NAME
lmc
- device driver for
LMC
(and some
SBE)
wide-area network interface cards
SYNOPSIS
FreeBSD Synopsis
The driver can be loaded into a kernel at boot-time by adding
-
if_lmc_load="YES"
to
/boot/loader.conf
;
see
loader.conf(5).
The driver can be loaded into a running kernel after boot-time with
kldload(8).
-
kldload if_lmc
securelevel
must be 0 to load modules after boot-time; see
init(8).
The driver can be built into a kernel by adding the following to
/sys/ARCH/conf/YOURKERNEL
:
-
device lmc
device sppp
device bpf
device pf # for altq
options ALTQ
options ALTQ_HFSC # for altq example
options NETGRAPH
options DEVICE_POLLING
The driver can send and receive raw IP packets even if
neither SPPP nor Netgraph are configured into the kernel.
NetBSD & OpenBSD Synopsis
This driver is built into the GENERIC kernel so it should "just work".
The driver can be loaded into a running kernel with
modload(8).
-
modload if_lmc.o
The kernel must be built with support for Loadable Kernel Modules,
lkm(8).
-
options LKM
option LKM
securelevel
must be 0 to load modules; see
init(8).
The driver can be built into a kernel by adding the following to
/sys/arch/ARCH/conf/YOURKERNEL
:
-
lmc* at pci?
options ALTQ
options ALTQ_HFSC # for altq example
option ALTQ
option ALTQ_HFSC # for altq example
pseudo-device sppp
pseudo-device bpfilter
The driver can send and receive raw IP packets even if
SPPP is not configured into the kernel.
BSD/OS Synopsis
The driver can be built into a custom kernel by adding the following to
/sys/i386/conf/YOURKERNEL
:
-
lmc* at pci?
options PPP
options FR
options CISCO_HDLC
pseudo-device bpfilter 16
The driver can send and receive raw IP packets even if
P2P is not configured into the kernel.
Linux Synopsis
Configure the driver and line protocol for your kernel with
-
make menuconfig
Navigating through the menus, select this device driver and the
GenericHDLC layer as loadable kernel modules or built into the kernel:
-
Device Drivers --->
Networking Support --->
Wan interfaces --->
Selecting LanMedia support selects the GenericHDLC and SyncPPP layers:
-
LanMedia (and some SBE) WAN card support
[*] Use NAPI to prevent livelock
[*] Use the Generic HDLC layer
[*] Use the SyncPPP layer
Selecting the GenericHDLC layer selects these protocols:
-
Generic HDLC layer
[*] Synchronous Point-to-Point Protocol (PPP) support
[*] Cisco-HDLC support
[*] Frame Relay support
[*] Raw HDLC support
[*] Raw HDLC Ethernet device support
Selecting the SyncPPP layer includes support
for PPP and Cisco-HDLC protocols.
-
SyncPPP layer
If configured as above, the kernel will recognize an LMC card
when it boots and load this driver and the GenericHDLC code.
Messages similar to the following will appear in
/var/log/messages
:
-
Cronyx Ltd, Synchronous PPP and CISCO HDLC (c) 1994
Linux port (c) 1998 Building Number Three Ltd & Jan 'Yenya' Kasprzak.
HDLC support module revision 1.18
LMC Driver version 2006/1/20; Options NAPI GEN_HDLC SYNC_PPP.
hdlc0: T3 card> at pci 0000:00:13.0 irq 11
The driver can send and receive raw IP packets even if
neither SyncPPP nor GenericHDLC are configured into the kernel.
The device name will be
hdlc
if GenericHDLC layer is configured, or
lmc
otherwise.
The driver accepts two optional parameters:
- debug
-
Sets the driver's
debug
flag at initialization time--earlier
than it can be set using
lmcconfig
or
ifconfig.
- verbose
-
Enables additional log messages describing the
hardware while booting or module-loading.
If the driver is built into the kernel,
parameters are passed on the kernel command line:
-
if_lmc.debug=1
If the driver is loaded into the kernel as a module,
parameters are passed in the
modprobe(8)
command:
-
modprobe if_lmc debug=1
DESCRIPTION
This is an open-source
Unix
device driver for PCI-bus wide-area network interface cards.
It sends and receives packets
in HDLC frames over synchronous circuits.
A computer plus
UNIX
plus some
LMC
cards makes an
open
wide-area network router.
The
lmc
driver works with
FreeBSD,
NetBSD,
OpenBSD,
BSD/OS,
and
Linux
OSs.
It has been tested on
i386
(SMP 32-bit little-end),
PowerPC
(32-bit big-end),
Alpha
(64-bit little-end), and
Sparc
(64-bit big-end) architectures.
The
lmc
driver works with the following cards:
- LMC5200
-
HSSI--High Speed Serial Interface,
EIA612/613, 50-pin connector,
0 to 52 Mb/s, DTE only.
- LMC5245
-
T3, 2xBNC conns, 75 ohm
C-Parity or M13 Framing,
DSX-3 up to 910 ft.
