7
Network Working Group R. Gilligan
8
Request for Comments: 3493 Intransa, Inc.
9
Obsoletes: 2553 S. Thomson
10
Category: Informational Cisco
18
Basic Socket Interface Extensions for IPv6
22
This memo provides information for the Internet community. It does
23
not specify an Internet standard of any kind. Distribution of this
28
Copyright (C) The Internet Society (2003). All Rights Reserved.
32
The de facto standard Application Program Interface (API) for TCP/IP
33
applications is the "sockets" interface. Although this API was
34
developed for Unix in the early 1980s it has also been implemented on
35
a wide variety of non-Unix systems. TCP/IP applications written
36
using the sockets API have in the past enjoyed a high degree of
37
portability and we would like the same portability with IPv6
38
applications. But changes are required to the sockets API to support
39
IPv6 and this memo describes these changes. These include a new
40
socket address structure to carry IPv6 addresses, new address
41
conversion functions, and some new socket options. These extensions
42
are designed to provide access to the basic IPv6 features required by
43
TCP and UDP applications, including multicasting, while introducing a
44
minimum of change into the system and providing complete
45
compatibility for existing IPv4 applications. Additional extensions
46
for advanced IPv6 features (raw sockets and access to the IPv6
47
extension headers) are defined in another document.
58
Gilligan, et al. Informational [Page 1]
60
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
65
1. Introduction................................................3
66
2. Design Considerations.......................................4
67
2.1 What Needs to be Changed...............................4
68
2.2 Data Types.............................................6
69
2.3 Headers................................................6
70
2.4 Structures.............................................6
71
3. Socket Interface............................................6
72
3.1 IPv6 Address Family and Protocol Family................6
73
3.2 IPv6 Address Structure.................................7
74
3.3 Socket Address Structure for 4.3BSD-Based Systems......7
75
3.4 Socket Address Structure for 4.4BSD-Based Systems......9
76
3.5 The Socket Functions...................................9
77
3.6 Compatibility with IPv4 Applications..................10
78
3.7 Compatibility with IPv4 Nodes.........................11
79
3.8 IPv6 Wildcard Address.................................11
80
3.9 IPv6 Loopback Address.................................13
81
3.10 Portability Additions.................................14
82
4. Interface Identification...................................16
83
4.1 Name-to-Index.........................................17
84
4.2 Index-to-Name.........................................17
85
4.3 Return All Interface Names and Indexes................18
86
4.4 Free Memory...........................................18
87
5. Socket Options.............................................18
88
5.1 Unicast Hop Limit.....................................19
89
5.2 Sending and Receiving Multicast Packets...............19
90
5.3 IPV6_V6ONLY option for AF_INET6 Sockets...............22
91
6. Library Functions..........................................22
92
6.1 Protocol-Independent Nodename and
93
Service Name Translation..............................23
94
6.2 Socket Address Structure to Node Name
95
and Service Name......................................28
96
6.3 Address Conversion Functions..........................31
97
6.4 Address Testing Macros................................33
98
7. Summary of New Definitions.................................33
99
8. Security Considerations....................................35
100
9. Changes from RFC 2553......................................35
101
10. Acknowledgments............................................36
102
11. References.................................................37
103
12. Authors' Addresses.........................................38
104
13. Full Copyright Statement...................................39
114
Gilligan, et al. Informational [Page 2]
116
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
121
While IPv4 addresses are 32 bits long, IPv6 addresses are 128 bits
122
long. The socket interface makes the size of an IP address quite
123
visible to an application; virtually all TCP/IP applications for
124
BSD-based systems have knowledge of the size of an IP address. Those
125
parts of the API that expose the addresses must be changed to
126
accommodate the larger IPv6 address size. IPv6 also introduces new
127
features, some of which must be made visible to applications via the
128
API. This memo defines a set of extensions to the socket interface
129
to support the larger address size and new features of IPv6. It
130
defines "basic" extensions that are of use to a broad range of
131
applications. A companion document, the "advanced" API [4], covers
132
extensions that are of use to more specialized applications, examples
133
of which include routing daemons, and the "ping" and "traceroute"
136
The development of this API was started in 1994 in the IETF IPng
137
working group. The API has evolved over the years, published first
138
in RFC 2133, then again in RFC 2553, and reaching its final form in
141
As the API matured and stabilized, it was incorporated into the Open
142
Group's Networking Services (XNS) specification, issue 5.2, which was
143
subsequently incorporated into a joint Open Group/IEEE/ISO standard
146
Effort has been made to ensure that this document and [3] contain the
147
same information with regard to the API definitions. However, the
148
reader should note that this document is for informational purposes
149
only, and that the official standard specification of the sockets API
152
It is expected that any future standardization work on this API would
153
be done by the Open Group Base Working Group [6].
155
It should also be noted that this document describes only those
156
portions of the API needed for IPv4 and IPv6 communications. Other
157
potential uses of the API, for example the use of getaddrinfo() and
158
getnameinfo() with the AF_UNIX address family, are beyond the scope
170
Gilligan, et al. Informational [Page 3]
172
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
175
2. Design Considerations
177
There are a number of important considerations in designing changes
178
to this well-worn API:
180
- The API changes should provide both source and binary
181
compatibility for programs written to the original API. That is,
182
existing program binaries should continue to operate when run on a
183
system supporting the new API. In addition, existing applications
184
that are re-compiled and run on a system supporting the new API
185
should continue to operate. Simply put, the API changes for IPv6
186
should not break existing programs. An additional mechanism for
187
implementations to verify this is to verify the new symbols are
188
protected by Feature Test Macros as described in [3]. (Such
189
Feature Test Macros are not defined by this RFC.)
191
- The changes to the API should be as small as possible in order to
192
simplify the task of converting existing IPv4 applications to
195
- Where possible, applications should be able to use this API to
196
interoperate with both IPv6 and IPv4 hosts. Applications should
197
not need to know which type of host they are communicating with.
199
- IPv6 addresses carried in data structures should be 64-bit
200
aligned. This is necessary in order to obtain optimum performance
201
on 64-bit machine architectures.
203
Because of the importance of providing IPv4 compatibility in the API,
204
these extensions are explicitly designed to operate on machines that
205
provide complete support for both IPv4 and IPv6. A subset of this
206
API could probably be designed for operation on systems that support
207
only IPv6. However, this is not addressed in this memo.
209
2.1 What Needs to be Changed
211
The socket interface API consists of a few distinct components:
213
- Core socket functions.
215
- Address data structures.
217
- Name-to-address translation functions.
219
- Address conversion functions.
226
Gilligan, et al. Informational [Page 4]
228
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
231
The core socket functions -- those functions that deal with such
232
things as setting up and tearing down TCP connections, and sending
233
and receiving UDP packets -- were designed to be transport
234
independent. Where protocol addresses are passed as function
235
arguments, they are carried via opaque pointers. A protocol-specific
236
address data structure is defined for each protocol that the socket
237
functions support. Applications must cast pointers to these
238
protocol-specific address structures into pointers to the generic
239
"sockaddr" address structure when using the socket functions. These
240
functions need not change for IPv6, but a new IPv6-specific address
241
data structure is needed.
243
The "sockaddr_in" structure is the protocol-specific data structure
244
for IPv4. This data structure actually includes 8-octets of unused
245
space, and it is tempting to try to use this space to adapt the
246
sockaddr_in structure to IPv6. Unfortunately, the sockaddr_in
247
structure is not large enough to hold the 16-octet IPv6 address as
248
well as the other information (address family and port number) that
249
is needed. So a new address data structure must be defined for IPv6.
251
IPv6 addresses are scoped [2] so they could be link-local, site,
252
organization, global, or other scopes at this time undefined. To
253
support applications that want to be able to identify a set of
254
interfaces for a specific scope, the IPv6 sockaddr_in structure must
255
support a field that can be used by an implementation to identify a
256
set of interfaces identifying the scope for an IPv6 address.
258
The IPv4 name-to-address translation functions in the socket
259
interface are gethostbyname() and gethostbyaddr(). These are left as
260
is, and new functions are defined which support both IPv4 and IPv6.
262
The IPv4 address conversion functions -- inet_ntoa() and inet_addr()
263
-- convert IPv4 addresses between binary and printable form. These
264
functions are quite specific to 32-bit IPv4 addresses. We have
265
designed two analogous functions that convert both IPv4 and IPv6
266
addresses, and carry an address type parameter so that they can be
267
extended to other protocol families as well.
269
Finally, a few miscellaneous features are needed to support IPv6. A
270
new interface is needed to support the IPv6 hop limit header field.
271
New socket options are needed to control the sending and receiving of
272
IPv6 multicast packets.
274
The socket interface will be enhanced in the future to provide access
275
to other IPv6 features. Some of these extensions are described in
282
Gilligan, et al. Informational [Page 5]
284
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
289
The data types of the structure elements given in this memo are
290
intended to track the relevant standards. uintN_t means an unsigned
291
integer of exactly N bits (e.g., uint16_t). The sa_family_t and
292
in_port_t types are defined in [3].
