xref: /csrg-svn/lib/libc/sys/mlock.2 (revision 61016)
1.\" Copyright (c) 1993 The Regents of the University of California.
2.\" All rights reserved.
3.\"
4.\" %sccs.include.redist.roff%
5.\"
6.\"	@(#)mlock.2	6.1 (Berkeley) 06/02/93
7.\"
8.Dd June 2, 1993
9.Dt MLOCK 2
10.Os
11.Sh NAME
12.Nm mlock ,
13.Nm munlock
14.Nd lock (unlock) physical pages in memory
15.Sh SYNOPSIS
16.Fd #include <sys/types.h>
17.Ft int
18.Fn mlock "caddr_t addr" "int len"
19.Ft int
20.Fn munlock "caddr_t addr" "int len"
21.Sh DESCRIPTION
22The
23.Nm mlock
24system call
25locks into memory the physical pages associated with the virtual address
26range starting at
27.Fa addr
28for
29.Fa len
30bytes.
31The
32.Nm munlock
33call reverses the effect of
34.Nm mlock .
35For both, the
36.Fa addr
37parameter should be aligned to a multiple of the page size.
38If the
39.Fa len
40parameter is not a multiple of the page size, it will be rounded up
41to be so.
42The entire range must be allocated.
43.Pp
44After an
45.Nm mlock
46call, the indicated pages will cause neither a non-resident page
47or address-translation fault until they are unlocked.
48They may still cause protection-violation faults or TLB-miss faults on
49architectures with software-managed TLBs.
50The physical pages remain in memory until all locked mappings for the pages
51are removed.
52Multiple processes may have the same physical pages locked via their own
53virtual address mappings.
54A single process may likewise have pages multiply-locked via different virtual
55mappings of the same pages or via nested
56.Nm mlock
57calls on the same address range.
58Unlocking is performed explicitly by
59.Nm munlock
60or implicitly by deallocation of the relevant address range.
61Locked mappings are not inherited by the child process after a
62.Xr fork 2 .
63.Pp
64Since physical memory is a potentially scarce resource, processes are
65limited in how much they can lock down.
66A single process can
67.Nm mlock
68the minimum of
69a system-wide ``wired pages'' limit and
70the per-process
71.Li RLIMIT_MEMLOCK
72resource limit.
73.Sh RETURN VALUES
74A return value of 0 indicates that the call
75succeeded and all pages in the range have either been locked or unlocked.
76A return value of -1 indicates an error occurred and the locked
77status of all pages in the range remains unchanged.
78In this case, the global location
79.Va errno
80is set to indicate the error.
81.Sh ERRORS
82.Fn Mlock
83and
84.Fn munlock
85will fail if:
86.Bl -tag -width Er
87.It Bq Er EINVAL
88The address given is not page aligned or the length is negative.
89.It Bq Er EAGAIN
90Locking the indicated range would exceed either the system or per-process
91limit for locked memory
92.Pq Nm mlock .
93.It Bq Er ENOMEM
94Some portion of the indicated address range is not allocated.
95There was an error faulting/mapping a page
96.Pq Nm mlock .
97Some portion of the indicated address range is not locked
98.Pq Nm munlock .
99.Sh "SEE ALSO"
100.Xr fork 2 ,
101.Xr setrlimit 2
102.Sh BUGS
103Unlike The Sun implementation, multiple
104.Nm mlock
105calls on the same address range require the corresponding number of
106.Nm munlock
107calls to actually unlock the pages, i.e.
108.Nm mlock
109nests.
110This should be considered a consequence of the implementation
111and not a feature.
112.Pp
113The per-process resource limit is a limit on the amount of virtual
114memory locked, while the system-wide limit is for the number of locked
115physical pages.
116Hence a process with two distinct locked mappings of the same physical page
117counts as 2 pages against the per-process limit and as only a single page
118in the system limit.
119.Sh HISTORY
120The
121.Fn mlock
122and
123.Fn munlock
124calls are
125.Ud .
126