Thanks Lutz for willing to do this. Here they are (you might have to reduce the text size one notch in your browser to read this, due to the fixed width of the `code' box.):
Code: Select all
mmap(2) mmap(2)
NAME
mmap, mmap64 - map pages of memory
SYNOPSIS
#include <sys/types.h>
#include <sys/mman.h>
void *mmap(void *addr, size_t len, int prot, int flags, int
fd, off_t off);
void *mmap64(void *addr, size_t len, int prot, int flags,
int fd, off64_t off);
DESCRIPTION
The functions mmap and mmap64 establish a mapping between a
process's address space and a virtual memory object. The
format of the call is as follows:
pa = mmap(addr, len, prot, flags, fd, off);
mmap establishes a mapping between the process's address
space at an address pa for len bytes to the memory object
represented by the file descriptor fd at offset off for len
bytes. The value of pa is an implementation-dependent
function of the parameter 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 a process and the
object in question, respectively.
The only difference between mmap and mmap64 is that in
mmap64 the off parameter is 64 bits long, so the file
offset can be greater than 2 gigabytes. This is useful for
certain filesystem types that support such file offsets.
The mapping established by mmap replaces any previous
mappings for the process's pages in the range [pa, pa +
len).
The parameter prot determines whether read (load), write
(store), execute, or some combination of accesses are
permitted to the pages being mapped. The protection
options are defined in <sys/mman.h> as:
PROT_READ Page can be read.
PROT_WRITE Page can be written.
PROT_EXEC Page can be executed.
PROT_NONE Page can not be accessed.
Not all implementations literally provide all possible
combinations. PROT_WRITE is often implemented as
PROT_READ|PROT_WRITE and PROT_EXEC as PROT_READ|PROT_EXEC.
This is true for all SGI implementations. In particular,
MIPS processors do not support a separate execute
permission. Any page that can be read can be executed
from, even if PROT_EXEC is not specified. Instead, the
operating system uses PROT_EXEC as a flag to indicate it
may need to perform certain platform dependent functions
(such as cache flushing) that may be needed to properly
execute instructions from the associated page. See
mprotect(2) for further details. However, no
implementation will permit a store to succeed where
PROT_WRITE has not been set. The behavior of PROT_WRITE
can be influenced by setting MAP_PRIVATE in the flags
parameter, described below.
The parameter flags provides other information about the
handling of the mapped pages. The options are defined in
<sys/mman.h> as:
MAP_SHARED Share changes
MAP_PRIVATE Changes are private
MAP_FIXED Interpret addr exactly
MAP_AUTOGROW Implicitly grow object
MAP_LOCAL Do not share with share group
MAP_AUTORESRV Reserve logical swap on demand
MAP_SGI_ANYADDR Use reserved area for mappings
MAP_SHARED and MAP_PRIVATE describe the disposition of
store references to the memory object. If MAP_SHARED is
specified, store references will change the memory object.
If MAP_PRIVATE is specified, the initial store reference
will create a private copy of the memory object page and
redirect the mapping to the copy. Either MAP_SHARED or
MAP_PRIVATE must be specified, but not both. The mapping
type is retained across a fork(2).
When MAP_SHARED is specified, and initially in all pages
when MAP_PRIVATE is specified, the contents of the mapped
segment change to reflect changes in the underlying memory
object. Changes can be caused by other processes that map
the same object with MAP_SHARED, or by processes using
write(2) or ftruncate(2). If the file is shortened, an
attempt to access a page of memory that is mapped to a part
of the file that no longer exists will cause a Bus Error
(SIGBUS) signal.
When MAP_PRIVATE is used, a private copy of a page is
created only when the process stores into the page. This
prevents changes from being seen by other processes that
map the same object, and prevents further changes made by
other processes from being visible. However, changes that
occur before the page is stored into are visible.
To protect the contents of a mapped file from changes or
truncation you can either use chmod(2) and lockf(3) to
enforce a mandatory file lock, or you can specify
MAP_PRIVATE and store into every page of the segment in
order to create a complete private copy of the data.
MAP_FIXED informs the system that the value of pa must be
addr, exactly. When MAP_FIXED is not set, the system uses
addr in an implementation- specific manner to arrive at pa.
The pa so chosen will be an area of the address space which
the system deems suitable for a mapping of len bytes to the
specified object.
All implementations interpret an addr value of zero as
granting the system 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 system
selects a value for pa, it will never place a mapping at
address 0, nor will it replace any extant mapping, and it
will attempt to map away from areas considered part of the
potential data or stack segments.
The MAP_FIXED directive should be used with caution. When
MAP_FIXED is set, any mappings (including text, heap, data,
and stack) in the range [addr, addr + len) will be replaced
with the new mapping.
To ensure best system hardware cache behavior, objects
should be mapped such that the low sixteen bits of the file
offset of the object match the low bits of the mapped
address.
