1 .\" Copyright (C) 2014 Michael Kerrisk <mtk.manpages@gmail.com>
2 .\" and Copyright (C) 2014 David Herrmann <dh.herrmann@gmail.com>
4 .\" %%%LICENSE_START(GPLv2+)
6 .\" This program is free software; you can redistribute it and/or modify
7 .\" it under the terms of the GNU General Public License as published by
8 .\" the Free Software Foundation; either version 2 of the License, or
9 .\" (at your option) any later version.
11 .\" This program is distributed in the hope that it will be useful,
12 .\" but WITHOUT ANY WARRANTY; without even the implied warranty of
13 .\" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 .\" GNU General Public License for more details.
16 .\" You should have received a copy of the GNU General Public
17 .\" License along with this manual; if not, see
18 .\" <http://www.gnu.org/licenses/>.
21 .TH MEMFD_CREATE 2 2018-02-02 Linux "Linux Programmer's Manual"
23 memfd_create \- create an anonymous file
25 .B #include <sys/mman.h>
27 .BI "int memfd_create(const char *" name ", unsigned int " flags ");"
30 creates an anonymous file and returns a file descriptor that refers to it.
31 The file behaves like a regular file, and so can be modified,
32 truncated, memory-mapped, and so on.
33 However, unlike a regular file,
34 it lives in RAM and has a volatile backing storage.
35 Once all references to the file are dropped, it is automatically released.
36 Anonymous memory is used for all backing pages of the file.
37 Therefore, files created by
39 have the same semantics as other anonymous
41 .\" memfd uses VM_NORESERVE so each page is accounted on first access.
42 .\" This means, the overcommit-limits (see __vm_enough_memory()) and the
43 .\" memory-cgroup limits (mem_cgroup_try_charge()) are applied. Note that
44 .\" those are accounted on "current" and "current->mm", that is, the
45 .\" process doing the first page access.
46 memory allocations such as those allocated using
52 The initial size of the file is set to 0.
53 Following the call, the file size should be set using
55 (Alternatively, the file may be populated by calls to
61 is used as a filename and will be displayed
62 as the target of the corresponding symbolic link in the directory
64 The displayed name is always prefixed with
66 and serves only for debugging purposes.
67 Names do not affect the behavior of the file descriptor,
68 and as such multiple files can have the same name without any side effects.
70 The following values may be bitwise ORed in
72 to change the behavior of
78 flag on the new file descriptor.
79 See the description of the
83 for reasons why this may be useful.
86 Allow sealing operations on this file.
87 See the discussion of the
93 and also NOTES, below.
94 The initial set of seals is empty.
95 If this flag is not set, the initial set of seals will be
97 meaning that no other seals can be set on the file.
98 .\" FIXME Why is the MFD_ALLOW_SEALING behavior not simply the default?
99 .\" Is it worth adding some text explaining this?
101 .BR MFD_HUGETLB " (since Linux 4.14)"
102 .\" commit 749df87bd7bee5a79cef073f5d032ddb2b211de8
103 The anonymous file will be created in the hugetlbfs filesystem using
105 See the Linux kernel source file
106 .I Documentation/vm/hugetlbpage.txt
107 for more information about hugetlbfs.
108 The hugetlbfs filesystem does not support file-sealing operations.
109 Therefore, specifying both
117 .BR MFD_HUGE_2MB ", " MFD_HUGE_1GB ", " "..."
118 Used in conjunction with
120 to select alternative hugetlb page sizes (respectively, 2\ MB, 1\ GB, ...)
121 on systems that support multiple hugetlb page sizes.
122 Definitions for known
123 huge page sizes are included in the header file
126 For details on encoding huge page sizes not included in the header file,
127 see the discussion of the similarly named constants in
136 returns a new file descriptor that can be used to refer to the file.
137 This file descriptor is opened for both reading and writing
141 is set for the file descriptor.
