1.\" $NetBSD: mremap.2,v 1.10 2023/10/25 18:23:10 riastradh Exp $ 2.\" 3.\" Copyright (c) 2007 Thomas Klausner and Joerg Sonnenberger 4.\" All rights reserved. 5.\" 6.\" Redistribution and use in source and binary forms, with or without 7.\" modification, are permitted provided that the following conditions 8.\" are met: 9.\" 1. Redistributions of source code must retain the above copyright 10.\" notice, this list of conditions and the following disclaimer. 11.\" 2. Redistributions in binary form must reproduce the above copyright 12.\" notice, this list of conditions and the following disclaimer in the 13.\" documentation and/or other materials provided with the distribution. 14.\" 15.\" THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND 16.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS BE LIABLE 19.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25.\" SUCH DAMAGE. 26.\" 27.\" ------------------------------------------------------------ 28.Dd May 2, 2018 29.Dt MREMAP 2 30.Os 31.Sh NAME 32.Nm mremap 33.Nd re-map a virtual memory address 34.\" ------------------------------------------------------------ 35.Sh LIBRARY 36.Lb libc 37.Sh SYNOPSIS 38.In sys/mman.h 39.Ft void * 40.Fn mremap "void *oldp" "size_t oldsize" "void *newp" "size_t newsize" \ 41"int flags" 42.\" ------------------------------------------------------------ 43.Sh DESCRIPTION 44The 45.Fn mremap 46function resizes the mapped range (see 47.Xr mmap 2 ) 48starting at 49.Ar oldp 50and having size 51.Ar oldsize 52to 53.Ar newsize . 54The following arguments can be 55.Dv OR Ap ed 56together in the 57.Ar flags 58argument: 59.Bl -tag -width XXMAPXALIGNEDXnXX 60.It Dv MAP_ALIGNED Ns Pq Ar n 61The allocation should be aligned to the given boundary, i.e. ensure 62that the lowest 63.Ar n 64bits of the address are zero. 65The parameter 66.Ar n 67should be the base 2 logarithm of the desired alignment (e.g., to 68request alignment to 16K, use 14 as the value for n). 69The alignment must be equal to or greater than the platform's page 70size as returned by 71.Xr sysconf 3 72with the 73.Dv _SC_PAGESIZE 74request. 75.It Dv MAP_FIXED 76.Ar newp 77is tried and 78.Fn mremap 79fails if that address can't be used as new base address for the range. 80Otherwise, 81.Ar oldp 82and 83.Ar newp 84are used as hints for the position, factoring in the given alignment. 85.It Dv MAP_REMAPDUP 86Duplicate the mapping. 87Both address ranges reference the same pages, but can have different protection flags. 88.El 89.Sh RETURN VALUES 90.Fn mremap 91returns the new address or 92.Dv MAP_FAILED , 93if the remap failed. 94.Sh EXAMPLES 95The following example program creates a two mappings for the same 96memory range, one RW- and one R-X. 97.Bd -literal 98#include <sys/types.h> 99#include <sys/mman.h> 100 101#include <stdio.h> 102#include <string.h> 103#include <unistd.h> 104#include <stdlib.h> 105#include <err.h> 106 107static int 108return_1(void) 109{ 110 return 1; 111} 112 113static void 114return_1_end(void) 115{ 116} 117 118static int 119return_2(void) 120{ 121 return 2; 122} 123 124static void 125return_2_end(void) 126{ 127} 128 129int 130main(int argc, char *argv[]) 131{ 132 void *maprw, *maprx; 133 int rv; 134 size_t page = (size_t)sysconf(_SC_PAGESIZE); 135 136 // Create the first mapping that has no protections, but intended 137 // protections only 138 maprw = mmap(NULL, page, 139 PROT_MPROTECT(PROT_EXEC|PROT_WRITE|PROT_READ), 140 MAP_ANON|MAP_PRIVATE, -1, 0); 141 if (maprw == MAP_FAILED) 142 err(EXIT_FAILURE, "mmap failed"); 143 144 // Create the second mapping for the same physical space, which 145 // again has no protections. 146 maprx = mremap(maprw, page, NULL, page, MAP_REMAPDUP); 147 if (maprx == MAP_FAILED) 148 err(EXIT_FAILURE, "mremap failed"); 149 150 // Set the first mapping read/write 151 if (mprotect(maprw, page, PROT_READ|PROT_WRITE) == -1) 152 err(EXIT_FAILURE, "mprotect(rw) failed"); 153 154 // Set the second mapping read/execute 155 if (mprotect(maprx, page, PROT_READ|PROT_EXEC) == -1) 156 err(EXIT_FAILURE, "mprotect(rx) failed"); 157 158#define XS(a) (size_t)((uintptr_t)(a ## _end) - (uintptr_t)(a)) 159 160 // Copy and run the first function 161 memcpy(maprw, return_1, XS(return_1)); 162 __builtin___clear_cache(maprw, (void *)((uintptr_t)maprw + page)); 163 rv = ((int (*)(void))maprx)(); 164 printf("%d\en", rv); 165 166 // Copy and run the second function 167 memcpy(maprw, return_2, XS(return_2)); 168 __builtin___clear_cache(maprw, (void *)((uintptr_t)maprw + page)); 169 rv = ((int (*)(void))maprx)(); 170 printf("%d\en", rv); 171 172 return EXIT_SUCCESS; 173} 174.Ed 175.Sh COMPATIBILITY 176The semantics of 177.Fn mremap 178differ from the one provided by glibc on Linux in that the 179.Ar newp 180argument was added and a different set of 181.Ar flags 182are implemented. 183.Sh ERRORS 184The 185.Fn mremap 186function fails if: 187.Bl -tag -width Er 188.It Bq Er E2BIG 189A request to extend 190.Fa oldp 191failed because of address overflow. 192.It Bq Er EINVAL 193If 194.Fa oldp 195or 196.Fa newp 197are not page aligned, or 198.Fa oldsize 199or 200.Fa newsize 201are not a multiple of the page size, or if 202.Fa oldp + 203.Fa oldsize 204or 205.Fa newp + 206.Fa newsize 207wrap around, or if an invalid alignment was requested in the 208.Fa flags 209argument, or if the request was to extend 210.Fa oldp 211and the extension address space either did not fit, was already occupied, 212or had set permissions. 213.It Bq Er ENOENT 214If the 215.Fa oldp 216segmented was not already mapped. 217.It Bq Er ENOMEM 218If there was either no space to allocate/move memory, 219or if a fixed allocation was requested in the 220.Fa flags 221that could not be accommodated. 222.El 223.Sh SEE ALSO 224.Xr mmap 2 , 225.Xr munmap 2 226.Sh HISTORY 227The 228.Fn mremap 229system call appeared in 230.Nx 5.0 . 231It was based on the code that supports 232.Fn mremap 233compatibility for Linux binaries. 234.Sh BUGS 235If a mapping is created by 236.Xr mmap 2 237with 238.Dv MAP_PRIVATE 239and then duplicated by 240.Nm 241with 242.Dv MAP_REMAPDUP , 243calling 244.Xr fork 2 245destroys the coupling between the original and duplicate mapping. 246Workarounds: 247.Bl -dash 248.It 249Create the original mapping with 250.Dv MAP_SHARED . 251.It 252Use 253.Xr minherit 2 254with 255.Dv MAP_INHERIT_NONE 256on the writable mapping. 257.It 258Don't use 259.Xr fork 2 ; 260use 261.Xr posix_spawn 2 . 262.El 263.Pp 264See 265.Lk https://gnats.NetBSD.org/55177 "PR lib/55177" 266for details. 267