1MLOCK(2) Linux Programmer's Manual MLOCK(2)
2
6 mlock, munlock, mlockall, munlockall - lock and unlock memory
7
9 #include <sys/mman.h>
10 int mlock(const void *addr, size_t len);
12 int munlock(const void *addr, size_t len);
14 int mlockall(int flags);
16 int munlockall(void);
18
20 mlock() and mlockall() respectively lock part or all of the calling
21 process's virtual address space into RAM, preventing that memory from
22 being paged to the swap area. munlock() and munlockall() perform the
23 converse operation, respectively unlocking part or all of the calling
24 process's virtual address space, so that pages in the specified virtual
25 address range may once more to be swapped out if required by the kernel
26 memory manager. Memory locking and unlocking are performed in units of
27 whole pages.
28 mlock() and munlock()
30 mlock() locks pages in the address range starting at addr and continu‐
31 ing for len bytes. All pages that contain a part of the specified
32 address range are guaranteed to be resident in RAM when the call
33 returns successfully; the pages are guaranteed to stay in RAM until
34 later unlocked.
35 munlock() unlocks pages in the address range starting at addr and con‐
37 tinuing for len bytes. After this call, all pages that contain a part
38 of the specified memory range can be moved to external swap space again
39 by the kernel.
40 mlockall() and munlockall()
42 mlockall() locks all pages mapped into the address space of the calling
43 process. This includes the pages of the code, data and stack segment,
44 as well as shared libraries, user space kernel data, shared memory, and
45 memory-mapped files. All mapped pages are guaranteed to be resident in
46 RAM when the call returns successfully; the pages are guaranteed to
47 stay in RAM until later unlocked.
48 The flags argument is constructed as the bitwise OR of one or more of
50 the following constants:
51 MCL_CURRENT Lock all pages which are currently mapped into the address
53 space of the process.
54 MCL_FUTURE Lock all pages which will become mapped into the address
56 space of the process in the future. These could be for
57 instance new pages required by a growing heap and stack as
58 well as new memory mapped files or shared memory regions.
59 If MCL_FUTURE has been specified, then a later system call (e.g.,
61 mmap(2), sbrk(2), malloc(3)), may fail if it would cause the number of
62 locked bytes to exceed the permitted maximum (see below). In the same
63 circumstances, stack growth may likewise fail: the kernel will deny
64 stack expansion and deliver a SIGSEGV signal to the process.
65 munlockall() unlocks all pages mapped into the address space of the
67 calling process.
68
70 Memory locking has two main applications: real-time algorithms and
71 high-security data processing. Real-time applications require determin‐
72 istic timing, and, like scheduling, paging is one major cause of unex‐
73 pected program execution delays. Real-time applications will usually
74 also switch to a real-time scheduler with sched_setscheduler(2). Cryp‐
75 tographic security software often handles critical bytes like passwords
76 or secret keys as data structures. As a result of paging, these secrets
77 could be transferred onto a persistent swap store medium, where they
78 might be accessible to the enemy long after the security software has
79 erased the secrets in RAM and terminated. (But be aware that the sus‐
80 pend mode on laptops and some desktop computers will save a copy of the
81 system's RAM to disk, regardless of memory locks.)
82 Real-time processes that are using mlockall() to prevent delays on page
84 faults should reserve enough locked stack pages before entering the
85 time-critical section, so that no page fault can be caused by function
86 calls. This can be achieved by calling a function that allocates a
87 sufficiently large automatic variable (an array) and writes to the mem‐
88 ory occupied by this array in order to touch these stack pages. This
89 way, enough pages will be mapped for the stack and can be locked into
90 RAM. The dummy writes ensure that not even copy-on-write page faults
91 can occur in the critical section.
92 Memory locks are not inherited by a child created via fork(2) and are
94 automatically removed (unlocked) during an execve(2) or when the
95 process terminates.
96 The memory lock on an address range is automatically removed if the
98 address range is unmapped via munmap(2).
99 Memory locks do not stack, i.e., pages which have been locked several
101 times by calls to mlock() or mlockall() will be unlocked by a single
102 call to munlock() for the corresponding range or by munlockall().
103 Pages which are mapped to several locations or by several processes
104 stay locked into RAM as long as they are locked at least at one loca‐
105 tion or by at least one process.
106
108 Under Linux, mlock() and munlock() automatically round addr down to the
109 nearest page boundary. However, POSIX.1-2001 allows an implementation
110 to require that addr is page aligned, so portable applications should
111 ensure this.
112 Limits and permissions
114 In Linux 2.6.8 and earlier, a process must be privileged (CAP_IPC_LOCK)
115 in order to lock memory and the RLIMIT_MEMLOCK soft resource limit
116 defines a limit on how much memory the process may lock.
117 Since Linux 2.6.9, no limits are placed on the amount of memory that a
119 privileged process can lock and the RLIMIT_MEMLOCK soft resource limit
120 instead defines a limit on how much memory an unprivileged process may
121 lock.
122
124 On success these system calls return 0. On error, -1 is returned,
125 errno is set appropriately, and no changes are made to any locks in the
126 address space of the process.
127
129 ENOMEM (Linux 2.6.9 and later) the caller had a non-zero RLIMIT_MEMLOCK
130 soft resource limit, but tried to lock more memory than the
131 limit permitted. This limit is not enforced if the process is
132 privileged (CAP_IPC_LOCK).
133 ENOMEM (Linux 2.4 and earlier) the calling process tried to lock more
135 than half of RAM.
136 EPERM (Linux 2.6.9 and later) the caller was not privileged
138 (CAP_IPC_LOCK) and its RLIMIT_MEMLOCK soft resource limit was 0.
139 EPERM (Linux 2.6.8 and earlier) The calling process has insufficient
141 privilege to call munlockall(). Under Linux the CAP_IPC_LOCK
142 capability is required.
143 For mlock() and munlock():
145 EINVAL len was negative.
147 EINVAL (Not on Linux) addr was not a multiple of the page size.
149 ENOMEM Some of the specified address range does not correspond to
151 mapped pages in the address space of the process.
152 For mlockall():
154 EINVAL Unknown flags were specified.
156 For munlockall():
158 EPERM (Linux 2.6.8 and earlier) The caller was not privileged
160 (CAP_IPC_LOCK).
161
163 In the 2.4 series Linux kernels up to and including 2.4.17, a bug
164 caused the mlockall() MCL_FUTURE flag to be inherited across a fork(2).
165 This was rectified in kernel 2.4.18.
166 Since kernel 2.6.9, if a privileged process calls mlockall(MCL_FUTURE)
168 and later drops privileges (loses the CAP_IPC_LOCK capability by, for
169 example, setting its effective UID to a non-zero value), then subse‐
170 quent memory allocations (e.g., mmap(2), brk(2)) will fail if the
171 RLIMIT_MEMLOCK resource limit is encountered.
172
174 On POSIX systems on which mlock() and munlock() are available,
175 _POSIX_MEMLOCK_RANGE is defined in <unistd.h> and the number of bytes
176 in a page can be determined from the constant PAGESIZE (if defined) in
177 <limits.h> or by calling sysconf(_SC_PAGESIZE).
178 On POSIX systems on which mlockall() and munlockall() are available,
180 _POSIX_MEMLOCK is defined in <unistd.h> to a value greater than 0. (See
181 also sysconf(3).)
182
184 POSIX.1-2001, SVr4
185
187 mmap(2), shmctl(2), setrlimit(2), sysconf(3), capabilities(7)
188Linux 2.6.15 2006-02-04 MLOCK(2)