296
When function prototypes and structures are shown we show the headers
297
that must be #included to cause that item to be defined.
301
When structures are described the members shown are the ones that
302
must appear in an implementation. Additional, nonstandard members
303
may also be defined by an implementation. As an additional
304
precaution nonstandard members could be verified by Feature Test
305
Macros as described in [3]. (Such Feature Test Macros are not
306
defined by this RFC.)
308
The ordering shown for the members of a structure is the recommended
309
ordering, given alignment considerations of multibyte members, but an
310
implementation may order the members differently.
314
This section specifies the socket interface changes for IPv6.
316
3.1 IPv6 Address Family and Protocol Family
318
A new address family name, AF_INET6, is defined in <sys/socket.h>.
319
The AF_INET6 definition distinguishes between the original
320
sockaddr_in address data structure, and the new sockaddr_in6 data
323
A new protocol family name, PF_INET6, is defined in <sys/socket.h>.
324
Like most of the other protocol family names, this will usually be
325
defined to have the same value as the corresponding address family
328
#define PF_INET6 AF_INET6
330
The AF_INET6 is used in the first argument to the socket() function
331
to indicate that an IPv6 socket is being created.
338
Gilligan, et al. Informational [Page 6]
340
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
343
3.2 IPv6 Address Structure
345
A new in6_addr structure holds a single IPv6 address and is defined
346
as a result of including <netinet/in.h>:
349
uint8_t s6_addr[16]; /* IPv6 address */
352
This data structure contains an array of sixteen 8-bit elements,
353
which make up one 128-bit IPv6 address. The IPv6 address is stored
354
in network byte order.
356
The structure in6_addr above is usually implemented with an embedded
357
union with extra fields that force the desired alignment level in a
358
manner similar to BSD implementations of "struct in_addr". Those
359
additional implementation details are omitted here for simplicity.
361
An example is as follows:
370
#define s6_addr _S6_un._S6_u8
372
3.3 Socket Address Structure for 4.3BSD-Based Systems
374
In the socket interface, a different protocol-specific data structure
375
is defined to carry the addresses for each protocol suite. Each
376
protocol-specific data structure is designed so it can be cast into a
377
protocol-independent data structure -- the "sockaddr" structure.
378
Each has a "family" field that overlays the "sa_family" of the
379
sockaddr data structure. This field identifies the type of the data
382
The sockaddr_in structure is the protocol-specific address data
383
structure for IPv4. It is used to pass addresses between
384
applications and the system in the socket functions. The following
385
sockaddr_in6 structure holds IPv6 addresses and is defined as a
386
result of including the <netinet/in.h> header:
394
Gilligan, et al. Informational [Page 7]
396
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
399
struct sockaddr_in6 {
400
sa_family_t sin6_family; /* AF_INET6 */
401
in_port_t sin6_port; /* transport layer port # */
402
uint32_t sin6_flowinfo; /* IPv6 flow information */
403
struct in6_addr sin6_addr; /* IPv6 address */
404
uint32_t sin6_scope_id; /* set of interfaces for a scope */
407
This structure is designed to be compatible with the sockaddr data
408
structure used in the 4.3BSD release.
410
The sin6_family field identifies this as a sockaddr_in6 structure.
411
This field overlays the sa_family field when the buffer is cast to a
412
sockaddr data structure. The value of this field must be AF_INET6.
414
The sin6_port field contains the 16-bit UDP or TCP port number. This
415
field is used in the same way as the sin_port field of the
416
sockaddr_in structure. The port number is stored in network byte
419
The sin6_flowinfo field is a 32-bit field intended to contain flow-
420
related information. The exact way this field is mapped to or from a
421
packet is not currently specified. Until such time as its use is
422
specified, applications should set this field to zero when
423
constructing a sockaddr_in6, and ignore this field in a sockaddr_in6
424
structure constructed by the system.
426
The sin6_addr field is a single in6_addr structure (defined in the
427
previous section). This field holds one 128-bit IPv6 address. The
428
address is stored in network byte order.
430
The ordering of elements in this structure is specifically designed
431
so that when sin6_addr field is aligned on a 64-bit boundary, the
432
start of the structure will also be aligned on a 64-bit boundary.
433
This is done for optimum performance on 64-bit architectures.
435
The sin6_scope_id field is a 32-bit integer that identifies a set of
436
interfaces as appropriate for the scope [2] of the address carried in
437
the sin6_addr field. The mapping of sin6_scope_id to an interface or
438
set of interfaces is left to implementation and future specifications
439
on the subject of scoped addresses.
441
Notice that the sockaddr_in6 structure will normally be larger than
442
the generic sockaddr structure. On many existing implementations the
443
sizeof(struct sockaddr_in) equals sizeof(struct sockaddr), with both
444
being 16 bytes. Any existing code that makes this assumption needs
445
to be examined carefully when converting to IPv6.
450
Gilligan, et al. Informational [Page 8]
452
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
455
3.4 Socket Address Structure for 4.4BSD-Based Systems
457
The 4.4BSD release includes a small, but incompatible change to the
458
socket interface. The "sa_family" field of the sockaddr data
459
structure was changed from a 16-bit value to an 8-bit value, and the
460
space saved used to hold a length field, named "sa_len". The
461
sockaddr_in6 data structure given in the previous section cannot be
462
correctly cast into the newer sockaddr data structure. For this
463
reason, the following alternative IPv6 address data structure is
464
provided to be used on systems based on 4.4BSD. It is defined as a
465
result of including the <netinet/in.h> header.
467
struct sockaddr_in6 {
468
uint8_t sin6_len; /* length of this struct */
469
sa_family_t sin6_family; /* AF_INET6 */
470
in_port_t sin6_port; /* transport layer port # */
471
uint32_t sin6_flowinfo; /* IPv6 flow information */
472
struct in6_addr sin6_addr; /* IPv6 address */
473
uint32_t sin6_scope_id; /* set of interfaces for a scope */
476
The only differences between this data structure and the 4.3BSD
477
variant are the inclusion of the length field, and the change of the
478
family field to a 8-bit data type. The definitions of all the other
479
fields are identical to the structure defined in the previous
482
Systems that provide this version of the sockaddr_in6 data structure
483
must also declare SIN6_LEN as a result of including the
484
<netinet/in.h> header. This macro allows applications to determine
485
whether they are being built on a system that supports the 4.3BSD or
486
4.4BSD variants of the data structure.
488
3.5 The Socket Functions
490
Applications call the socket() function to create a socket descriptor
491
that represents a communication endpoint. The arguments to the
492
socket() function tell the system which protocol to use, and what
493
format address structure will be used in subsequent functions. For
494
example, to create an IPv4/TCP socket, applications make the call:
496
s = socket(AF_INET, SOCK_STREAM, 0);
498
To create an IPv4/UDP socket, applications make the call:
500
s = socket(AF_INET, SOCK_DGRAM, 0);
506
Gilligan, et al. Informational [Page 9]
508
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
511
Applications may create IPv6/TCP and IPv6/UDP sockets (which may also
512
handle IPv4 communication as described in section 3.7) by simply
513
using the constant AF_INET6 instead of AF_INET in the first argument.
514
For example, to create an IPv6/TCP socket, applications make the
517
s = socket(AF_INET6, SOCK_STREAM, 0);
519
To create an IPv6/UDP socket, applications make the call:
521
s = socket(AF_INET6, SOCK_DGRAM, 0);
523
Once the application has created a AF_INET6 socket, it must use the
524
sockaddr_in6 address structure when passing addresses in to the
525
system. The functions that the application uses to pass addresses
533
The system will use the sockaddr_in6 address structure to return
534
addresses to applications that are using AF_INET6 sockets. The
535
functions that return an address from the system to an application
544
No changes to the syntax of the socket functions are needed to
545
support IPv6, since all of the "address carrying" functions use an
546
opaque address pointer, and carry an address length as a function
549
3.6 Compatibility with IPv4 Applications
551
In order to support the large base of applications using the original
552
API, system implementations must provide complete source and binary
553
compatibility with the original API. This means that systems must
554
continue to support AF_INET sockets and the sockaddr_in address
555
structure. Applications must be able to create IPv4/TCP and IPv4/UDP
556
sockets using the AF_INET constant in the socket() function, as
562
Gilligan, et al. Informational [Page 10]
564
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
567
described in the previous section. Applications should be able to
568
hold a combination of IPv4/TCP, IPv4/UDP, IPv6/TCP and IPv6/UDP
569
sockets simultaneously within the same process.
571
Applications using the original API should continue to operate as
572
they did on systems supporting only IPv4. That is, they should
573
continue to interoperate with IPv4 nodes.
