mmap(2)

DESCRIPTION

The mmap() function establishes a mapping between a process' address space and a file. The format of the call is as follows:

The mmap() function establishes a mapping between the process' address space at an address pa for len bytes and the file associated with the file descriptor fildes at offset off for len bytes. The value of pa is an unspecified function of the argument addr and values of flags, further described below. A successful mmap() call returns pa as its result. The address ranges covered by [pa, pa+len] and [off, off+len] must be legitimate for the possible (not necessarily current) address space of the process and the file, respectively.

If the size of the mapped file changes after the call to mmap(), the effect of references to portions of the mapped region that correspond to added or removed portions of the file is unspecified.

The mmap() function is supported for regular files. Support for any other type of file is unspecified except for /dev/zero, which is supported in HP-UX. See also zero(7).

The prot argument determines whether read, write, execute, or some combination of accesses are permitted to the pages being mapped. The protection options are defined in <sys/mman.h>:

The prot argument can be PROT_NONE, or any combination of the bit-wise inclusive OR of the values of PROT_READ, PROT_WRITE, and PROT_EXEC. If PROT_NONE is not specified, the system may grant other access permissions to the region in addition to those explicitly requested, except that write access will not be granted unless PROT_WRITE is specified.

The following combinations of protection modes are supported:

The flags argument provides other information about the handling of the mapped pages. The options are defined in <sys/mman.h>:

The MAP_PRIVATE and MAP_SHARED flags control the visibility of write references to the memory region. Exactly one of these flags must be specified. The mapping type is retained across a fork().

If MAP_SHARED is set in flags, write references to the memory region by the calling process may change the file and are visible in all MAP_SHARED mappings of the same portion of the file by any process of the same executable type. That is, an application compiled as a 32-bit process will be able to share the same mappings with other 32-bit processes, and an application compiled as a 64-bit process will be able to share the same mappings with other 64-bit processes.

If a 64-bit and a 32-bit application want to share the same mapping, both MAP_ADDR32 and MAP_SHARED must be set in flags by the 64-bit application. The 32-bit application does not need to set MAP_ADDR32 in flags. When MAP_SHARED is set in flags, write references to the memory region by the calling process may change the file. Changes are visible in all 32-bit processes which specify MAP_SHARED and by all 64-bit processes which specify both MAP_SHARED and MAP_ADDR32 for the same portion of the file. Mappings created using the MAP_IO flag do not currently support usage of the MAP_ADDR32 flag. Applications may use MAP_IO with 32-bit and 64-bit processes, but the pa returned to a 32-bit process may differ from the pa returned to a 64-bit process.

On Itanium®-based platforms, 64-bit MPAS (Mostly Private Address Space) and MGAS (Mostly Global Address Space) processes can share between themselves without incurring the cost of aliasing by specifying the MAP_GLOBAL flag without specifying the MAP_ADDR32 flag. Such mappings are not accessible by 32-bit processes. See the Adaptive Address Space Whitepaper for details.

If MAP_PRIVATE is set in flags, write references to the memory region by the calling process do not change the file and are not visible to any process in other mappings of the same portion of the file.

It is unspecified whether write references by processes that have mapped the memory region using MAP_SHARED are visible to processes that have mapped the same portion of the file using MAP_PRIVATE.

The MAP_VARIABLE and MAP_FIXED flags control the placement of the region as described below. Exactly one of these flags must be selected, unless MAP_IO is set. MAP_IO implicitly defaults to using MAP_VARIABLE and does not support MAP_FIXED.

If MAP_VARIABLE is set in flags, and the requested address addr is NULL, or if it is not possible for the system to place the region at the requested address, the region is placed at an address selected by the system. If the requested address is not a multiple of the page size returned by sysconf(_SC_PAGE_SIZE), the system treats the address as if it were rounded up to the next larger page size multiple.

When MAP_FIXED is set in the flags argument, the implementation is informed that the value of pa must be addr, exactly. If MAP_FIXED is set, mmap() may return MAP_FAILED and set errno to EINVAL. If a MAP_FIXED request is successful, the mapping established by mmap() replaces any previous mappings for the process' pages in the range [pa, pa+len]. If mmap() fails for reasons other than EBADF or EINVAL, some of the mappings in the address range starting at addr and continuing for len bytes may have been unmapped.

