1MLOCK(2)                   Linux Programmer's Manual                  MLOCK(2)
2
3
4

NAME

6       mlock, munlock, mlockall, munlockall - lock and unlock memory
7

SYNOPSIS

9       #include <sys/mman.h>
10
11       int mlock(const void *addr, size_t len);
12       int munlock(const void *addr, size_t len);
13
14       int mlockall(int flags);
15       int munlockall(void);
16

DESCRIPTION

18       mlock()  and  mlockall()  respectively  lock part or all of the calling
19       process's virtual address space into RAM, preventing that  memory  from
20       being  paged  to the swap area.  munlock() and munlockall() perform the
21       converse operation, respectively unlocking part or all of  the  calling
22       process's virtual address space, so that pages in the specified virtual
23       address range may once more to be swapped out if required by the kernel
24       memory manager.  Memory locking and unlocking are performed in units of
25       whole pages.
26
27   mlock() and munlock()
28       mlock() locks pages in the address range starting at addr and  continu‐
29       ing  for  len  bytes.   All  pages that contain a part of the specified
30       address range are guaranteed to  be  resident  in  RAM  when  the  call
31       returns  successfully;  the  pages  are guaranteed to stay in RAM until
32       later unlocked.
33
34       munlock() unlocks pages in the address range starting at addr and  con‐
35       tinuing  for len bytes.  After this call, all pages that contain a part
36       of the specified memory range can be moved to external swap space again
37       by the kernel.
38
39   mlockall() and munlockall()
40       mlockall() locks all pages mapped into the address space of the calling
41       process.  This includes the pages of the code, data and stack  segment,
42       as well as shared libraries, user space kernel data, shared memory, and
43       memory-mapped files.  All mapped pages are guaranteed to be resident in
44       RAM  when  the  call  returns successfully; the pages are guaranteed to
45       stay in RAM until later unlocked.
46
47       The flags argument is constructed as the bitwise OR of one or  more  of
48       the following constants:
49
50       MCL_CURRENT Lock  all pages which are currently mapped into the address
51                   space of the process.
52
53       MCL_FUTURE  Lock all pages which will become mapped  into  the  address
54                   space  of  the  process  in the future.  These could be for
55                   instance new pages required by a growing heap and stack  as
56                   well as new memory mapped files or shared memory regions.
57
58       If  MCL_FUTURE  has  been  specified,  then  a later system call (e.g.,
59       mmap(2), sbrk(2), malloc(3)), may fail if it would cause the number  of
60       locked  bytes to exceed the permitted maximum (see below).  In the same
61       circumstances, stack growth may likewise fail:  the  kernel  will  deny
62       stack expansion and deliver a SIGSEGV signal to the process.
63
64       munlockall()  unlocks  all  pages  mapped into the address space of the
65       calling process.
66

RETURN VALUE

68       On success these system calls return 0.   On  error,  -1  is  returned,
69       errno is set appropriately, and no changes are made to any locks in the
70       address space of the process.
71

ERRORS

73       ENOMEM (Linux 2.6.9 and later) the caller had a non-zero RLIMIT_MEMLOCK
74              soft  resource  limit,  but  tried  to lock more memory than the
75              limit permitted.  This limit is not enforced if the  process  is
76              privileged (CAP_IPC_LOCK).
77
78       ENOMEM (Linux  2.4  and earlier) the calling process tried to lock more
79              than half of RAM.
80
81       EPERM  (Linux  2.6.9  and  later)  the  caller   was   not   privileged
82              (CAP_IPC_LOCK) and its RLIMIT_MEMLOCK soft resource limit was 0.
83
84       EPERM  (Linux  2.6.8  and earlier) The calling process has insufficient
85              privilege to call munlockall().  Under  Linux  the  CAP_IPC_LOCK
86              capability is required.
87
88       For mlock() and munlock():
89
90       EAGAIN Some or all of the specified address range could not be locked.
91
92       EINVAL len was negative.
93
94       EINVAL (Not on Linux) addr was not a multiple of the page size.
95
96       ENOMEM Some  of  the  specified  address  range  does not correspond to
97              mapped pages in the address space of the process.
98
99       For mlockall():
100
101       EINVAL Unknown flags were specified.
102
103       For munlockall():
104
105       EPERM  (Linux  2.6.8  and  earlier)  The  caller  was  not   privileged
106              (CAP_IPC_LOCK).
107

