DLPI(7)DLPI(7)NAMEdlpi - data link provider interface
DESCRIPTION
The DLPI is a kernel level interface, based on the STREAMS development
environment. The DLPI interface is accessed from STREAMS protocol
modules via STREAMS messages or directly from user programs via the
putmsg(2) and getmsg(2) system calls.
The Data Link Service (DLS) provider is configured as a STREAMS driver
and the DLS user accesses the provider using open(2) /dev/llc2 (llc2(7))
to establish a stream to the DLS provider. The stream acts as a
communication endpoint between a DLS user and the DLS provider. The
current implementation is a style 2 provider (a different minor device
for each DLS user) and is based on the DLPI Specification, Revision
2.0.0.
Protocol stacks register themselves with DLPI which indicates the types
of packets which the stack will process. The DLPI directs incoming
packets to the appropriate protocol stack.
Outbound packets are directed to DLPI which sends them to the appropriate
MAC layer drivers i.e Ethernet, FDDI or Token Ring.
CONFIGURATION
A.) Snif master file and interfaces
The DLPI includes three master files: /var/sysgen/master.d/llc2,
/var/sysgen/master.d/xtimer, and /var/sysgen/master.d/snif. You
should not need to change the llc2 master file and the xtimer
master file. However, the snif master file provides a mapping
between the minor number of an interface and its name. The snif
master file shipped with the DLPI subsystem has three interfaces
pre-defined: one Token Ring interface, one Ethernet interface,
and one FDDI interface. The name of the interface should match to
the name as seen by netstat -ia(1) command. The sequence of
the interfaces defined in the iftab[] table of the snif master
file determines the minor number of an interface (the major number
is always 117.)
Upon installation the DLPI creates (mknod(1M)) three devices
with minor numbers corresponding to the pre-defined interfaces
of the snif master file. If you must modify the snif master file,
be sure your interface name is a valid interface (can be seen by
the netstat -ia command) and each interface in the iftab[]
table has its corresponding device defined. Please note that the
loopback interface, 'lo0' is not a valid interface for DLPI.
For example, to create a fv1 interface update the snif master
file iftab[] table with an entry for the fv1 device and issue the
command:
Page 1
DLPI(7)DLPI(7)
# mknod /dev/fv1 c 117 3
In this example the minor number is 3 and the fv1 device would be
the fourth device listed in the iftab[] table (the device with the
minor device 0 is the first device listed in the iftab[] table.)
After the installation of the DLPI software or after the snif
master file has been modified, you must autoconfig(1M) to
create a new kernel and reboot(1M) your kernel:
# autoconfig -f
# reboot
B.) Snetd daemon and its config file
Upon system startup, if the state of the DLPI daemon snetd(1M)
is on, the daemon is started by /etc/init.d/network. You may use
chkconfig(1M) to check the state of the daemon, and use ifconfig
(1M) to change its state.
When the snetd(1M) is started, it reads a configuration from
the /var/config/snetd.options file. Shipped with the software is a
sample snetd.options file that includes the sample configuration
lines for the three interfaces pre-defined in the snif master
file. As 'sample' configuration lines, the configuration lines are
commented out with an '#' sign at column one. To activate a
particular interface, the '#' signs must be removed. For example,
if your sample snetd.options is as below (Please note that the line
numbers will not be in the shipped file but they are put in here
for reference purposes):
/*line1*/ #fv0 d /dev/fv0
/*line2*/ #ec0 d /dev/ec0
/*line3*/ #ipg0 d /dev/ipg0
/*line4*/ llc2 dc /dev/llc2
/*line5*/
/*line6*/ %%
/*line7*/
/*line8*/ #llc2 fv0 LL_SET_SNID=A \
/*line9*/ # SHELL="lltune -P -p llc2 -s A -d /dev/llc2 def.tokenring"
/*line10*/#llc2 ec0 LL_SET_SNID=B \
/*line11*/# SHELL="lltune -P -p llc2 -s B -d /dev/llc2 def.ether"
/*line12*/#llc2 ipg0 LL_SET_SNID=C \
/*line13*/# SHELL="lltune -P -p llc2 -s C -d /dev/llc2 def.fddi"
To activate the ethernet interface ec0, you need only take out
the '#' sign at the beginning of the line #2, #10, and #11.
Please also note that the LL_SET_SNID parameter is equivalent to
the Point of Physical Attachment (PPA) as referenced in the
DLPI Specification. The value 'B' is used in the DL_ATTACH_REQ
Page 2
DLPI(7)DLPI(7)
primitive, or in the case of 4DDN product, this value will be
used in the file '/var/opt/dn/dn_config' for the 'DPPA' parameter.
In general, the LL_SET_SNID parameter correspond to the associated
minor device number , ie. minor 0 is 'A', minor 1 is 'B', etc.