The address range from 0x30000000 to 0x40000000 is normally
reserved for MAP_FIXED mappings except when MAP_SGI_ANYADDR
is specified or when the SGI_UNSUPPORTED_MAP_RESERVED_RANGE
option is enabled with syssgi(2). Note that this space may
not be useful for programs which require a very large heap,
since by default program heaps start near 0x10000000 and
grow toward higher addresses. This range will never be
used when zero is passed as the value for addr unless
either MAP_SGI_ANYADDR or
SGI_UNSUPPORTED_MAP_RESERVED_RANGE is used. See
sgi_use_anyaddr(1) for further details.
If MAP_AUTOGROW is specified with MAP_SHARED, the mapped
object will be implicitly grown when referenced by a store
operation to a page which maps beyond the current end of
the object; the object will be grown and zero-filled to
fulfill the mapping up to the next page boundary or to the
end of the mapping, whichever is less. If used with
MAP_PRIVATE, MAP_AUTOGROW allocates private zero-filled
pages for references beyond the end of the object, but does
not grow the object.
MAP_AUTOGROW requires that the object is mapped with
PROT_WRITE permission. Load references to mapped pages
following the end of a object will result in the delivery
of a SIGSEGV signal, as will various filesystem conditions
on stores. Whenever a SIGSEGV signal is delivered, the
second argument to the signal handler contains a value that
indicates the reason for the delivery of the signal; these
values are defined in /usr/include/sys/errno.h.
When MAP_AUTOGROW is specified, len determines the maximum
size of the memory map, as opposed to the initial size.
The size of the map is extended automatically by storing to
any location beyond the current working size, up to the
size limited by len. The mapped file will also be extended,
even if it is closed. Any read or write beyond the end of
the len specified area results in delivery of a SIGSEGV
signal. Therefore the use of MAP_AUTOGROW must anticipate
the maximum len as to prevent failed accesses beyond the
end of the mmapped area.
If MAP_LOCAL is used and the process does an sproc(2) each
process will receive a private copy of the object's
mapping. All subsequent load reference of objects mapped
MAP_PRIVATE will cause private copies of the object to be
created. In addition, the share group processes will be
able to independently unmap the object from their address
spaces.
The system reserves len bytes of logical swap space when
MAP_PRIVATE mappings of regular files are created, as well
as for all mappings of /dev/zero. (See swap(1m) for a
discussion of logical swap space.) If insufficient logical
swap space is available, mmap fails with EAGAIN. The
MAP_AUTORESRV flag causes logical swap space to be
automatically reserved as each page is first referenced
with a store operation instead of when the mapping is
created. When this flag is used, no logical swap space is
reserved when the mapping is created. Therefore, the
system cannot guarantee that space will be available when
needed. If all the logical swap space has been taken by
other processes when a page in a MAP_AUTORESRV mapping is
first stored to, then the process will be sent SIGBUS.
The parameter off is constrained to be aligned and sized
according to the value returned by getpagesize(2) or
sysconf(_SC_PAGESIZE). When MAP_FIXED is specified, the
parameter addr as well as off must be aligned according to
the value returned by sysconf(_SC_MMAP_FIXED_ALIGNMENT).
The system performs mapping operations over whole pages.
Thus, while the parameter len need not meet a size or
alignment constraint, the system will include, in any
mapping operation, any partial page specified by the range
[pa, pa + len).
The system will always zero-fill any partial page at the
end of an object. Further, the system will never write out
any modified portions of the last page of an object which
are beyond its end. References to whole pages following
the end of an object will result in the delivery of a
SIGBUS signal. SIGBUS signals may also be delivered on
various filesystem conditions, including quota exceeded
errors, and for physical device errors (such as unreadable
disk blocks). The signal handler may examine the si_code
and si_errno fields of the siginfo structure for
information about the nature of the error.
RETURN VALUE
On success, mmap returns the address at which the mapping
was placed (pa). On failure it returns MAP_FAILED and sets
errno to indicate an error.
ERRORS
Under the following conditions, mmap fails and sets errno
to:
EAGAIN The mapping could not be locked in memory.
EAGAIN The amount of logical swap space required is
temporarily unavailable.
EBADF fd is not open.
EACCES fd is not open for read, regardless of the
protection specified, or fd is not open for write
and PROT_WRITE was specified for a MAP_SHARED type
mapping.
EACCES prot has extraneous bits set.
EINVAL The arguments addr (if MAP_FIXED was specified) or
off are not multiples of the value returned by
sysconf(_SC_MMAP_FIXED_ALIGNMENT).
EINVAL The arguments flags is invalid (neither MAP_PRIVATE
or MAP_SHARED).
EINVAL The argument addr specifies an unmappable address.
ENXIO Addresses in the range [off, off + len) are invalid
for fd.
ENXIO The argument len has a value less than or equal to 0.
ENODEV fd refers to an object for which mmap is
meaningless, such as a terminal.
ENOSYS fd refers to an object for which mmap is not
permitted.
ENOMEM zero was passed as the value of addr, and
insufficient space was available in the standard
address ranges. This is primarily an issue for 32
bit programs requesting 1GByte or more, because the
range from 0x30000000 to 0x40000000 is reserved for
MAP_FIXED as described above.
ENOMEM MAP_FIXED was specified and the range [addr, addr +
len) is invalid or exceeds that allowed for the
address space of a process, or MAP_FIXED was not
specified and there is insufficient room in the
address space to effect the mapping.