When MAP_FIXED is not set, the implementation uses addr in an unspecified manner to arrive at pa. The pa so chosen will be an area of the address space which the implementation deems suitable for a mapping of len bytes to the file. All implementations interpret an addr value of 0 as granting the implementation complete freedom in selecting pa, subject to constraints described below. A non-zero value of addr is taken to be a suggestion of a process address near which the mapping should be placed. When the implementation selects a value for pa, it never places a mapping at address 0, replaces any extant mapping, or maps into dynamic memory allocation areas.

The off argument is constrained to be aligned and sized according to the value returned by sysconf() when passed _SC_PAGESIZE or _SC_PAGE_SIZE. When MAP_FIXED is specified, the argument addr must also meet these constraints. The implementation performs mapping operations over whole pages. Thus, while the argument len need not meet a size or alignment constraint, the implementation will include, in any unmapping operation, any partial page specified by the range [pa, pa+len].

The MAP_FILE and MAP_ANONYMOUS flags control whether the region to be mapped is a mapped file region or an anonymous shared memory region. Exactly one of these flags must be selected.

If MAP_FILE is set in flags, a new mapped file region is created, mapping the file associated with fildes. off specifies the file byte offset at which the mapping starts. This offset must be a multiple of the page size returned by sysconf(_SC_PAGE_SIZE). If the end of the mapped file region is beyond the end of the file, any reference to an address in the mapped file region corresponding to an offset beyond the end of the file results in the delivery of a SIGBUS signal to the process, unless the address lies in the last partial page corresponding to the range beyond the end of the file. The last partial page mapping the range beyond the end of the file is always initialized to zeros, and any modified portions of the last page of a file which are beyond the file's end are not written back to the file.

If MAP_ANONYMOUS is set in flags, a new memory region is created and initialized to all zeros. This memory region can be shared only with descendants of the current process. If the fildes argument is not -1, an EINVAL error is generated. This is not enforced in the current release, but will be enforced in the next release. The value of off is meaningless because there is no underlying file object for the memory region.

If MAP_NORESERVE is set in flags, no swap space is initially reserved for the private mapping. Without this flag, the creation of a MAP_PRIVATE region reserves swap space equal to the size of the mapping. When a page in the mapping is first modified (written to), a private page is created and the swap space reserved is used to hold the private copy of the data in the event of a page-out. An initial write into a page of a MAP_NORESERVE mapping produces results which depend on the current availability of system swap space, since the swap space reservation occurs at the time of the first write and only for the affected page. If the swap space reservation can be made for the page, the write succeeds as described above. If not, the write fails and a SIGBUS signal is posted to the writing process for the effective virtual address. madvise(...,MADV_DONTNEED) on a MAP_NORESERVE object will release swap space reservations for relevant pages.

Specifying MAP_GLOBAL in a call to mmap() forces the allocation to happen from the global quadrant as compared to the private quadrants. See setmemwindow(1M) for further details.

If MAP_IO is set in flags:

Note that the flags MAP_MEM_INTERLEAVED, MAP_MEM_LOCAL, and MAP_MEM_FIRST_TOUCH involve the placement of physical memory. This is important only on Cache Coherent Non-Uniform Memory Architecture (ccNUMA) systems. They will have no effect otherwise. For more information, see the mpctl(2) manual page. These flags are hints to the system. If memory of the desired type is not available, the next most suitable memory is returned instead.

mmap() cannot create a mapped file region unless the file descriptor used to map the file is open for reading. For a mapped file region that is mapped with MAP_SHARED, mmap() grants write access permission only if the file descriptor is open for writing. If a region was mapped with either MAP_PRIVATE or MAP_ANONYMOUS, mmap() grants all requested access permissions.

After the successful completion of mmap(), fildes can be closed without effect on the mapped region or on the contents of the mapped file. Each mapped region creates a file reference, similar to an open file descriptor, that prevents the file data from being deallocated.

If an enforcement-mode file lock is in effect for any range of a file, a call to mmap() to map any range of the file with access rights that would violate the lock fails. The msem_lock() and msem_unlock() semaphore interfaces can be used to coordinate shared access to a region created with the MAP_SHARED flag. The advisory locks of the lockf() or fcntl() interfaces have no effect on memory mapped access, but they can be used to coordinate shared access to a MAP_SHARED mapped file region.