If you add a new interface to the snif master file, you need to
create the configuration lines for the new interface as well. For
example, if you add fv1 as shown in the example of the "Snif master
file and interfaces" section, insert the following line between
line3 and line4:
fv1 d /dev/fv1
and the next two lines after line13:
llc2 fv1 LL_SET_SNID=D \
SHELL="lltune -P -p llc2 -s D -d /dev/llc2 def.tokenring"
Please refer to the manual page snetd(1M) for more information
on the configuration format. X.25 users should refer to its release
notes for a configuration utility that writes more X.25 related
configuration information into this snetd.options file.
PRIMITIVES
The DLPI supports three modes of communication service: connection
(circuit-oriented), connectionless (message-oriented) and acknowledged
connectionless. IRIX supports both connection (DL_CODLS) service and
connectionless (DL_CLDLS) service.
The following primitives are provided and are divided into the following
phases. More details may be found in <sys/dlpi.h> and the DLPI
Specification.
A.) Local Management
1. Information Reporting Primitives
DL_INFO_REQ
DL_INFO_ACK
DL_ERROR_ACK
2. Attach Primitives
DL_ATTACH_REQ
DL_DL_DETACH_REQ
DL_OK_ACK
DL_ERROR_ACK
3. Bind Primitives
DL_BIND_REQ
DL_BIND_ACK
Page 3
DLPI(7)DLPI(7)
DL_SUBS_BIND_REQ
DL_SUBS_BIND_ACK
DL_UNBIND_REQ
DL_SUBS_UNBIND_REQ
DL_OK_ACK
DL_ERROR_ACK
4. Other Primitives
DL_ENABMULTI_REQ
DL_DISABMULTI_REQ
DL_OK_ACK
DL_ERROR_ACK
DL_PHYS_ADDR_REQ
DL_PHYS_ADDR_ACK
DL_SET_PHYS_ADDR_REQ
B.) Connection Establishment
1. Connection Establishment Primitives
DL_CONNECT_REQ
DL_CONNECT_IND
DL_CONNECT_RES
DL_CONNECT_CON
DL_DISCONNECT_REQ
DL_DISCONNECT_IND
DL_TOKEN_REQ
DL_TOKEN_ACK
DL_OK_ACK
DL_ERROR_ACK
C.) Connection-mode Data Transfer
1. Data Transfer
To send an connection-mode data frame, a message consisting of one
or more M_DATA message blocks needs to be sent to the DLS
provider. This will cause an I-frame to be sent over the network,
containing the data passed in the M_DATA message blocks.
Incoming connection-mode data will be passed upstream in
messages consisting of one or more M_DATA message blocks.
2. Reset Primitives
DL_RESET_REQ
DL_RESET_IND
DL_RESET_RES
DL_RESET_CON
DL_OK_ACK
DL_ERROR_ACK
D.) Connection Release
1. Connection Release Primitives
DL_DISCONNECT_REQ
Page 4
DLPI(7)DLPI(7)
DL_DISCONNECT_IND
DL_OK_ACK
DL_ERROR_ACK
E.) Connectionless-mode Data Transfer
1. Data Transfer Primitives
DL_UNITDATA_REQ
DL_UNITDATA_IND
2. Error Reporting Primitives
DL_UDERROR_IND
F.) XID and TEST Services
1. XID Primitives
DL_XID_REQ
DL_XID_IND
DL_XID_RES
DL_XID_CON
2. TEST Primitives
DL_TEST_REQ
DL_TEST_IND
DL_TEST_RES
DL_TEST_CON
DLSAP ADDRESS FORMATS
A DLS user's identity is established by associating it with a DLS access
point (DLSAP), which is the point through wich the user will communicate
with the DLS provider. A DLSAP is identified by a DLSAP address. A DLSAP
address contains the hardware address (usually 6 bytes in length)
followed by SAP information. Only the SAP information is needed to build
a DL_BIND_REQ primitive. The full DLSAP address is then returned to the
DLS users in the DL_BIND_ACK primitive and should be used where a DLSAP
address is required to construct a DLPI primitive as specified in the
DLPI Specification.
802.2 LLC has two modes of operation to support the various types of SAP
that a DLS user may request. The mode of operation of a DLS user is
determined by the contents of the SAP information provided in the
DL_BIND_REQ primitive. The modes of operation should not be confused
with the three modes of 802.2 LLC communication service; namely,
connection service mode, connectionless service mode and acknowledge
connectionless service mode.
A.) Normal Mode
The Normal mode of operation is when a bind is performed
with the value of the SAP information being in range 0x02
to 0xFE (a one-byte, even value). This is the SAP as
Page 5
DLPI(7)DLPI(7)
specified under 802.2 LLC, and is the only mode of operation
for the connection (i.e. LLC2) service mode. The
Sub-Network Access Protocol (SNAP) also uses this mode of
operation.