ENOMEM The calling process has the automatic memory locking
of future mappings enabled [see mlockall(3C)] and
there is insufficient physical memory available for
the mapping.
NOTES
mmap allows access to resources via address space
manipulations instead of the read/write interface. Once a
file is mapped, all a process has to do to access it is use
the data at the address to which the object was mapped.
Consider the following pseudo-code:
fd = open(...)
lseek(fd, offset)
read(fd, buf, len)
/* use data in buf */
Here is a rewrite using mmap:
fd = open(...)
address = mmap(NULL, len, (PROT_READ | PROT_WRITE),
MAP_PRIVATE, fd, offset)
/* use data at address */
Previous IRIX releases have only required MAP_FIXED address
alignment to the system page size returned with
getpagesize(2). This is no longer true for the MIPS
R4000PC, R4600 and R5000 processors. Requests now must be
aligned to the size returned by
sysconf(_SC_MMAP_FIXED_ALIGNMENT).
SEE ALSO
sgi_use_anyaddr(1), fcntl(2), fork(2), madvise(2),
mprotect(2), msync(2), munmap(2), plock(2), sproc(2),
sysconf(2), lockf(3C).
The discussion on MAP_SGI_ANYADDR says to "[s]ee sgi_use_anyaddr", and here it is:
Code: Select all
sgi_use_anyaddr(1) sgi_use_anyaddr(1)
NAME
sgi_use_anyaddr - disable reserved range for auto-placed
mappings
SYNOPSIS
sgi_use_anyaddr command [arguments]
DESCRIPTION
sgi_use_anyaddr executes command with the reserved range of
the virtual address space (from 0x30000000 to 0x40000000)
disabled. Normally, the kernel does not auto-place
mappings made by mmap(2) and shmat(2) in the reserved
range. By disabling the reserved range with
sgi_use_anyaddr, the kernel is permitted to use any
available virtual address for mapping requests made by
command. sgi_use_anyaddr is implemented by invoking
syssgi(2) with request parameter
SGI_UNSUPPORTED_MAP_RESERVED_RANGE and then performing an
exec of command. The effect of this option is the same as
if MAP_SGI_ANYADDR and SHM_SGI_ANYADDR were added to all
future calls to mmap(2) and shmat(2) respectively by
command. Note that this syssgi option is inherited across
fork and exec, so command and all its descendants will be
affected by this.
As described in mmap(2), the kernel chooses a virtual
address for a new mapping created by mmap(2) or shmat(2)
when zero is passed as the attach address. When the kernel
auto-places a mapping in this manner, it is guaranteed that
it will not place the mapping in the reserved range. The
only mappings that can be placed in the reserved range are
those where the user explicitly specifies an attach address
that is in the reserved range.
The reserved range was originally created in the earlier
90's by a MIPS/UNIX standards body at the request of
application writers. The application writers wanted a
portion of the virtual address space where they could be
assured they could place their own mappings via mmap(2) and
shmat(2) without the risk of overlapping other mappings
placed by the kernel. A 256 MB region from 0x30000000 to
0x40000000 was reserved for use by applications to use by
specifying an explicit attach address with mmap(2) or
shmat(2). It is guaranteed the kernel will never auto-place
a mapping in this range when zero is passed as the attach
address to mmap(2) and shmat(2) because this range has been
reserved for applications that specify explicit attach
addresses in this range.
At the time the reserved range was created, applications
were still relatively small by today's standards, so having
this 256 MB hole in the middle of the 2 GB user virtual
address space was not an issue. Application program size
has grown substantially over the years and some
applications find they can no longer easily fit in the 2 GB
address space available when running in 32-bit mode.
Having the reserved range take away 256 MB makes it
difficult for some applications to allocate the memory
space they require on 32-bit systems. This command, the
SGI_UNSUPPORTED_MAP_RESERVED_RANGE syssgi(2) option, the
MAP_SGI_ANYADDR flag to mmap(2), and the SHM_SGI_ANYADDR
flag to shmat(2) together allow the application to control
when the reserved range is used so applications which so
choose may disable it. For compatibility, the reserved
range is normally enabled.
The reserved range is in effect for 64-bit applications as
well, but since the virtual address space is so large for
these it is rarely an issue. In general, 64-bit
applications do not need to be concerned with these new
flags and options to control the reserved range.
FILES
/usr/lib/locale/locale/LC_MESSAGES/uxue
language-specific message file [See LANG on environ(5).]
SEE ALSO
mmap(2), shmat(2). syssgi(2).
WARNINGS
Some existing applications are dependent on the reserved
range being honored by the kernel. Using this command to
disable the reserved range for a command that requires it
may cause it to fail, dump core, or otherwise execute
incorrectly. This applies to any other programs the
command may run via fork/exec since the option is
inherited. Therefore, this command and the syssgi(2)
option must be used with extreme care. It should only be
used with programs the user is certain are not dependent on
the reserved range. The effect this command may have on
applications that are in some way dependent on the reserved
range is strictly unsupported by SGI.