575
3.7 Compatibility with IPv4 Nodes
577
The API also provides a different type of compatibility: the ability
578
for IPv6 applications to interoperate with IPv4 applications. This
579
feature uses the IPv4-mapped IPv6 address format defined in the IPv6
580
addressing architecture specification [2]. This address format
581
allows the IPv4 address of an IPv4 node to be represented as an IPv6
582
address. The IPv4 address is encoded into the low-order 32 bits of
583
the IPv6 address, and the high-order 96 bits hold the fixed prefix
584
0:0:0:0:0:FFFF. IPv4-mapped addresses are written as follows:
586
::FFFF:<IPv4-address>
588
These addresses can be generated automatically by the getaddrinfo()
589
function, as described in Section 6.1.
591
Applications may use AF_INET6 sockets to open TCP connections to IPv4
592
nodes, or send UDP packets to IPv4 nodes, by simply encoding the
593
destination's IPv4 address as an IPv4-mapped IPv6 address, and
594
passing that address, within a sockaddr_in6 structure, in the
595
connect() or sendto() call. When applications use AF_INET6 sockets
596
to accept TCP connections from IPv4 nodes, or receive UDP packets
597
from IPv4 nodes, the system returns the peer's address to the
598
application in the accept(), recvfrom(), or getpeername() call using
599
a sockaddr_in6 structure encoded this way.
601
Few applications will likely need to know which type of node they are
602
interoperating with. However, for those applications that do need to
603
know, the IN6_IS_ADDR_V4MAPPED() macro, defined in Section 6.4, is
606
3.8 IPv6 Wildcard Address
608
While the bind() function allows applications to select the source IP
609
address of UDP packets and TCP connections, applications often want
610
the system to select the source address for them. With IPv4, one
611
specifies the address as the symbolic constant INADDR_ANY (called the
612
"wildcard" address) in the bind() call, or simply omits the bind()
618
Gilligan, et al. Informational [Page 11]
620
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
623
Since the IPv6 address type is a structure (struct in6_addr), a
624
symbolic constant can be used to initialize an IPv6 address variable,
625
but cannot be used in an assignment. Therefore systems provide the
626
IPv6 wildcard address in two forms.
628
The first version is a global variable named "in6addr_any" that is an
629
in6_addr structure. The extern declaration for this variable is
630
defined in <netinet/in.h>:
632
extern const struct in6_addr in6addr_any;
634
Applications use in6addr_any similarly to the way they use INADDR_ANY
635
in IPv4. For example, to bind a socket to port number 23, but let
636
the system select the source address, an application could use the
639
struct sockaddr_in6 sin6;
641
sin6.sin6_family = AF_INET6;
642
sin6.sin6_flowinfo = 0;
643
sin6.sin6_port = htons(23);
644
sin6.sin6_addr = in6addr_any; /* structure assignment */
646
if (bind(s, (struct sockaddr *) &sin6, sizeof(sin6)) == -1)
649
The other version is a symbolic constant named IN6ADDR_ANY_INIT and
650
is defined in <netinet/in.h>. This constant can be used to
651
initialize an in6_addr structure:
653
struct in6_addr anyaddr = IN6ADDR_ANY_INIT;
655
Note that this constant can be used ONLY at declaration time. It can
656
not be used to assign a previously declared in6_addr structure. For
657
example, the following code will not work:
659
/* This is the WRONG way to assign an unspecified address */
660
struct sockaddr_in6 sin6;
662
sin6.sin6_addr = IN6ADDR_ANY_INIT; /* will NOT compile */
664
Be aware that the IPv4 INADDR_xxx constants are all defined in host
665
byte order but the IPv6 IN6ADDR_xxx constants and the IPv6
666
in6addr_xxx externals are defined in network byte order.
674
Gilligan, et al. Informational [Page 12]
676
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
679
3.9 IPv6 Loopback Address
681
Applications may need to send UDP packets to, or originate TCP
682
connections to, services residing on the local node. In IPv4, they
683
can do this by using the constant IPv4 address INADDR_LOOPBACK in
684
their connect(), sendto(), or sendmsg() call.
686
IPv6 also provides a loopback address to contact local TCP and UDP
687
services. Like the unspecified address, the IPv6 loopback address is
688
provided in two forms -- a global variable and a symbolic constant.
690
The global variable is an in6_addr structure named
691
"in6addr_loopback." The extern declaration for this variable is
692
defined in <netinet/in.h>:
694
extern const struct in6_addr in6addr_loopback;
696
Applications use in6addr_loopback as they would use INADDR_LOOPBACK
697
in IPv4 applications (but beware of the byte ordering difference
698
mentioned at the end of the previous section). For example, to open
699
a TCP connection to the local telnet server, an application could use
702
struct sockaddr_in6 sin6;
704
sin6.sin6_family = AF_INET6;
705
sin6.sin6_flowinfo = 0;
706
sin6.sin6_port = htons(23);
707
sin6.sin6_addr = in6addr_loopback; /* structure assignment */
709
if (connect(s, (struct sockaddr *) &sin6, sizeof(sin6)) == -1)
712
The symbolic constant is named IN6ADDR_LOOPBACK_INIT and is defined
713
in <netinet/in.h>. It can be used at declaration time ONLY; for
716
struct in6_addr loopbackaddr = IN6ADDR_LOOPBACK_INIT;
718
Like IN6ADDR_ANY_INIT, this constant cannot be used in an assignment
719
to a previously declared IPv6 address variable.
730
Gilligan, et al. Informational [Page 13]
732
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
735
3.10 Portability Additions
737
One simple addition to the sockets API that can help application
738
writers is the "struct sockaddr_storage". This data structure can
739
simplify writing code that is portable across multiple address
740
families and platforms. This data structure is designed with the
743
- Large enough to accommodate all supported protocol-specific address
746
- Aligned at an appropriate boundary so that pointers to it can be
747
cast as pointers to protocol specific address structures and used
748
to access the fields of those structures without alignment
751
The sockaddr_storage structure contains field ss_family which is of
752
type sa_family_t. When a sockaddr_storage structure is cast to a
753
sockaddr structure, the ss_family field of the sockaddr_storage
754
structure maps onto the sa_family field of the sockaddr structure.
755
When a sockaddr_storage structure is cast as a protocol specific
756
address structure, the ss_family field maps onto a field of that
757
structure that is of type sa_family_t and that identifies the
758
protocol's address family.
786
Gilligan, et al. Informational [Page 14]
788
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
791
An example implementation design of such a data structure would be as
795
* Desired design of maximum size and alignment
797
#define _SS_MAXSIZE 128 /* Implementation specific max size */
798
#define _SS_ALIGNSIZE (sizeof (int64_t))
799
/* Implementation specific desired alignment */
801
* Definitions used for sockaddr_storage structure paddings design.
803
#define _SS_PAD1SIZE (_SS_ALIGNSIZE - sizeof (sa_family_t))
804
#define _SS_PAD2SIZE (_SS_MAXSIZE - (sizeof (sa_family_t) +
805
_SS_PAD1SIZE + _SS_ALIGNSIZE))
806
struct sockaddr_storage {
807
sa_family_t ss_family; /* address family */
808
/* Following fields are implementation specific */
809
char __ss_pad1[_SS_PAD1SIZE];
810
/* 6 byte pad, this is to make implementation
811
/* specific pad up to alignment field that */
812
/* follows explicit in the data structure */
813
int64_t __ss_align; /* field to force desired structure */
814
/* storage alignment */
815
char __ss_pad2[_SS_PAD2SIZE];
816
/* 112 byte pad to achieve desired size, */
817
/* _SS_MAXSIZE value minus size of ss_family */
818
/* __ss_pad1, __ss_align fields is 112 */
821
The above example implementation illustrates a data structure which
822
will align on a 64-bit boundary. An implementation-specific field
823
"__ss_align" along with "__ss_pad1" is used to force a 64-bit
824
alignment which covers proper alignment good enough for the needs of
825
sockaddr_in6 (IPv6), sockaddr_in (IPv4) address data structures. The
826
size of padding field __ss_pad1 depends on the chosen alignment
827
boundary. The size of padding field __ss_pad2 depends on the value
828
of overall size chosen for the total size of the structure. This
829
size and alignment are represented in the above example by
830
implementation specific (not required) constants _SS_MAXSIZE (chosen
831
value 128) and _SS_ALIGNSIZE (with chosen value 8). Constants
832
_SS_PAD1SIZE (derived value 6) and _SS_PAD2SIZE (derived value 112)
833
are also for illustration and not required. The derived values
834
assume sa_family_t is 2 bytes. The implementation specific
835
definitions and structure field names above start with an underscore
836
to denote implementation private namespace. Portable code is not
837
expected to access or reference those fields or constants.
842
Gilligan, et al. Informational [Page 15]
844
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
847
On implementations where the sockaddr data structure includes a
848
"sa_len" field this data structure would look like this:
851
* Definitions used for sockaddr_storage structure paddings design.