After a call to fork(), the child process inherits all mapped regions with the same data and the same sharing and protection attributes as in the parent process. Each mapped file and anonymous memory region created with mmap() is unmapped upon process exit, and by a successful call to any of the exec functions.

MAP_NORESERVE attribute is inherited across a fork() call; at the time of the fork(), swap space for a mapping is reserved in the child only for dirtied private pages that currently exist in the parent. Thereafter, the child's mapping reservation policy is as described above.

A SIGBUS signal is delivered to a process when a write reference to a mapped file region would cause a file system error condition, such as, exceeding quota or file system space limits.

A SIGBUS or SIGSEGV signal is delivered to a process upon a write reference to a region without PROT_WRITE protection, or any reference to a region with PROT_NONE protection.

A call to mmap() with PROT_EXECUTE specified, but without PROT_WRITE specified for a MAP_SHARED|MAP_FILE mapping is treated by the system as the execution of the underlying file. This implies that such a call fails if the file is currently open for writing or mapped with MAP_SHARED|PROT_WRITE options by any process, and if the call succeeds, the file cannot be opened for writing or subsequently mapped with MAP_SHARED|PROT_WRITE options as long as such mappings are present. A file's status as an active executable file is determined only at the time of an exec(), exit(), or mmap() operation. mprotect() operations on a MAP_FILE|MAP_SHARED mapping have no effect on the underlying file's status as an active executable file.

The implementation always zero-fills any partial page at the end of a memory region. Further, the implementation never writes out any modified portions of the last page of a file that are beyond the end of the mapped portion of the file. If the mapping established by mmap() extends into pages beyond the page containing the last byte of the file, an application reference to any of the pages in the mapping that are beyond the last page results in the delivery of a SIGBUS or SIGSEGV signal. The mmap() function adds an extra reference to the file associated with the file descriptor fildes which is not removed by a subsequent close() on that file descriptor. This reference is removed when there are no more mappings to the file. The st_atime field of the mapped file may be marked for update at any time between the mmap() call and the corresponding munmap() call. The initial read or write reference to a mapped region will cause the file's st_atime field to be marked for update if it has not already been marked for update.

The st_ctime, st_mtime, and st_atime fields of a file that is mapped with MAP_SHARED and PROT_WRITE, will be marked for update at some point in the interval between a write reference to the mapped region and the next call to msync() with MS_ASYNC or MS_SYNC for that portion of the file by any process. If there is no such call, these fields may be marked for update at any time after a write reference if the underlying file is modified as a result.

There may be implementation-dependent limits on the number of memory regions that can be mapped (per process or per system). If such a limit is imposed, whether the number of memory regions that can be mapped by a process is decreased by the use of shmat() is implementation-dependent.

RETURN VALUE

Upon successful completion, mmap() returns the address, (pa), at which the mapping was placed. Otherwise, it returns MAP_FAILED (defined in <sys/mman.h>) and sets errno to indicate the error.

ERRORS

The mmap() function will fail if:

DEPENDENCIES

APPLICATION USAGE

Use of mmap() may reduce the amount of memory available to other memory allocation functions.

Use of MAP_FIXED may result in unspecified behavior later in the use of brk(), sbrk(), malloc(), and shmat(). The use of MAP_FIXED is discouraged, as it may prevent an implementation from making the most effective use of resources.

The application must ensure correct synchronization when using mmap() in conjunction with any other file access method, such as read() and write(), standard input/output, and shmat().

The mmap() function allows access to resources via address space manipulations, instead of read()/write(). Once a file is mapped, all a process has to do to access it is use the data at the address to which the file was mapped. Using pseudo-code to illustrate the way an existing program might be changed to use mmap(), the following:

becomes:

AUTHOR

mmap() was developed by HP, AT&T, and OSF.

SEE ALSO

setmemwindow(1M), creat64(2), exec(2), fcntl(2), fork(2), lockf(2), madvise(2), mpctl(2), mprotect(2), msem_init(2), msem_lock(2), msem_unlock(2), msync(2), munmap(2), open(2), pset_ctl(2), shmop(2), sysconf(2), truncate(2), mman(5), privileges(5), stat(5), zero(7).

Adaptive Address Space Whitepaper in http://www.hp.com/products1/unix/operating/infolibrary.

STANDARDS CONFORMANCE

mmap(): AES, SVID3