- LMC1000
-
SSI--Synchronous Serial Interface,
V.35, X.21, EIA449, EIA530(A), EIA232,
0 to 10 Mb/s, DTE or DCE.
- LMC1200
-
T1/E1, RJ45 conn, 100 or 120 ohms,
T1-B8ZS-ESF, T1-AMI-SF, E1-HDB3-many,
DSX-1 up to 1500 ft; CSU up to 6 Kft.
LMC
cards contain a high-performance
PCI
interface, an
HDLC
function and
either integrated
modems
(T1, T3) or
modem
interfaces (HSSI and SSI).
- PCI
-
The PCI interface is a
DEC 21140A Tulip
Fast Ethernet chip.
This chip has an efficient PCI implementation with scatter/gather DMA,
and can run at 100 Mb/s full duplex (twice as fast as needed here).
- HDLC
-
The HDLC functions (ISO-3309: flags, bit-stuffing, CRC) are implemented
in a Field Programmable Gate Array (FPGA) which talks to the Ethernet
chip through a Media Independent Interface (MII).
The hardware in the FPGA translates between Ethernet packets and
HDLC frames on-the-fly; think of it as a WAN PHY chip for Ethernet.
- Modem
-
The modem chips are the main differences between cards.
HSSI cards use ECL10K chips to implement the EIA-612/613 interface.
T3 cards use a
TranSwitch TXC-03401
framer chip.
SSI cards use
Linear Technology LTC1343
modem interface chips.
T1 cards use a
BrookTree/Conexant/Mindspeed Bt8370
framer and line interface chip.
Line protocol stacks exist above device drivers
and below internet protocol stacks.
They typically encapsulate packets in HDLC frames and deal with
higher-level issues like protocol multiplexing and security.
The driver is compatible with several line protocol stacks:
- Netgraph
-
Netgraph(4)
implements many basic packet-handling functions as kernel loadable modules.
They can be interconnected in a graph to implement many protocols.
Configuration is done from userland without rebuilding the kernel.
ASCII configuration control messages are
not
currently supported.
- SPPP
-
sppp(4)
implements Synchronous-PPP, Cisco-HDLC and Frame-Relay in the kernel.
- SPPP
-
sppp(4)
implements Synchronous-PPP and Cisco-HDLC in the kernel.
- P2P
-
p2p(4)
implements Synchronous-PPP, Cisco-HDLC and Frame-Relay in the kernel.
- GenericHDLC
-
implements Synchronous-PPP, Cisco-HDLC, Frame-Relay,
Ether-in-HDLC, IPv4-in-HDLC, and X.25 in the kernel.
- SyncPPP
-
implements Synchronous-PPP and Cisco-HDLC in the kernel.
GenericHDLC uses SyncPPP as it's in-kernel PPP implementation.
- RawIP
-
The null line protocol, built into the driver, sends and receives
raw IPv4 and IPv6 packets in HDLC frames with no extra bytes of
overhead and no state at the end points.
EXAMPLES
ifconfig and lmcconfig
The program
lmcconfig(8)
manipulates interface parameters beyond the scope of
ifconfig(8).
lmcconfig
has many flags and options,
but in normal operation only a few are needed.
-
lmcconfig lmc0
displays interface configuration and status.
-
lmcconfig lmc0 -X 1
selects the built-in RawIP line protocol stack.
-
lmcconfig lmc0 -X 2 -x 2
selects the SPPP stack and the PPP protocol.
-
lmcconfig lmc0 -X 3 -x 2
selects the P2P stack and the PPP protocol.
-
lmcconfig lmc0 -X 4 -x 2
selects the GenericHDLC stack and the PPP protocol.
-
lmcconfig lmc0 -X 5 -x 2
selects the SyncPPP stack and the PPP protocol.
-
lmcconfig lmc0 -X 6
selects the Netgraph stack.
Some configuration options are available through
ifconfig
as well as
lmcconfig.
-
ifconfig -m lmc0
lists the available media options.
-
ifconfig lmc0 mediaopt loopback
loops the interface transmitter to the receiver for testing.
This loopback uses a path present in every card type.
lmcconfig
can select card-specific loopbacks, such as outbound payload loopback.
-
ifconfig lmc0 media e1 timeslot all
selects E1 format using all 32 timeslots.
-
ifconfig lmc0 mediaopt ppp
selects Point-to-Point Protocol.
-
ifconfig lmc0 mediaopt master
selects the on-board crystal oscillator as the transmitter clock.
-
ifconfig lmc0 debug
enables debugging output from the device driver and from
the line protocol stack above it.
-
lmcconfig lmc0 -D
enables debugging output from the device driver.
Debugging messages that appear on the console are also
written to file
/var/log/messages
.
Caution:
when
things go very wrong, a torrent of debugging messages
can swamp the console and bring a machine to its knees.