853
#define _SS_PAD1SIZE (_SS_ALIGNSIZE -
854
(sizeof (uint8_t) + sizeof (sa_family_t))
855
#define _SS_PAD2SIZE (_SS_MAXSIZE -
856
(sizeof (uint8_t) + sizeof (sa_family_t) +
857
_SS_PAD1SIZE + _SS_ALIGNSIZE))
858
struct sockaddr_storage {
859
uint8_t ss_len; /* address length */
860
sa_family_t ss_family; /* address family */
861
/* Following fields are implementation specific */
862
char __ss_pad1[_SS_PAD1SIZE];
863
/* 6 byte pad, this is to make implementation
864
/* specific pad up to alignment field that */
865
/* follows explicit in the data structure */
866
int64_t __ss_align; /* field to force desired structure */
867
/* storage alignment */
868
char __ss_pad2[_SS_PAD2SIZE];
869
/* 112 byte pad to achieve desired size, */
870
/* _SS_MAXSIZE value minus size of ss_len, */
871
/* __ss_family, __ss_pad1, __ss_align fields is 112 */
874
4. Interface Identification
876
This API uses an interface index (a small positive integer) to
877
identify the local interface on which a multicast group is joined
878
(Section 5.2). Additionally, the advanced API [4] uses these same
879
interface indexes to identify the interface on which a datagram is
880
received, or to specify the interface on which a datagram is to be
883
Interfaces are normally known by names such as "le0", "sl1", "ppp2",
884
and the like. On Berkeley-derived implementations, when an interface
885
is made known to the system, the kernel assigns a unique positive
886
integer value (called the interface index) to that interface. These
887
are small positive integers that start at 1. (Note that 0 is never
888
used for an interface index.) There may be gaps so that there is no
889
current interface for a particular positive interface index.
891
This API defines two functions that map between an interface name and
892
index, a third function that returns all the interface names and
893
indexes, and a fourth function to return the dynamic memory allocated
894
by the previous function. How these functions are implemented is
898
Gilligan, et al. Informational [Page 16]
900
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
903
left up to the implementation. 4.4BSD implementations can implement
904
these functions using the existing sysctl() function with the
905
NET_RT_IFLIST command. Other implementations may wish to use ioctl()
910
The first function maps an interface name into its corresponding
915
unsigned int if_nametoindex(const char *ifname);
917
If ifname is the name of an interface, the if_nametoindex() function
918
shall return the interface index corresponding to name ifname;
919
otherwise, it shall return zero. No errors are defined.
923
The second function maps an interface index into its corresponding
928
char *if_indextoname(unsigned int ifindex, char *ifname);
930
When this function is called, the ifname argument shall point to a
931
buffer of at least IF_NAMESIZE bytes. The function shall place in
932
this buffer the name of the interface with index ifindex.
933
(IF_NAMESIZE is also defined in <net/if.h> and its value includes a
934
terminating null byte at the end of the interface name.) If ifindex
935
is an interface index, then the function shall return the value
936
supplied in ifname, which points to a buffer now containing the
937
interface name. Otherwise, the function shall return a NULL pointer
938
and set errno to indicate the error. If there is no interface
939
corresponding to the specified index, errno is set to ENXIO. If
940
there was a system error (such as running out of memory), errno would
941
be set to the proper value (e.g., ENOMEM).
954
Gilligan, et al. Informational [Page 17]
956
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
959
4.3 Return All Interface Names and Indexes
961
The if_nameindex structure holds the information about a single
962
interface and is defined as a result of including the <net/if.h>
965
struct if_nameindex {
966
unsigned int if_index; /* 1, 2, ... */
967
char *if_name; /* null terminated name: "le0", ... */
970
The final function returns an array of if_nameindex structures, one
971
structure per interface.
975
struct if_nameindex *if_nameindex(void);
977
The end of the array of structures is indicated by a structure with
978
an if_index of 0 and an if_name of NULL. The function returns a NULL
979
pointer upon an error, and would set errno to the appropriate value.
981
The memory used for this array of structures along with the interface
982
names pointed to by the if_name members is obtained dynamically.
983
This memory is freed by the next function.
987
The following function frees the dynamic memory that was allocated by
992
void if_freenameindex(struct if_nameindex *ptr);
994
The ptr argument shall be a pointer that was returned by
995
if_nameindex(). After if_freenameindex() has been called, the
996
application shall not use the array of which ptr is the address.
1000
A number of new socket options are defined for IPv6. All of these
1001
new options are at the IPPROTO_IPV6 level. That is, the "level"
1002
parameter in the getsockopt() and setsockopt() calls is IPPROTO_IPV6
1003
when using these options. The constant name prefix IPV6_ is used in
1004
all of the new socket options. This serves to clearly identify these
1005
options as applying to IPv6.
1010
Gilligan, et al. Informational [Page 18]
1012
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1015
The declaration for IPPROTO_IPV6, the new IPv6 socket options, and
1016
related constants defined in this section are obtained by including
1017
the header <netinet/in.h>.
1019
5.1 Unicast Hop Limit
1021
A new setsockopt() option controls the hop limit used in outgoing
1022
unicast IPv6 packets. The name of this option is IPV6_UNICAST_HOPS,
1023
and it is used at the IPPROTO_IPV6 layer. The following example
1024
illustrates how it is used:
1028
if (setsockopt(s, IPPROTO_IPV6, IPV6_UNICAST_HOPS,
1029
(char *) &hoplimit, sizeof(hoplimit)) == -1)
1030
perror("setsockopt IPV6_UNICAST_HOPS");
1032
When the IPV6_UNICAST_HOPS option is set with setsockopt(), the
1033
option value given is used as the hop limit for all subsequent
1034
unicast packets sent via that socket. If the option is not set, the
1035
system selects a default value. The integer hop limit value (called
1036
x) is interpreted as follows:
1038
x < -1: return an error of EINVAL
1039
x == -1: use kernel default
1040
0 <= x <= 255: use x
1041
x >= 256: return an error of EINVAL
1043
The IPV6_UNICAST_HOPS option may be used with getsockopt() to
1044
determine the hop limit value that the system will use for subsequent
1045
unicast packets sent via that socket. For example:
1048
socklen_t len = sizeof(hoplimit);
1050
if (getsockopt(s, IPPROTO_IPV6, IPV6_UNICAST_HOPS,
1051
(char *) &hoplimit, &len) == -1)
1052
perror("getsockopt IPV6_UNICAST_HOPS");
1054
printf("Using %d for hop limit.\n", hoplimit);
1056
5.2 Sending and Receiving Multicast Packets
1058
IPv6 applications may send multicast packets by simply specifying an
1059
IPv6 multicast address as the destination address, for example in the
1060
destination address argument of the sendto() function.
1066
Gilligan, et al. Informational [Page 19]
1068
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1071
Three socket options at the IPPROTO_IPV6 layer control some of the
1072
parameters for sending multicast packets. Setting these options is
1073
not required: applications may send multicast packets without using
1074
these options. The setsockopt() options for controlling the sending
1075
of multicast packets are summarized below. These three options can
1076
also be used with getsockopt().
1080
Set the interface to use for outgoing multicast packets. The
1081
argument is the index of the interface to use. If the
1082
interface index is specified as zero, the system selects the
1083
interface (for example, by looking up the address in a routing
1084
table and using the resulting interface).
1086
Argument type: unsigned int
1090
Set the hop limit to use for outgoing multicast packets. (Note
1091
a separate option - IPV6_UNICAST_HOPS - is provided to set the
1092
hop limit to use for outgoing unicast packets.)
1094
The interpretation of the argument is the same as for the
1095
IPV6_UNICAST_HOPS option:
1097
x < -1: return an error of EINVAL
1098
x == -1: use kernel default
1099
0 <= x <= 255: use x
1100
x >= 256: return an error of EINVAL
1102
If IPV6_MULTICAST_HOPS is not set, the default is 1
1103
(same as IPv4 today)
1109
If a multicast datagram is sent to a group to which the sending
1110
host itself belongs (on the outgoing interface), a copy of the
1111
datagram is looped back by the IP layer for local delivery if
1112
this option is set to 1. If this option is set to 0 a copy is
1113
not looped back. Other option values return an error of
1122
Gilligan, et al. Informational [Page 20]
1124
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1127
If IPV6_MULTICAST_LOOP is not set, the default is 1 (loopback;
1128
same as IPv4 today).
1130
Argument type: unsigned int
1132
The reception of multicast packets is controlled by the two
1133
setsockopt() options summarized below. An error of EOPNOTSUPP is
1134
returned if these two options are used with getsockopt().
1138
Join a multicast group on a specified local interface.
1139
If the interface index is specified as 0,
1140
the kernel chooses the local interface.
1141
For example, some kernels look up the multicast group
1142
in the normal IPv6 routing table and use the resulting
1145
Argument type: struct ipv6_mreq
1149
Leave a multicast group on a specified interface.
1150
If the interface index is specified as 0, the system
1151
may choose a multicast group membership to drop by
1152
matching the multicast address only.