147 the usual semantics apply for the file descriptor created by
149 A copy of the file descriptor is inherited by the child produced by
151 and refers to the same file.
152 The file descriptor is preserved across
154 unless the close-on-exec flag has been set.
158 returns a new file descriptor.
159 On error, \-1 is returned and
161 is set to indicate the error.
167 points to invalid memory.
171 included unknown bits.
177 .\" NAME_MAX - strlen("memfd:")
178 249 bytes, excluding the terminating null byte.)
189 The per-process limit on the number of open file descriptors has been reached.
192 The system-wide limit on the total number of open files has been reached.
195 There was insufficient memory to create a new anonymous file.
199 system call first appeared in Linux 3.17;
200 glibc support was added in version 2.27.
204 system call is Linux-specific.
207 .\" See also http://lwn.net/Articles/593918/
208 .\" and http://lwn.net/Articles/594919/ and http://lwn.net/Articles/591108/
211 system call provides a simple alternative to manually mounting a
213 filesystem and creating and opening a file in that filesystem.
214 The primary purpose of
216 is to create files and associated file descriptors that are
217 used with the file-sealing APIs provided by
222 system call also has uses without file sealing
223 (which is why file-sealing is disabled, unless explicitly requested with the
224 .BR MFD_ALLOW_SEALING
226 In particular, it can be used as an alternative to creating files in
228 or as an alternative to using the
231 in cases where there is no intention to actually link the
232 resulting file into the filesystem.
234 In the absence of file sealing,
235 processes that communicate via shared memory must either trust each other,
236 or take measures to deal with the possibility that an untrusted peer
237 may manipulate the shared memory region in problematic ways.
238 For example, an untrusted peer might modify the contents of the
239 shared memory at any time, or shrink the shared memory region.
240 The former possibility leaves the local process vulnerable to
241 time-of-check-to-time-of-use race conditions
242 (typically dealt with by copying data from
243 the shared memory region before checking and using it).
244 The latter possibility leaves the local process vulnerable to
246 signals when an attempt is made to access a now-nonexistent
247 location in the shared memory region.
248 (Dealing with this possibility necessitates the use of a handler for the
252 Dealing with untrusted peers imposes extra complexity on
253 code that employs shared memory.
254 Memory sealing enables that extra complexity to be eliminated,
255 by allowing a process to operate secure in the knowledge that
256 its peer can't modify the shared memory in an undesired fashion.
258 An example of the usage of the sealing mechanism is as follows:
260 The first process creates a
264 The call yields a file descriptor used in subsequent steps.
267 sizes the file created in the previous step using
271 and populates the shared memory with the desired data.
273 The first process uses the
276 operation to place one or more seals on the file,
277 in order to restrict further modifications on the file.
280 then it will be necessary to first unmap the shared writable mapping
281 created in the previous step.)
283 A second process obtains a file descriptor for the
286 Among the possible ways in which this could happen are the following:
289 The process that called
291 could transfer the resulting file descriptor to the second process
292 via a UNIX domain socket (see
296 The second process then maps the file using
299 The second process is created via
301 and thus automatically inherits the file descriptor and mapping.
302 (Note that in this case and the next,
303 there is a natural trust relationship between the two processes,
304 since they are running under the same user ID.
305 Therefore, file sealing would not normally be necessary.)
307 The second process opens the file
308 .IR /proc/<pid>/fd/<fd> ,
311 is the PID of the first process (the one that called
312 .BR memfd_create ()),
315 is the number of the file descriptor returned by the call to
318 The second process then maps the file using
322 The second process uses the
325 operation to retrieve the bit mask of seals
326 that has been applied to the file.
327 This bit mask can be inspected in order to determine
328 what kinds of restrictions have been placed on file modifications.
329 If desired, the second process can apply further seals
330 to impose additional restrictions (so long as the
332 seal has not yet been applied).
