1MLOCK(2) Linux Programmer's Manual MLOCK(2)
2
6 mlock, mlock2, munlock, mlockall, munlockall - lock and unlock memory
7
9 #include <sys/mman.h>
10 int mlock(const void *addr, size_t len);
12 int mlock2(const void *addr, size_t len, int flags);
13 int munlock(const void *addr, size_t len);
14 int mlockall(int flags);
16 int munlockall(void);
17
19 mlock(), mlock2(), and mlockall() lock part or all of the calling
20 process's virtual address space into RAM, preventing that memory from
21 being paged to the swap area.
22 munlock() and munlockall() perform the converse operation, unlocking
24 part or all of the calling process's virtual address space, so that
25 pages in the specified virtual address range may once more to be
26 swapped out if required by the kernel memory manager.
27 Memory locking and unlocking are performed in units of whole pages.
29 mlock(), mlock2(), and munlock()
31 mlock() locks pages in the address range starting at addr and continu‐
32 ing for len bytes. All pages that contain a part of the specified ad‐
33 dress range are guaranteed to be resident in RAM when the call returns
34 successfully; the pages are guaranteed to stay in RAM until later un‐
35 locked.
36 mlock2() also locks pages in the specified range starting at addr and
38 continuing for len bytes. However, the state of the pages contained in
39 that range after the call returns successfully will depend on the value
40 in the flags argument.
41 The flags argument can be either 0 or the following constant:
43 MLOCK_ONFAULT
45 Lock pages that are currently resident and mark the entire range
46 so that the remaining nonresident pages are locked when they are
47 populated by a page fault.
48 If flags is 0, mlock2() behaves exactly the same as mlock().
50 munlock() unlocks pages in the address range starting at addr and con‐
52 tinuing for len bytes. After this call, all pages that contain a part
53 of the specified memory range can be moved to external swap space again
54 by the kernel.
55 mlockall() and munlockall()
57 mlockall() locks all pages mapped into the address space of the calling
58 process. This includes the pages of the code, data and stack segment,
59 as well as shared libraries, user space kernel data, shared memory, and
60 memory-mapped files. All mapped pages are guaranteed to be resident in
61 RAM when the call returns successfully; the pages are guaranteed to
62 stay in RAM until later unlocked.
63 The flags argument is constructed as the bitwise OR of one or more of
65 the following constants:
66 MCL_CURRENT
68 Lock all pages which are currently mapped into the address space
69 of the process.
70 MCL_FUTURE
72 Lock all pages which will become mapped into the address space
73 of the process in the future. These could be, for instance, new
74 pages required by a growing heap and stack as well as new mem‐
75 ory-mapped files or shared memory regions.
76 MCL_ONFAULT (since Linux 4.4)
78 Used together with MCL_CURRENT, MCL_FUTURE, or both. Mark all
79 current (with MCL_CURRENT) or future (with MCL_FUTURE) mappings
80 to lock pages when they are faulted in. When used with MCL_CUR‐
81 RENT, all present pages are locked, but mlockall() will not
82 fault in non-present pages. When used with MCL_FUTURE, all fu‐
83 ture mappings will be marked to lock pages when they are faulted
84 in, but they will not be populated by the lock when the mapping
85 is created. MCL_ONFAULT must be used with either MCL_CURRENT or
86 MCL_FUTURE or both.
87 If MCL_FUTURE has been specified, then a later system call (e.g.,
89 mmap(2), sbrk(2), malloc(3)), may fail if it would cause the number of
90 locked bytes to exceed the permitted maximum (see below). In the same
91 circumstances, stack growth may likewise fail: the kernel will deny
92 stack expansion and deliver a SIGSEGV signal to the process.
93 munlockall() unlocks all pages mapped into the address space of the
95 calling process.
96
98 On success, these system calls return 0. On error, -1 is returned, er‐
99 rno is set appropriately, and no changes are made to any locks in the
100 address space of the process.