1154
Argument type: struct ipv6_mreq
1156
The argument type of both of these options is the ipv6_mreq
1157
structure, defined as a result of including the <netinet/in.h>
1161
struct in6_addr ipv6mr_multiaddr; /* IPv6 multicast addr */
1162
unsigned int ipv6mr_interface; /* interface index */
1165
Note that to receive multicast datagrams a process must join the
1166
multicast group to which datagrams will be sent. UDP applications
1167
must also bind the UDP port to which datagrams will be sent. Some
1168
processes also bind the multicast group address to the socket, in
1169
addition to the port, to prevent other datagrams destined to that
1170
same port from being delivered to the socket.
1178
Gilligan, et al. Informational [Page 21]
1180
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1183
5.3 IPV6_V6ONLY option for AF_INET6 Sockets
1185
This socket option restricts AF_INET6 sockets to IPv6 communications
1186
only. As stated in section <3.7 Compatibility with IPv4 Nodes>,
1187
AF_INET6 sockets may be used for both IPv4 and IPv6 communications.
1188
Some applications may want to restrict their use of an AF_INET6
1189
socket to IPv6 communications only. For these applications the
1190
IPV6_V6ONLY socket option is defined. When this option is turned on,
1191
the socket can be used to send and receive IPv6 packets only. This
1192
is an IPPROTO_IPV6 level option. This option takes an int value.
1193
This is a boolean option. By default this option is turned off.
1195
Here is an example of setting this option:
1199
if (setsockopt(s, IPPROTO_IPV6, IPV6_V6ONLY,
1200
(char *)&on, sizeof(on)) == -1)
1201
perror("setsockopt IPV6_V6ONLY");
1203
printf("IPV6_V6ONLY set\n");
1205
Note - This option has no effect on the use of IPv4 Mapped addresses
1206
which enter a node as a valid IPv6 addresses for IPv6 communications
1207
as defined by Stateless IP/ICMP Translation Algorithm (SIIT) [5].
1209
An example use of this option is to allow two versions of the same
1210
server process to run on the same port, one providing service over
1211
IPv6, the other providing the same service over IPv4.
1213
6. Library Functions
1215
New library functions are needed to perform a variety of operations
1216
with IPv6 addresses. Functions are needed to lookup IPv6 addresses
1217
in the Domain Name System (DNS). Both forward lookup (nodename-to-
1218
address translation) and reverse lookup (address-to-nodename
1219
translation) need to be supported. Functions are also needed to
1220
convert IPv6 addresses between their binary and textual form.
1222
We note that the two existing functions, gethostbyname() and
1223
gethostbyaddr(), are left as-is. New functions are defined to handle
1224
both IPv4 and IPv6 addresses.
1226
The commonly used function gethostbyname() is inadequate for many
1227
applications, first because it provides no way for the caller to
1228
specify anything about the types of addresses desired (IPv4 only,
1229
IPv6 only, IPv4-mapped IPv6 are OK, etc.), and second because many
1230
implementations of this function are not thread safe. RFC 2133
1234
Gilligan, et al. Informational [Page 22]
1236
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1239
defined a function named gethostbyname2() but this function was also
1240
inadequate, first because its use required setting a global option
1241
(RES_USE_INET6) when IPv6 addresses were required, and second because
1242
a flag argument is needed to provide the caller with additional
1243
control over the types of addresses required. The gethostbyname2()
1244
function was deprecated in RFC 2553 and is no longer part of the
1247
6.1 Protocol-Independent Nodename and Service Name Translation
1249
Nodename-to-address translation is done in a protocol-independent
1250
fashion using the getaddrinfo() function.
1252
#include <sys/socket.h>
1256
int getaddrinfo(const char *nodename, const char *servname,
1257
const struct addrinfo *hints, struct addrinfo **res);
1259
void freeaddrinfo(struct addrinfo *ai);
1262
int ai_flags; /* AI_PASSIVE, AI_CANONNAME,
1263
AI_NUMERICHOST, .. */
1264
int ai_family; /* AF_xxx */
1265
int ai_socktype; /* SOCK_xxx */
1266
int ai_protocol; /* 0 or IPPROTO_xxx for IPv4 and IPv6 */
1267
socklen_t ai_addrlen; /* length of ai_addr */
1268
char *ai_canonname; /* canonical name for nodename */
1269
struct sockaddr *ai_addr; /* binary address */
1270
struct addrinfo *ai_next; /* next structure in linked list */
1273
The getaddrinfo() function translates the name of a service location
1274
(for example, a host name) and/or a service name and returns a set of
1275
socket addresses and associated information to be used in creating a
1276
socket with which to address the specified service.
1278
The nodename and servname arguments are either null pointers or
1279
pointers to null-terminated strings. One or both of these two
1280
arguments must be a non-null pointer.
1282
The format of a valid name depends on the address family or families.
1283
If a specific family is not given and the name could be interpreted
1284
as valid within multiple supported families, the implementation will
1285
attempt to resolve the name in all supported families and, in absence
1286
of errors, one or more results shall be returned.
1290
Gilligan, et al. Informational [Page 23]
1292
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1295
If the nodename argument is not null, it can be a descriptive name or
1296
can be an address string. If the specified address family is
1297
AF_INET, AF_INET6, or AF_UNSPEC, valid descriptive names include host
1298
names. If the specified address family is AF_INET or AF_UNSPEC,
1299
address strings using Internet standard dot notation as specified in
1300
inet_addr() are valid. If the specified address family is AF_INET6
1301
or AF_UNSPEC, standard IPv6 text forms described in inet_pton() are
1304
If nodename is not null, the requested service location is named by
1305
nodename; otherwise, the requested service location is local to the
1308
If servname is null, the call shall return network-level addresses
1309
for the specified nodename. If servname is not null, it is a null-
1310
terminated character string identifying the requested service. This
1311
can be either a descriptive name or a numeric representation suitable
1312
for use with the address family or families. If the specified
1313
address family is AF_INET, AF_INET6 or AF_UNSPEC, the service can be
1314
specified as a string specifying a decimal port number.
1316
If the argument hints is not null, it refers to a structure
1317
containing input values that may direct the operation by providing
1318
options and by limiting the returned information to a specific socket
1319
type, address family and/or protocol. In this hints structure every
1320
member other than ai_flags, ai_family, ai_socktype and ai_protocol
1321
shall be set to zero or a null pointer. A value of AF_UNSPEC for
1322
ai_family means that the caller shall accept any address family. A
1323
value of zero for ai_socktype means that the caller shall accept any
1324
socket type. A value of zero for ai_protocol means that the caller
1325
shall accept any protocol. If hints is a null pointer, the behavior
1326
shall be as if it referred to a structure containing the value zero
1327
for the ai_flags, ai_socktype and ai_protocol fields, and AF_UNSPEC
1328
for the ai_family field.
1332
1. If the caller handles only TCP and not UDP, for example, then the
1333
ai_protocol member of the hints structure should be set to
1334
IPPROTO_TCP when getaddrinfo() is called.
1336
2. If the caller handles only IPv4 and not IPv6, then the ai_family
1337
member of the hints structure should be set to AF_INET when
1338
getaddrinfo() is called.
1346
Gilligan, et al. Informational [Page 24]
1348
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1351
The ai_flags field to which hints parameter points shall be set to
1352
zero or be the bitwise-inclusive OR of one or more of the values
1353
AI_PASSIVE, AI_CANONNAME, AI_NUMERICHOST, AI_NUMERICSERV,
1354
AI_V4MAPPED, AI_ALL, and AI_ADDRCONFIG.
1356
If the AI_PASSIVE flag is specified, the returned address information
1357
shall be suitable for use in binding a socket for accepting incoming
1358
connections for the specified service (i.e., a call to bind()). In
1359
this case, if the nodename argument is null, then the IP address
1360
portion of the socket address structure shall be set to INADDR_ANY
1361
for an IPv4 address or IN6ADDR_ANY_INIT for an IPv6 address. If the
1362
AI_PASSIVE flag is not specified, the returned address information
1363
shall be suitable for a call to connect() (for a connection-mode
1364
protocol) or for a call to connect(), sendto() or sendmsg() (for a
1365
connectionless protocol). In this case, if the nodename argument is
1366
null, then the IP address portion of the socket address structure
1367
shall be set to the loopback address. This flag is ignored if the
1368
nodename argument is not null.
1370
If the AI_CANONNAME flag is specified and the nodename argument is
1371
not null, the function shall attempt to determine the canonical name
1372
corresponding to nodename (for example, if nodename is an alias or
1373
shorthand notation for a complete name).
1375
If the AI_NUMERICHOST flag is specified, then a non-null nodename
1376
string supplied shall be a numeric host address string. Otherwise,
1377
an [EAI_NONAME] error is returned. This flag shall prevent any type
1378
of name resolution service (for example, the DNS) from being invoked.
1380
If the AI_NUMERICSERV flag is specified, then a non-null servname
1381
string supplied shall be a numeric port string. Otherwise, an
1382
[EAI_NONAME] error shall be returned. This flag shall prevent any
1383
type of name resolution service (for example, NIS+) from being
1386
If the AI_V4MAPPED flag is specified along with an ai_family of
1387
AF_INET6, then getaddrinfo() shall return IPv4-mapped IPv6 addresses
1388
on finding no matching IPv6 addresses (ai_addrlen shall be 16).