334 Below are shown two example programs that demonstrate the use of
336 and the file sealing API.
339 .IR t_memfd_create.c ,
344 sets a size for the file, maps it into memory,
345 and optionally places some seals on the file.
346 The program accepts up to three command-line arguments,
347 of which the first two are required.
348 The first argument is the name to associate with the file,
349 the second argument is the size to be set for the file,
350 and the optional third argument is a string of characters that specify
351 seals to be set on file.
355 can be used to open an existing file that was created via
357 and inspect the set of seals that have been applied to that file.
359 The following shell session demonstrates the use of these programs.
362 file and set some seals on it:
366 $ \fB./t_memfd_create my_memfd_file 4096 sw &\fP
368 PID: 11775; fd: 3; /proc/11775/fd/3
374 program continues to run in the background.
375 From another program, we can obtain a file descriptor for the
380 file that corresponds to the file descriptor opened by
382 Using that pathname, we inspect the content of the
384 symbolic link, and use our
386 program to view the seals that have been placed on the file:
390 $ \fBreadlink /proc/11775/fd/3\fP
391 /memfd:my_memfd_file (deleted)
392 $ \fB./t_get_seals /proc/11775/fd/3\fP
393 Existing seals: WRITE SHRINK
396 .SS Program source: t_memfd_create.c
399 #include <sys/mman.h>
406 #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \\
410 main(int argc, char *argv[])
415 char *name, *seals_arg;
419 fprintf(stderr, "%s name size [seals]\\n", argv[0]);
420 fprintf(stderr, "\\t\(aqseals\(aq can contain any of the "
421 "following characters:\\n");
422 fprintf(stderr, "\\t\\tg \- F_SEAL_GROW\\n");
423 fprintf(stderr, "\\t\\ts \- F_SEAL_SHRINK\\n");
424 fprintf(stderr, "\\t\\tw \- F_SEAL_WRITE\\n");
425 fprintf(stderr, "\\t\\tS \- F_SEAL_SEAL\\n");
433 /* Create an anonymous file in tmpfs; allow seals to be
434 placed on the file */
436 fd = memfd_create(name, MFD_ALLOW_SEALING);
438 errExit("memfd_create");
440 /* Size the file as specified on the command line */
442 if (ftruncate(fd, len) == \-1)
445 printf("PID: %ld; fd: %d; /proc/%ld/fd/%d\\n",
446 (long) getpid(), fd, (long) getpid(), fd);
448 /* Code to map the file and populate the mapping with data
451 /* If a \(aqseals\(aq command\-line argument was supplied, set some
454 if (seals_arg != NULL) {
457 if (strchr(seals_arg, \(aqg\(aq) != NULL)
458 seals |= F_SEAL_GROW;
459 if (strchr(seals_arg, \(aqs\(aq) != NULL)
460 seals |= F_SEAL_SHRINK;
461 if (strchr(seals_arg, \(aqw\(aq) != NULL)
462 seals |= F_SEAL_WRITE;
463 if (strchr(seals_arg, \(aqS\(aq) != NULL)
464 seals |= F_SEAL_SEAL;
466 if (fcntl(fd, F_ADD_SEALS, seals) == \-1)
470 /* Keep running, so that the file created by memfd_create()
471 continues to exist */
478 .SS Program source: t_get_seals.c
481 #include <sys/mman.h>
488 #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \\
492 main(int argc, char *argv[])
498 fprintf(stderr, "%s /proc/PID/fd/FD\\n", argv[0]);
502 fd = open(argv[1], O_RDWR);
506 seals = fcntl(fd, F_GET_SEALS);
510 printf("Existing seals:");
511 if (seals & F_SEAL_SEAL)
513 if (seals & F_SEAL_GROW)
515 if (seals & F_SEAL_WRITE)
517 if (seals & F_SEAL_SHRINK)
521 /* Code to map the file and access the contents of the
522 resulting mapping omitted */