101
103 ENOMEM (Linux 2.6.9 and later) the caller had a nonzero RLIMIT_MEMLOCK
104 soft resource limit, but tried to lock more memory than the
105 limit permitted. This limit is not enforced if the process is
106 privileged (CAP_IPC_LOCK).
107 ENOMEM (Linux 2.4 and earlier) the calling process tried to lock more
109 than half of RAM.
110 EPERM The caller is not privileged, but needs privilege (CAP_IPC_LOCK)
112 to perform the requested operation.
113 For mlock(), mlock2(), and munlock():
115 EAGAIN Some or all of the specified address range could not be locked.
117 EINVAL The result of the addition addr+len was less than addr (e.g.,
119 the addition may have resulted in an overflow).
120 EINVAL (Not on Linux) addr was not a multiple of the page size.
122 ENOMEM Some of the specified address range does not correspond to
124 mapped pages in the address space of the process.
125 ENOMEM Locking or unlocking a region would result in the total number
127 of mappings with distinct attributes (e.g., locked versus un‐
128 locked) exceeding the allowed maximum. (For example, unlocking
129 a range in the middle of a currently locked mapping would result
130 in three mappings: two locked mappings at each end and an un‐
131 locked mapping in the middle.)
132 For mlock2():
134 EINVAL Unknown flags were specified.
136 For mlockall():
138 EINVAL Unknown flags were specified or MCL_ONFAULT was specified with‐
140 out either MCL_FUTURE or MCL_CURRENT.
141 For munlockall():
143 EPERM (Linux 2.6.8 and earlier) The caller was not privileged
145 (CAP_IPC_LOCK).
146
148 mlock2() is available since Linux 4.4; glibc support was added in ver‐
149 sion 2.27.
150
152 POSIX.1-2001, POSIX.1-2008, SVr4.
153 mlock2() is Linux specific.
155 On POSIX systems on which mlock() and munlock() are available,
157 _POSIX_MEMLOCK_RANGE is defined in <unistd.h> and the number of bytes
158 in a page can be determined from the constant PAGESIZE (if defined) in
159 <limits.h> or by calling sysconf(_SC_PAGESIZE).
160 On POSIX systems on which mlockall() and munlockall() are available,
162 _POSIX_MEMLOCK is defined in <unistd.h> to a value greater than 0.
163 (See also sysconf(3).)
164
166 Memory locking has two main applications: real-time algorithms and
167 high-security data processing. Real-time applications require deter‐
168 ministic timing, and, like scheduling, paging is one major cause of un‐
169 expected program execution delays. Real-time applications will usually
170 also switch to a real-time scheduler with sched_setscheduler(2). Cryp‐
171 tographic security software often handles critical bytes like passwords
172 or secret keys as data structures. As a result of paging, these se‐
173 crets could be transferred onto a persistent swap store medium, where
174 they might be accessible to the enemy long after the security software
175 has erased the secrets in RAM and terminated. (But be aware that the
176 suspend mode on laptops and some desktop computers will save a copy of
177 the system's RAM to disk, regardless of memory locks.)
178 Real-time processes that are using mlockall() to prevent delays on page
180 faults should reserve enough locked stack pages before entering the
181 time-critical section, so that no page fault can be caused by function
182 calls. This can be achieved by calling a function that allocates a
183 sufficiently large automatic variable (an array) and writes to the mem‐
184 ory occupied by this array in order to touch these stack pages. This
185 way, enough pages will be mapped for the stack and can be locked into
186 RAM. The dummy writes ensure that not even copy-on-write page faults
187 can occur in the critical section.
188 Memory locks are not inherited by a child created via fork(2) and are
190 automatically removed (unlocked) during an execve(2) or when the
191 process terminates. The mlockall() MCL_FUTURE and MCL_FUTURE | MCL_ON‐
192 FAULT settings are not inherited by a child created via fork(2) and are
193 cleared during an execve(2).