1390
For example, when using the DNS, if no AAAA records are found then
1391
a query is made for A records and any found are returned as IPv4-
1392
mapped IPv6 addresses.
1394
The AI_V4MAPPED flag shall be ignored unless ai_family equals
1397
If the AI_ALL flag is used with the AI_V4MAPPED flag, then
1398
getaddrinfo() shall return all matching IPv6 and IPv4 addresses.
1402
Gilligan, et al. Informational [Page 25]
1404
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1407
For example, when using the DNS, queries are made for both AAAA
1408
records and A records, and getaddrinfo() returns the combined
1409
results of both queries. Any IPv4 addresses found are returned as
1410
IPv4-mapped IPv6 addresses.
1412
The AI_ALL flag without the AI_V4MAPPED flag is ignored.
1416
When ai_family is not specified (AF_UNSPEC), AI_V4MAPPED and
1417
AI_ALL flags will only be used if AF_INET6 is supported.
1419
If the AI_ADDRCONFIG flag is specified, IPv4 addresses shall be
1420
returned only if an IPv4 address is configured on the local system,
1421
and IPv6 addresses shall be returned only if an IPv6 address is
1422
configured on the local system. The loopback address is not
1423
considered for this case as valid as a configured address.
1425
For example, when using the DNS, a query for AAAA records should
1426
occur only if the node has at least one IPv6 address configured
1427
(other than IPv6 loopback) and a query for A records should occur
1428
only if the node has at least one IPv4 address configured (other
1429
than the IPv4 loopback).
1431
The ai_socktype field to which argument hints points specifies the
1432
socket type for the service, as defined for socket(). If a specific
1433
socket type is not given (for example, a value of zero) and the
1434
service name could be interpreted as valid with multiple supported
1435
socket types, the implementation shall attempt to resolve the service
1436
name for all supported socket types and, in the absence of errors,
1437
all possible results shall be returned. A non-zero socket type value
1438
shall limit the returned information to values with the specified
1441
If the ai_family field to which hints points has the value AF_UNSPEC,
1442
addresses shall be returned for use with any address family that can
1443
be used with the specified nodename and/or servname. Otherwise,
1444
addresses shall be returned for use only with the specified address
1445
family. If ai_family is not AF_UNSPEC and ai_protocol is not zero,
1446
then addresses are returned for use only with the specified address
1447
family and protocol; the value of ai_protocol shall be interpreted as
1448
in a call to the socket() function with the corresponding values of
1449
ai_family and ai_protocol.
1451
The freeaddrinfo() function frees one or more addrinfo structures
1452
returned by getaddrinfo(), along with any additional storage
1453
associated with those structures (for example, storage pointed to by
1454
the ai_canonname and ai_addr fields; an application must not
1458
Gilligan, et al. Informational [Page 26]
1460
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1463
reference this storage after the associated addrinfo structure has
1464
been freed). If the ai_next field of the structure is not null, the
1465
entire list of structures is freed. The freeaddrinfo() function must
1466
support the freeing of arbitrary sublists of an addrinfo list
1467
originally returned by getaddrinfo().
1469
Functions getaddrinfo() and freeaddrinfo() must be thread-safe.
1471
A zero return value for getaddrinfo() indicates successful
1472
completion; a non-zero return value indicates failure. The possible
1473
values for the failures are listed below under Error Return Values.
1475
Upon successful return of getaddrinfo(), the location to which res
1476
points shall refer to a linked list of addrinfo structures, each of
1477
which shall specify a socket address and information for use in
1478
creating a socket with which to use that socket address. The list
1479
shall include at least one addrinfo structure. The ai_next field of
1480
each structure contains a pointer to the next structure on the list,
1481
or a null pointer if it is the last structure on the list. Each
1482
structure on the list shall include values for use with a call to the
1483
socket() function, and a socket address for use with the connect()
1484
function or, if the AI_PASSIVE flag was specified, for use with the
1485
bind() function. The fields ai_family, ai_socktype, and ai_protocol
1486
shall be usable as the arguments to the socket() function to create a
1487
socket suitable for use with the returned address. The fields
1488
ai_addr and ai_addrlen are usable as the arguments to the connect()
1489
or bind() functions with such a socket, according to the AI_PASSIVE
1492
If nodename is not null, and if requested by the AI_CANONNAME flag,
1493
the ai_canonname field of the first returned addrinfo structure shall
1494
point to a null-terminated string containing the canonical name
1495
corresponding to the input nodename; if the canonical name is not
1496
available, then ai_canonname shall refer to the nodename argument or
1497
a string with the same contents. The contents of the ai_flags field
1498
of the returned structures are undefined.
1500
All fields in socket address structures returned by getaddrinfo()
1501
that are not filled in through an explicit argument (for example,
1502
sin6_flowinfo) shall be set to zero.
1504
Note: This makes it easier to compare socket address structures.
1514
Gilligan, et al. Informational [Page 27]
1516
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1519
Error Return Values:
1521
The getaddrinfo() function shall fail and return the corresponding
1524
[EAI_AGAIN] The name could not be resolved at this time. Future
1525
attempts may succeed.
1527
[EAI_BADFLAGS] The flags parameter had an invalid value.
1529
[EAI_FAIL] A non-recoverable error occurred when attempting to
1532
[EAI_FAMILY] The address family was not recognized.
1534
[EAI_MEMORY] There was a memory allocation failure when trying to
1535
allocate storage for the return value.
1537
[EAI_NONAME] The name does not resolve for the supplied
1538
parameters. Neither nodename nor servname were
1539
supplied. At least one of these must be supplied.
1541
[EAI_SERVICE] The service passed was not recognized for the
1542
specified socket type.
1544
[EAI_SOCKTYPE] The intended socket type was not recognized.
1546
[EAI_SYSTEM] A system error occurred; the error code can be found
1549
The gai_strerror() function provides a descriptive text string
1550
corresponding to an EAI_xxx error value.
1554
const char *gai_strerror(int ecode);
1556
The argument is one of the EAI_xxx values defined for the
1557
getaddrinfo() and getnameinfo() functions. The return value points
1558
to a string describing the error. If the argument is not one of the
1559
EAI_xxx values, the function still returns a pointer to a string
1560
whose contents indicate an unknown error.
1562
6.2 Socket Address Structure to Node Name and Service Name
1564
The getnameinfo() function is used to translate the contents of a
1565
socket address structure to a node name and/or service name.
1570
Gilligan, et al. Informational [Page 28]
1572
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1575
#include <sys/socket.h>
1578
int getnameinfo(const struct sockaddr *sa, socklen_t salen,
1579
char *node, socklen_t nodelen,
1580
char *service, socklen_t servicelen,
1583
The getnameinfo() function shall translate a socket address to a node
1584
name and service location, all of which are defined as in
1587
The sa argument points to a socket address structure to be
1590
The salen argument holds the size of the socket address structure
1593
If the socket address structure contains an IPv4-mapped IPv6 address
1594
or an IPv4-compatible IPv6 address, the implementation shall extract
1595
the embedded IPv4 address and lookup the node name for that IPv4
1598
Note: The IPv6 unspecified address ("::") and the IPv6 loopback
1599
address ("::1") are not IPv4-compatible addresses. If the address
1600
is the IPv6 unspecified address ("::"), a lookup is not performed,
1601
and the [EAI_NONAME] error is returned.
1603
If the node argument is non-NULL and the nodelen argument is nonzero,
1604
then the node argument points to a buffer able to contain up to
1605
nodelen characters that receives the node name as a null-terminated
1606
string. If the node argument is NULL or the nodelen argument is
1607
zero, the node name shall not be returned. If the node's name cannot
1608
be located, the numeric form of the node's address is returned
1609
instead of its name.
1611
If the service argument is non-NULL and the servicelen argument is
1612
non-zero, then the service argument points to a buffer able to
1613
contain up to servicelen bytes that receives the service name as a
1614
null-terminated string. If the service argument is NULL or the
1615
servicelen argument is zero, the service name shall not be returned.
1616
If the service's name cannot be located, the numeric form of the
1617
service address (for example, its port number) shall be returned
1618
instead of its name.
1620
The arguments node and service cannot both be NULL.
1626
Gilligan, et al. Informational [Page 29]
1628
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1631
The flags argument is a flag that changes the default actions of the
1632
function. By default the fully-qualified domain name (FQDN) for the
1633
host shall be returned, but:
1635
- If the flag bit NI_NOFQDN is set, only the node name portion of
1636
the FQDN shall be returned for local hosts.
1638
- If the flag bit NI_NUMERICHOST is set, the numeric form of the
1639
host's address shall be returned instead of its name, under all
1642
- If the flag bit NI_NAMEREQD is set, an error shall be returned if
1643
the host's name cannot be located.