CONFORMING TO

109       POSIX.1-2001, SVr4.
110

AVAILABILITY

112       On  POSIX  systems  on  which  mlock()  and  munlock()  are  available,
113       _POSIX_MEMLOCK_RANGE is defined in <unistd.h> and the number  of  bytes
114       in  a page can be determined from the constant PAGESIZE (if defined) in
115       <limits.h> or by calling sysconf(_SC_PAGESIZE).
116
117       On POSIX systems on which mlockall() and  munlockall()  are  available,
118       _POSIX_MEMLOCK  is  defined  in  <unistd.h>  to a value greater than 0.
119       (See also sysconf(3).)
120

NOTES

122       Memory locking has two  main  applications:  real-time  algorithms  and
123       high-security  data  processing.  Real-time applications require deter‐
124       ministic timing, and, like scheduling, paging is  one  major  cause  of
125       unexpected  program execution delays.  Real-time applications will usu‐
126       ally also switch to a real-time scheduler  with  sched_setscheduler(2).
127       Cryptographic security software often handles critical bytes like pass‐
128       words or secret keys as data structures.  As a result of paging,  these
129       secrets could be transferred onto a persistent swap store medium, where
130       they might be accessible to the enemy long after the security  software
131       has  erased  the secrets in RAM and terminated.  (But be aware that the
132       suspend mode on laptops and some desktop computers will save a copy  of
133       the system's RAM to disk, regardless of memory locks.)
134
135       Real-time processes that are using mlockall() to prevent delays on page
136       faults should reserve enough locked stack  pages  before  entering  the
137       time-critical  section, so that no page fault can be caused by function
138       calls.  This can be achieved by calling a  function  that  allocates  a
139       sufficiently large automatic variable (an array) and writes to the mem‐
140       ory occupied by this array in order to touch these stack  pages.   This
141       way,  enough  pages will be mapped for the stack and can be locked into
142       RAM.  The dummy writes ensure that not even copy-on-write  page  faults
143       can occur in the critical section.
144
145       Memory  locks  are not inherited by a child created via fork(2) and are
146       automatically removed  (unlocked)  during  an  execve(2)  or  when  the
147       process terminates.
148
149       The  memory  lock  on  an address range is automatically removed if the
150       address range is unmapped via munmap(2).
151
152       Memory locks do not stack, that is, pages which have been  locked  sev‐
153       eral times by calls to mlock() or mlockall() will be unlocked by a sin‐
154       gle call to munlock() for the corresponding range or  by  munlockall().
155       Pages  which  are  mapped  to several locations or by several processes
156       stay locked into RAM as long as they are locked at least at  one  loca‐
157       tion or by at least one process.
158
159   Linux Notes
160       Under Linux, mlock() and munlock() automatically round addr down to the
161       nearest page boundary.  However, POSIX.1-2001 allows an  implementation
162       to  require  that addr is page aligned, so portable applications should
163       ensure this.
164
165   Limits and permissions
166       In Linux 2.6.8 and earlier, a process must be privileged (CAP_IPC_LOCK)
167       in  order  to  lock  memory  and the RLIMIT_MEMLOCK soft resource limit
168       defines a limit on how much memory the process may lock.
169
170       Since Linux 2.6.9, no limits are placed on the amount of memory that  a
171       privileged  process can lock and the RLIMIT_MEMLOCK soft resource limit
172       instead defines a limit on how much memory an unprivileged process  may
173       lock.
174

BUGS

176       In  the  2.4  series  Linux  kernels  up to and including 2.4.17, a bug
177       caused the mlockall() MCL_FUTURE flag to be inherited across a fork(2).
178       This was rectified in kernel 2.4.18.
179
180       Since  kernel 2.6.9, if a privileged process calls mlockall(MCL_FUTURE)
181       and later drops privileges (loses the CAP_IPC_LOCK capability  by,  for
182       example,  setting  its  effective UID to a non-zero value), then subse‐
183       quent memory allocations (e.g.,  mmap(2),  brk(2))  will  fail  if  the
184       RLIMIT_MEMLOCK resource limit is encountered.
185

SEE ALSO

187       mmap(2), setrlimit(2), shmctl(2), sysconf(3), capabilities(7)
188

COLOPHON

190       This  page  is  part of release 3.22 of the Linux man-pages project.  A
191       description of the project, and information about reporting  bugs,  can
192       be found at http://www.kernel.org/doc/man-pages/.
193
194
195
196Linux                             2008-09-25                          MLOCK(2)
Impressum