The DLSAP addresses for Normal mode have the following format:
struct llc_dlsap {
u_char dl_mac[6]; /* hardware address */
u_char dl_sap; /* LLC SAP */
};
The DLSAP address may be modified through DL_SUBS_BIND_REQ
primitive when the SNAP is used to extend the LLC header.
The extended SNAP DLSAP addresses have the following format:
struct llc_snap_dlsap {
u_char dl_mac[6]; /* hardware address */
u_char dl_sap; /* SNAP sap: 0xAA */
u_char dl_oui[3]; /* OUI information */
u_char dl_proto[2]; /* protocol ID */
};
DLS users should use llc_dlsap format in constructing the
DL_UNITDATA_REQ primitive and it is DLS users' responsibility
to put the OUI information and protocol ID in front of their
data. Upon receipt of DL_UNITDATA_IND, the DLSAP addresses
are also of llc_dlsap format. It is DLS users' responsibility
to skip the OUI information and protocol ID for users' data.
The DLSAP address may also be modified if source routing is
used for Token Ring networks through TEST and/or XID
primitives. The source routing information field (rif) is
appended to the end of the llc_dlsap format. The DL_CONNECT_*
primitives should also use this llc_sri_dlsap format when
source routing information is present. The extended SRI
DLSAP addresses have the following format:
struct llc_sri_dlsap {
u_char dl_mac[6]; /* hardware address */
u_char dl_sap; /* LLC SAP */
u_char dl_rif; /* start of rif */
};
B.) Ethernet Mode
The Ethernet mode of operation occurs when a bind is performed
for two bytes (the high byte being non-zero). When this occurs
the binding driver will be sent packets for the Ethernet types
registered for.
The DLSAP addresses for Ethernet mode have the following format:
Page 6
DLPI(7)DLPI(7)
struct llc_eth_dlsap {
u_char dl_mac[6]; /* hardware address */
u_short dl_sap; /* Ethernet SAP */
};
EXAMPLE
The example program below will format an DL_INFO_REQ message and send it
to the provider opened via the open(2) system call. It will then check
for a valid response; if one is found, it will print the DLPI provider
style to the console.
#include <stdio.h>
#include <fcntl.h>
#include <errno.h>
#include <sys/stream.h>
#include <sys/stropts.h>
#include <sys/dlpi.h>
#define DLPIDEV "/dev/llc2"
int
main(int argc, char **argv)
{
int fd;
if ((fd = open(DLPIDEV, O_RDWR)) < 0 ) {
perror ("open");
exit(0);
}
if (info(fd) < 0) {
perror ("info");
exit (0);
}
close(fd);
}
int
print_info(dl_info_ack_t *info)
{
printf("DLPI Provider Style %d\n",
(info->dl_provider_style == DL_STYLE1) ? 1 : 2);
}
int
info(int fd)
{
dl_info_req_t info_req;
dl_info_ack_t *info_ack;
Page 7
DLPI(7)DLPI(7)
dl_error_ack_t *error_ack;
struct strbuf ctlbuf;
int flags;
char buffer[BUFSIZ];
char *cp = buffer;
int len;
/* Build info_req primitive */
info_req.dl_primitive = DL_INFO_REQ;
ctlbuf.len = DL_INFO_REQ_SIZE;
ctlbuf.buf = (char *)&info_req;
flags = RS_HIPRI;
if (putmsg(fd, &ctlbuf, NULL, flags ) < 0 )
return(-1);
/*
* Now wait for the reply, this should be
* dl_info_ack or dl_error_ack
*/
ctlbuf.maxlen = sizeof(buffer);
ctlbuf.len = 0;
ctlbuf.buf = buffer;
flags = RS_HIPRI;
if (getmsg(fd, &ctlbuf, NULL, &flags) < 0)
return(-1);
/* Is what we got sensible? */
if (ctlbuf.len < sizeof(long)) {
errno = EPROTO;
return(-1);
}
/* Switch on type */
switch (*(unsigned long *)cp){
default:
errno = EPROTO;
return(-1);
case DL_INFO_ACK:
info_ack = (dl_info_ack_t *)cp;
print_info(info_ack);
return(fd);
case DL_ERROR_ACK:
if (ctlbuf.len < DL_ERROR_ACK_SIZE) {
errno = EPROTO;
return(-1);
}
error_ack = (dl_error_ack_t *)cp;
errno = error_ack->dl_unix_errno;
return(-1);
Page 8
DLPI(7)DLPI(7)
}
}
FILES
<sys/dlpi.h>,
SEE ALSOsnetd(1M), chkconfig(1M), ifconfig(1M), llc2(7)
Page 9