194 Note that fork(2) will prepare the address space for a copy-on-write
196 operation. The consequence is that any write access that follows will
197 cause a page fault that in turn may cause high latencies for a real-
198 time process. Therefore, it is crucial not to invoke fork(2) after an
199 mlockall() or mlock() operation—not even from a thread which runs at a
200 low priority within a process which also has a thread running at ele‐
201 vated priority.
202 The memory lock on an address range is automatically removed if the ad‐
204 dress range is unmapped via munmap(2).
205 Memory locks do not stack, that is, pages which have been locked sev‐
207 eral times by calls to mlock(), mlock2(), or mlockall() will be un‐
208 locked by a single call to munlock() for the corresponding range or by
209 munlockall(). Pages which are mapped to several locations or by sev‐
210 eral processes stay locked into RAM as long as they are locked at least
211 at one location or by at least one process.
212 If a call to mlockall() which uses the MCL_FUTURE flag is followed by
214 another call that does not specify this flag, the changes made by the
215 MCL_FUTURE call will be lost.
216 The mlock2() MLOCK_ONFAULT flag and the mlockall() MCL_ONFAULT flag al‐
218 low efficient memory locking for applications that deal with large map‐
219 pings where only a (small) portion of pages in the mapping are touched.
220 In such cases, locking all of the pages in a mapping would incur a sig‐
221 nificant penalty for memory locking.
222 Linux notes
224 Under Linux, mlock(), mlock2(), and munlock() automatically round addr
225 down to the nearest page boundary. However, the POSIX.1 specification
226 of mlock() and munlock() allows an implementation to require that addr
227 is page aligned, so portable applications should ensure this.
228 The VmLck field of the Linux-specific /proc/[pid]/status file shows how
230 many kilobytes of memory the process with ID PID has locked using
231 mlock(), mlock2(), mlockall(), and mmap(2) MAP_LOCKED.
232 Limits and permissions
234 In Linux 2.6.8 and earlier, a process must be privileged (CAP_IPC_LOCK)
235 in order to lock memory and the RLIMIT_MEMLOCK soft resource limit de‐
236 fines a limit on how much memory the process may lock.
237 Since Linux 2.6.9, no limits are placed on the amount of memory that a
239 privileged process can lock and the RLIMIT_MEMLOCK soft resource limit
240 instead defines a limit on how much memory an unprivileged process may
241 lock.
242
244 In Linux 4.8 and earlier, a bug in the kernel's accounting of locked
245 memory for unprivileged processes (i.e., without CAP_IPC_LOCK) meant
246 that if the region specified by addr and len overlapped an existing
247 lock, then the already locked bytes in the overlapping region were
248 counted twice when checking against the limit. Such double accounting
249 could incorrectly calculate a "total locked memory" value for the
250 process that exceeded the RLIMIT_MEMLOCK limit, with the result that
251 mlock() and mlock2() would fail on requests that should have succeeded.
252 This bug was fixed in Linux 4.9.
253 In the 2.4 series Linux kernels up to and including 2.4.17, a bug
255 caused the mlockall() MCL_FUTURE flag to be inherited across a fork(2).
256 This was rectified in kernel 2.4.18.
257 Since kernel 2.6.9, if a privileged process calls mlockall(MCL_FUTURE)
259 and later drops privileges (loses the CAP_IPC_LOCK capability by, for
260 example, setting its effective UID to a nonzero value), then subsequent
261 memory allocations (e.g., mmap(2), brk(2)) will fail if the RLIMIT_MEM‐
262 LOCK resource limit is encountered.
263
265 mincore(2), mmap(2), setrlimit(2), shmctl(2), sysconf(3), proc(5), ca‐
266 pabilities(7)
267
269 This page is part of release 5.10 of the Linux man-pages project. A
270 description of the project, information about reporting bugs, and the
271 latest version of this page, can be found at
272 https://www.kernel.org/doc/man-pages/.
273Linux 2020-04-11 MLOCK(2)