1645
- If the flag bit NI_NUMERICSERV is set, the numeric form of the
1646
service address shall be returned (for example, its port number)
1647
instead of its name, under all circumstances.
1649
- If the flag bit NI_DGRAM is set, this indicates that the service
1650
is a datagram service (SOCK_DGRAM). The default behavior shall
1651
assume that the service is a stream service (SOCK_STREAM).
1655
1. The NI_NUMERICxxx flags are required to support the "-n" flags
1656
that many commands provide.
1658
2. The NI_DGRAM flag is required for the few AF_INET and AF_INET6
1659
port numbers (for example, [512,514]) that represent different
1660
services for UDP and TCP.
1662
The getnameinfo() function shall be thread safe.
1664
A zero return value for getnameinfo() indicates successful
1665
completion; a non-zero return value indicates failure.
1667
Upon successful completion, getnameinfo() shall return the node and
1668
service names, if requested, in the buffers provided. The returned
1669
names are always null-terminated strings.
1682
Gilligan, et al. Informational [Page 30]
1684
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1687
Error Return Values:
1689
The getnameinfo() function shall fail and return the corresponding
1692
[EAI_AGAIN] The name could not be resolved at this time.
1693
Future attempts may succeed.
1695
[EAI_BADFLAGS] The flags had an invalid value.
1697
[EAI_FAIL] A non-recoverable error occurred.
1699
[EAI_FAMILY] The address family was not recognized or the address
1700
length was invalid for the specified family.
1702
[EAI_MEMORY] There was a memory allocation failure.
1704
[EAI_NONAME] The name does not resolve for the supplied parameters.
1705
NI_NAMEREQD is set and the host's name cannot be
1706
located, or both nodename and servname were null.
1708
[EAI_OVERFLOW] An argument buffer overflowed.
1710
[EAI_SYSTEM] A system error occurred. The error code can be found
1713
6.3 Address Conversion Functions
1715
The two IPv4 functions inet_addr() and inet_ntoa() convert an IPv4
1716
address between binary and text form. IPv6 applications need similar
1717
functions. The following two functions convert both IPv6 and IPv4
1720
#include <arpa/inet.h>
1722
int inet_pton(int af, const char *src, void *dst);
1724
const char *inet_ntop(int af, const void *src,
1725
char *dst, socklen_t size);
1727
The inet_pton() function shall convert an address in its standard
1728
text presentation form into its numeric binary form. The af argument
1729
shall specify the family of the address. The AF_INET and AF_INET6
1730
address families shall be supported. The src argument points to the
1731
string being passed in. The dst argument points to a buffer into
1732
which the function stores the numeric address; this shall be large
1733
enough to hold the numeric address (32 bits for AF_INET, 128 bits for
1734
AF_INET6). The inet_pton() function shall return 1 if the conversion
1738
Gilligan, et al. Informational [Page 31]
1740
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1743
succeeds, with the address pointed to by dst in network byte order.
1744
It shall return 0 if the input is not a valid IPv4 dotted-decimal
1745
string or a valid IPv6 address string, or -1 with errno set to
1746
EAFNOSUPPORT if the af argument is unknown.
1748
If the af argument of inet_pton() is AF_INET, the src string shall be
1749
in the standard IPv4 dotted-decimal form:
1753
where "ddd" is a one to three digit decimal number between 0 and 255.
1754
The inet_pton() function does not accept other formats (such as the
1755
octal numbers, hexadecimal numbers, and fewer than four numbers that
1756
inet_addr() accepts).
1758
If the af argument of inet_pton() is AF_INET6, the src string shall
1759
be in one of the standard IPv6 text forms defined in Section 2.2 of
1760
the addressing architecture specification [2].
1762
The inet_ntop() function shall convert a numeric address into a text
1763
string suitable for presentation. The af argument shall specify the
1764
family of the address. This can be AF_INET or AF_INET6. The src
1765
argument points to a buffer holding an IPv4 address if the af
1766
argument is AF_INET, or an IPv6 address if the af argument is
1767
AF_INET6; the address must be in network byte order. The dst
1768
argument points to a buffer where the function stores the resulting
1769
text string; it shall not be NULL. The size argument specifies the
1770
size of this buffer, which shall be large enough to hold the text
1771
string (INET_ADDRSTRLEN characters for IPv4, INET6_ADDRSTRLEN
1772
characters for IPv6).
1774
In order to allow applications to easily declare buffers of the
1775
proper size to store IPv4 and IPv6 addresses in string form, the
1776
following two constants are defined in <netinet/in.h>:
1778
#define INET_ADDRSTRLEN 16
1779
#define INET6_ADDRSTRLEN 46
1781
The inet_ntop() function shall return a pointer to the buffer
1782
containing the text string if the conversion succeeds, and NULL
1783
otherwise. Upon failure, errno is set to EAFNOSUPPORT if the af
1784
argument is invalid or ENOSPC if the size of the result buffer is
1794
Gilligan, et al. Informational [Page 32]
1796
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1799
6.4 Address Testing Macros
1801
The following macros can be used to test for special IPv6 addresses.
1803
#include <netinet/in.h>
1805
int IN6_IS_ADDR_UNSPECIFIED (const struct in6_addr *);
1806
int IN6_IS_ADDR_LOOPBACK (const struct in6_addr *);
1807
int IN6_IS_ADDR_MULTICAST (const struct in6_addr *);
1808
int IN6_IS_ADDR_LINKLOCAL (const struct in6_addr *);
1809
int IN6_IS_ADDR_SITELOCAL (const struct in6_addr *);
1810
int IN6_IS_ADDR_V4MAPPED (const struct in6_addr *);
1811
int IN6_IS_ADDR_V4COMPAT (const struct in6_addr *);
1813
int IN6_IS_ADDR_MC_NODELOCAL(const struct in6_addr *);
1814
int IN6_IS_ADDR_MC_LINKLOCAL(const struct in6_addr *);
1815
int IN6_IS_ADDR_MC_SITELOCAL(const struct in6_addr *);
1816
int IN6_IS_ADDR_MC_ORGLOCAL (const struct in6_addr *);
1817
int IN6_IS_ADDR_MC_GLOBAL (const struct in6_addr *);
1819
The first seven macros return true if the address is of the specified
1820
type, or false otherwise. The last five test the scope of a
1821
multicast address and return true if the address is a multicast
1822
address of the specified scope or false if the address is either not
1823
a multicast address or not of the specified scope.
1825
Note that IN6_IS_ADDR_LINKLOCAL and IN6_IS_ADDR_SITELOCAL return true
1826
only for the two types of local-use IPv6 unicast addresses (Link-
1827
Local and Site-Local) defined in [2], and that by this definition,
1828
the IN6_IS_ADDR_LINKLOCAL macro returns false for the IPv6 loopback
1829
address (::1). These two macros do not return true for IPv6
1830
multicast addresses of either link-local scope or site-local scope.
1832
7. Summary of New Definitions
1834
The following list summarizes the constants, structure, and extern
1835
definitions discussed in this memo, sorted by header.
1837
<net/if.h> IF_NAMESIZE
1838
<net/if.h> struct if_nameindex{};
1840
<netdb.h> AI_ADDRCONFIG
1842
<netdb.h> AI_CANONNAME
1843
<netdb.h> AI_NUMERICHOST
1844
<netdb.h> AI_NUMERICSERV
1845
<netdb.h> AI_PASSIVE
1846
<netdb.h> AI_V4MAPPED
1850
Gilligan, et al. Informational [Page 33]
1852
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1856
<netdb.h> EAI_BADFLAGS
1858
<netdb.h> EAI_FAMILY
1859
<netdb.h> EAI_MEMORY
1860
<netdb.h> EAI_NONAME
1861
<netdb.h> EAI_OVERFLOW
1862
<netdb.h> EAI_SERVICE
1863
<netdb.h> EAI_SOCKTYPE
1864
<netdb.h> EAI_SYSTEM
1866
<netdb.h> NI_NAMEREQD
1868
<netdb.h> NI_NUMERICHOST
1869
<netdb.h> NI_NUMERICSERV
1870
<netdb.h> struct addrinfo{};
1872
<netinet/in.h> IN6ADDR_ANY_INIT
1873
<netinet/in.h> IN6ADDR_LOOPBACK_INIT
1874
<netinet/in.h> INET6_ADDRSTRLEN
1875
<netinet/in.h> INET_ADDRSTRLEN
1876
<netinet/in.h> IPPROTO_IPV6
1877
<netinet/in.h> IPV6_JOIN_GROUP
1878
<netinet/in.h> IPV6_LEAVE_GROUP
1879
<netinet/in.h> IPV6_MULTICAST_HOPS
1880
<netinet/in.h> IPV6_MULTICAST_IF
1881
<netinet/in.h> IPV6_MULTICAST_LOOP
1882
<netinet/in.h> IPV6_UNICAST_HOPS
1883
<netinet/in.h> IPV6_V6ONLY
1884
<netinet/in.h> SIN6_LEN
1885
<netinet/in.h> extern const struct in6_addr in6addr_any;
1886
<netinet/in.h> extern const struct in6_addr in6addr_loopback;
1887
<netinet/in.h> struct in6_addr{};
1888
<netinet/in.h> struct ipv6_mreq{};
1889
<netinet/in.h> struct sockaddr_in6{};
1891
<sys/socket.h> AF_INET6
1892
<sys/socket.h> PF_INET6
1893
<sys/socket.h> struct sockaddr_storage;
1895
The following list summarizes the function and macro prototypes
1896
discussed in this memo, sorted by header.
