major characteris‐
326 tics:
327 * Atomic/non-atomic: A write is atomic if the whole amount written in
329 one operation is not interleaved with data from any other process.
330 This is useful when there are multiple writers sending data to a
331 single reader. Applications need to know how large a write request
332 can be expected to be performed atomically. This maximum is called
333 {PIPE_BUF}. This volume of IEEE Std 1003.1-2001 does not say whether
334 write requests for more than {PIPE_BUF} bytes are atomic, but
335 requires that writes of {PIPE_BUF} or fewer bytes shall be atomic.
336 * Blocking/immediate: Blocking is only possible with O_NONBLOCK clear.
338 If there is enough space for all the data requested to be written
339 immediately, the implementation should do so. Otherwise, the process
340 may block; that is, pause until enough space is available for writ‐
341 ing. The effective size of a pipe or FIFO (the maximum amount that
342 can be written in one operation without blocking) may vary dynami‐
343 cally, depending on the implementation, so it is not possible to
344 specify a fixed value for it.
345 * Complete/partial/deferred: A write request:
347 int fildes;
350 size_t nbyte;
351 ssize_t ret;
352 char *buf;
353 ret = write(fildes, buf, nbyte);
356 may return:
358 Complete
360 ret=nbyte
361 Partial
363 ret<nbyte
364 This shall never happen if nbyte<= {PIPE_BUF}. If it does happen
366 (with nbyte> {PIPE_BUF}), this volume of IEEE Std 1003.1-2001
367 does not guarantee atomicity, even if ret<= {PIPE_BUF}, because
368 atomicity is guaranteed according to the amount requested, not
369 the amount written.
370 Deferred:
372 ret=-1, errno=[EAGAIN]
373 This error indicates that a later request may succeed. It does
375 not indicate that it shall succeed, even if nbyte<= {PIPE_BUF},
376 because if no process reads from the pipe or FIFO, the write
377 never succeeds. An application could usefully count the number
378 of times [EAGAIN] is caused by a particular value of nbyte>
379 {PIPE_BUF} and perhaps do later writes with a smaller value, on
380 the assumption that the effective size of the pipe may have
381 decreased.
382 Partial and deferred writes are only possible with O_NONBLOCK set.
385 The relations of these properties are shown in the following tables:
387 Write to a Pipe or FIFO with O_NONBLOCK clear
389 Immediately Writable: None Some nbyte
390 nbyte<={PIPE_BUF} Atomic blocking Atomic blocking Atomic immediate
391 nbyte nbyte nbyte
392 nbyte>{PIPE_BUF} Blocking nbyte Blocking nbyte Blocking nbyte
393 If the O_NONBLOCK flag is clear, a write request shall block if the
395 amount writable immediately is less than that requested. If the flag is
396 set (by fcntl()), a write request shall never block.
397 Write to a Pipe or FIFO with O_NONBLOCK set
399 Immediately Writable: None Some nbyte
400 nbyte<={PIPE_BUF} -1, [EAGAIN] -1, [EAGAIN] Atomic nbyte
401 nbyte>{PIPE_BUF} -1, [EAGAIN] <nbyte or -1, <=nbyte or -1,
402 [EAGAIN] [EAGAIN]
403 There is no exception regarding partial writes when O_NONBLOCK is set.
405 With the exception of writing to an empty pipe, this volume of
406 IEEE Std 1003.1-2001 does not specify exactly when a partial write is
407 performed since that would require specifying internal details of the
408 implementation. Every application should be prepared to handle partial
409 writes when O_NONBLOCK is set and the requested amount is greater than
410 {PIPE_BUF}, just as every application should be prepared to handle par‐
411 tial writes on other kinds of file descriptors.
412 The intent of forcing writing at least one byte if any can be written
414 is to assure that each write makes progress if there is any room in the
415 pipe. If the pipe is empty, {PIPE_BUF} bytes must be written; if not,
416 at least some progress must have been made.
417 Where this volume of IEEE Std 1003.1-2001 requires -1 to be returned
419 and errno set to [EAGAIN], most historical implementations return zero
420 (with the O_NDELAY flag set, which is the historical predecessor of
421 O_NONBLOCK, but is not itself in this volume of IEEE Std 1003.1-2001).
422 The error indications in this volume of IEEE Std 1003.1-2001 were cho‐
423 sen so that an application can distinguish these cases from end-of-
424 file. While write() cannot receive an indication of end-of-file,
425 read() can, and the two functions have similar return values. Also,
426 some existing systems (for example, Eighth Edition) permit a write of
427 zero bytes to mean that the reader should get an end-of-file indica‐
428 tion; for those systems, a return value of zero from write() indicates
429 a successful write of an end-of-file indication.
430 Implementations are allowed, but not required, to perform error check‐
432 ing for write() requests of zero bytes.
433 The concept of a {PIPE_MAX} limit (indicating the maximum number of
435 bytes that can be written to a pipe in a single operation) was consid‐
436 ered, but rejected, because this concept would unnecessarily limit
437 application writing.
438 See also the discussion of O_NONBLOCK in read() .
440 Writes can be serialized with respect to other reads and writes. If a
442 read() of file data can be proven (by any means) to occur after a
443 write() of the data, it must reflect that write(), even if the calls
444 are made by different processes. A similar requirement applies to mul‐
445 tiple write operations to the same file position. This is needed to
446 guarantee the propagation of data from write() calls to subsequent
447 read() calls. This requirement is particularly significant for net‐
448 worked file systems, where some caching schemes violate these seman‐
449 tics.
450 Note that this is specified in terms of read() and write(). The XSI
452 extensions readv() and writev() also obey these semantics. A new "high-
453 performance" write analog that did not follow these serialization
454 requirements would also be permitted by this wording. This volume of
455 IEEE Std 1003.1-2001 is also silent about any effects of application-
456 level caching (such as that done by stdio).
457 This volume of IEEE Std 1003.1-2001 does not specify the value of the
459 file offset after an error is returned; there are too many cases. For
460 programming errors, such as [EBADF], the concept is meaningless since
461 no file is involved. For errors that are detected immediately, such as
462 [EAGAIN], clearly the pointer should not change. After an interrupt or
463 hardware error, however, an updated value would be very useful and is
464 the behavior of many implementations.
465 This volume of IEEE Std 1003.1-2001 does not specify behavior of con‐
467 current writes to a file from multiple processes. Applications should
468 use some form of concurrency control.
469
471 None.
472
474 chmod() , creat() , dup() , fcntl() , getrlimit() , lseek() , open() ,
475 pipe() , ulimit() , writev() , the Base Definitions volume of
476 IEEE Std 1003.1-2001, <limits.h>, <stropts.h>, <sys/uio.h>, <unistd.h>
477
479 Portions of this text are reprinted and reproduced in electronic form
480 from IEEE Std 1003.1, 2003 Edition, Standard for Information Technology
481 -- Portable Operating System Interface (POSIX), The Open Group Base
482 Specifications Issue 6, Copyright (C) 2001-2003 by the Institute of
483 Electrical and Electronics Engineers, Inc and The Open Group. In the
484 event of any discrepancy between this version and the original IEEE and
485 The Open Group Standard, the original IEEE and The Open Group Standard
486 is the referee document. The original Standard can be obtained online
487 at http://www.opengroup.org/unix/online.html .
488IEEE/The Open Group 2003 WRITE(P)