1898
<arpa/inet.h> int inet_pton(int, const char *, void *);
1899
<arpa/inet.h> const char *inet_ntop(int, const void *,
1906
Gilligan, et al. Informational [Page 34]
1908
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1911
<net/if.h> char *if_indextoname(unsigned int, char *);
1912
<net/if.h> unsigned int if_nametoindex(const char *);
1913
<net/if.h> void if_freenameindex(struct if_nameindex *);
1914
<net/if.h> struct if_nameindex *if_nameindex(void);
1916
<netdb.h> int getaddrinfo(const char *, const char *,
1917
const struct addrinfo *,
1918
struct addrinfo **);
1919
<netdb.h> int getnameinfo(const struct sockaddr *, socklen_t,
1920
char *, socklen_t, char *, socklen_t, int);
1921
<netdb.h> void freeaddrinfo(struct addrinfo *);
1922
<netdb.h> const char *gai_strerror(int);
1924
<netinet/in.h> int IN6_IS_ADDR_LINKLOCAL(const struct in6_addr *);
1925
<netinet/in.h> int IN6_IS_ADDR_LOOPBACK(const struct in6_addr *);
1926
<netinet/in.h> int IN6_IS_ADDR_MC_GLOBAL(const struct in6_addr *);
1927
<netinet/in.h> int IN6_IS_ADDR_MC_LINKLOCAL(const struct in6_addr *);
1928
<netinet/in.h> int IN6_IS_ADDR_MC_NODELOCAL(const struct in6_addr *);
1929
<netinet/in.h> int IN6_IS_ADDR_MC_ORGLOCAL(const struct in6_addr *);
1930
<netinet/in.h> int IN6_IS_ADDR_MC_SITELOCAL(const struct in6_addr *);
1931
<netinet/in.h> int IN6_IS_ADDR_MULTICAST(const struct in6_addr *);
1932
<netinet/in.h> int IN6_IS_ADDR_SITELOCAL(const struct in6_addr *);
1933
<netinet/in.h> int IN6_IS_ADDR_UNSPECIFIED(const struct in6_addr *);
1934
<netinet/in.h> int IN6_IS_ADDR_V4COMPAT(const struct in6_addr *);
1935
<netinet/in.h> int IN6_IS_ADDR_V4MAPPED(const struct in6_addr *);
1937
8. Security Considerations
1939
IPv6 provides a number of new security mechanisms, many of which need
1940
to be accessible to applications. Companion memos detailing the
1941
extensions to the socket interfaces to support IPv6 security are
1944
9. Changes from RFC 2553
1946
1. Add brief description of the history of this API and its relation
1947
to the Open Group/IEEE/ISO standards.
1949
2. Alignments with [3].
1951
3. Removed all references to getipnodebyname() and getipnodebyaddr(),
1952
which are deprecated in favor of getaddrinfo() and getnameinfo().
1954
4. Added IPV6_V6ONLY IP level socket option to permit nodes to not
1955
process IPv4 packets as IPv4 Mapped addresses in implementations.
1957
5. Added SIIT to references and added new contributors.
1962
Gilligan, et al. Informational [Page 35]
1964
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
1967
6. In previous versions of this specification, the sin6_flowinfo
1968
field was associated with the IPv6 traffic class and flow label,
1969
but its usage was not completely specified. The complete
1970
definition of the sin6_flowinfo field, including its association
1971
with the traffic class or flow label, is now deferred to a future
1976
This specification's evolution and completeness were significantly
1977
influenced by the efforts of Richard Stevens, who has passed on.
1978
Richard's wisdom and talent made the specification what it is today.
1979
The co-authors will long think of Richard with great respect.
1981
Thanks to the many people who made suggestions and provided feedback
1982
to this document, including:
1984
Werner Almesberger, Ran Atkinson, Fred Baker, Dave Borman, Andrew
1985
Cherenson, Alex Conta, Alan Cox, Steve Deering, Richard Draves,
1986
Francis Dupont, Robert Elz, Brian Haberman, Jun-ichiro itojun Hagino,
1987
Marc Hasson, Tom Herbert, Bob Hinden, Wan-Yen Hsu, Christian Huitema,
1988
Koji Imada, Markus Jork, Ron Lee, Alan Lloyd, Charles Lynn, Dan
1989
McDonald, Dave Mitton, Finnbarr Murphy, Thomas Narten, Josh Osborne,
1990
Craig Partridge, Jean-Luc Richier, Bill Sommerfield, Erik Scoredos,
1991
Keith Sklower, JINMEI Tatuya, Dave Thaler, Matt Thomas, Harvey
1992
Thompson, Dean D. Throop, Karen Tracey, Glenn Trewitt, Paul Vixie,
1993
David Waitzman, Carl Williams, Kazu Yamamoto, Vlad Yasevich, Stig
1994
Venaas, and Brian Zill.
1996
The getaddrinfo() and getnameinfo() functions are taken from an
1997
earlier document by Keith Sklower. As noted in that document,
1998
William Durst, Steven Wise, Michael Karels, and Eric Allman provided
1999
many useful discussions on the subject of protocol-independent name-
2000
to-address translation, and reviewed early versions of Keith
2001
Sklower's original proposal. Eric Allman implemented the first
2002
prototype of getaddrinfo(). The observation that specifying the pair
2003
of name and service would suffice for connecting to a service
2004
independent of protocol details was made by Marshall Rose in a
2005
proposal to X/Open for a "Uniform Network Interface".
2007
Craig Metz, Jack McCann, Erik Nordmark, Tim Hartrick, and Mukesh
2008
Kacker made many contributions to this document. Ramesh Govindan
2009
made a number of contributions and co-authored an earlier version of
2018
Gilligan, et al. Informational [Page 36]
2020
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
2025
[1] Deering, S. and R. Hinden, "Internet Protocol, Version 6 (IPv6)
2026
Specification", RFC 2460, December 1998.
2028
[2] Hinden, R. and S. Deering, "IP Version 6 Addressing
2029
Architecture", RFC 2373, July 1998.
2031
[3] IEEE Std. 1003.1-2001 Standard for Information Technology --
2032
Portable Operating System Interface (POSIX). Open Group
2033
Technical Standard: Base Specifications, Issue 6, December 2001.
2034
ISO/IEC 9945:2002. http://www.opengroup.org/austin
2036
[4] Stevens, W. and M. Thomas, "Advanced Sockets API for IPv6", RFC
2037
2292, February 1998.
2039
[5] Nordmark, E., "Stateless IP/ICMP Translation Algorithm (SIIT)",
2040
RFC 2765, February 2000.
2042
[6] The Open Group Base Working Group
2043
http://www.opengroup.org/platform/base.html
2074
Gilligan, et al. Informational [Page 37]
2076
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
2079
12. Authors' Addresses
2087
EMail: gilligan@intransa.com
2092
499 Thornall Street, 8th floor
2096
EMail: sethomso@cisco.com
2100
Hewlett-Packard Company
2101
110 Spitbrook Road ZKO3-3/W20
2105
EMail: Jim.Bound@hp.com
2109
Hewlett-Packard Company
2110
110 Spitbrook Road ZKO3-3/W20
2114
EMail: Jack.McCann@hp.com
2130
Gilligan, et al. Informational [Page 38]
2132
RFC 3493 Basic Socket Interface Extensions for IPv6 February 2003
2135
13. Full Copyright Statement
2137
Copyright (C) The Internet Society (2003). All Rights Reserved.
2139
This document and translations of it may be copied and furnished to
2140
others, and derivative works that comment on or otherwise explain it
2141
or assist in its implementation may be prepared, copied, published
2142
and distributed, in whole or in part, without restriction of any
2143
kind, provided that the above copyright notice and this paragraph are
2144
included on all such copies and derivative works. However, this
2145
document itself may not be modified in any way, such as by removing
2146
the copyright notice or references to the Internet Society or other
2147
Internet organizations, except as needed for the purpose of
2148
developing Internet standards in which case the procedures for
2149
copyrights defined in the Internet Standards process must be
2150
followed, or as required to translate it into languages other than
2153
The limited permissions granted above are perpetual and will not be
2154
revoked by the Internet Society or its successors or assigns.
2156
This document and the information contained herein is provided on an
2157
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
2158
TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
2159
BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
2160
HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
2161
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
2165
Funding for the RFC Editor function is currently provided by the
2186
Gilligan, et al. Informational [Page 39]