FreeBSD Operation
Configure a PPP link using SPPP and Netgraph with
-
lmcconfig lmc0 -X 6
ngctl mkpeer lmc0: sppp rawdata downstream
ifconfig sppp0 10.0.0.1 10.0.0.2
Configure a PPP link using only SPPP with
-
lmcconfig lmc0 -X 2 -x 2
ifconfig lmc0 10.0.0.1 10.0.0.2
Configure a Cisco-HDLC link using SPPP and Netgraph with
-
lmcconfig lmc0 -X 6
ngctl mkpeer lmc0: sppp rawdata downstream
ifconfig sppp0 10.0.0.1 10.0.0.2 link2
Configure a Cisco-HDLC link using only SPPP with
-
lmcconfig lmc0 -X 2 -x 3
ifconfig lmc0 10.0.0.1 10.0.0.2
Configure a Cisco-HDLC link using only Netgraph with
-
lmcconfig lmc0 -X 6
ngctl mkpeer lmc0: cisco rawdata downstream
ngctl mkpeer lmc0:rawdata iface inet inet
ifconfig ng0 10.0.0.1 10.0.0.2
Configure a Frame-Relay DTE link using SPPP with
-
lmcconfig lmc0 -X 2 -x 4
ifconfig lmc0 10.0.0.1 10.0.0.2
SPPP implements the ANSI
link management interface (LMI).
Configure a Frame-Relay DTE link using Netgraph with
-
lmcconfig lmc0 -X 6
ngctl mkpeer lmc0: frame_relay rawdata downstream
ngctl mkpeer lmc0:rawdata lmi dlci0 auto0
ngctl connect lmc0:rawdata dlci0 dlci1023 auto1023
ngctl mkpeer lmc0:rawdata rfc1490 dlci500 downstream
ngctl mkpeer lmc0:rawdata.dlci500 iface inet inet
ifconfig ng0 10.0.0.1 10.0.0.2
Netgraph implements ANSI, ITU, and FRIF
link management interfaces (LMIs).
Configure a RAWIP link using only the driver with
-
lmcconfig lmc0 -X 1
ifconfig lmc0 10.0.0.1 10.0.0.2
Configure a RAWIP link using Netgraph with
-
lmcconfig lmc0 -X 6
ngctl mkpeer lmc0: iface rawdata inet
ifconfig ng0 10.0.0.1 10.0.0.2
NetBSD & OpenBSD Operation
Configure a PPP link using SPPP with
-
lmcconfig lmc0 -X 2 -x 2
ifconfig lmc0 10.0.0.1 10.0.0.2
Configure a Cisco-HDLC link using SPPP with
-
lmcconfig lmc0 -X 2 -x 3
ifconfig lmc0 10.0.0.1 10.0.0.2
Configure a RAWIP link with
-
lmcconfig lmc0 -X 1
ifconfig lmc0 10.0.0.1 10.0.0.2
BSD/OS Operation
Configure a PPP link using P2P by creating file
/etc/ppp.sys
containing
-
Plmc0: :device=lmc0:
:local-addr=10.0.0.1:
:remote-addr=10.0.0.2:
:immediate:dialout:direct:
:-pfc:-acfc:-tcpc:
Then run
ppp(8):
-
ppp -bd Plmc0
Add
-X debug-all
to watch protocol events happen.
Configure a Cisco-HDLC link by setting LINKTYPE with
-
ifconfig lmc0 10.0.0.1 10.0.0.2 linktype chdlc
Configure a Fame-Relay link with
-
ifconfig lmc0 linktype fr
frconfig lmc0 lmi ansi [type dce]
frconfig lmc0 dlci 500 10.0.0.2
ifconfig lmc0 10.0.0.1 10.0.0.2
Adding
``type dce''
after
``ansi''
configures it as a DCE (switch).
P2P implements ANSI, ITU and FRIF
link management interfaces (LMIs).
Configure a RAWIP link with
-
lmcconfig lmc0 -X 1
ifconfig lmc0 10.0.0.1 10.0.0.2
Linux operation
The
sethdlc
program configures the GenericHDLC code.
-
sethdlc hdlc0
(or
pvc0
for frame relay)
displays the current settings of a given device.
Note that
sethdlc
must be run before
ifconfig.
Sethdlc
and the GenericHDLC kernel code are documented in
/usr/src/linux/Documentation/networking/generic-hdlc.txt
and at
http://www.kernel.org:/pub/linux/utils/net/hdlc
Configure a PPP link using GenericHDLC with
-
lmcconfig lmc0 -X 4 -x 2
sethdlc hdlc0 ppp
ifconfig hdlc0 10.0.0.1 pointopoint 10.0.0.2
Configure a PPP link using SyncPPP with
-
lmcconfig hdlc0 -X 5 -x 2
ifconfig hdlc0 10.0.0.1 pointopoint 10.0.0.2
Configure a Cisco-HDLC link using GenericHDLC with
-
lmcconfig lmc0 -X 4
sethdlc hdlc0 cisco
ifconfig hdlc0 10.0.0.1 pointopoint 10.0.0.2
Configure a Cisco-HDLC link using SyncPPP with
-
lmcconfig hdlc0 -X 5 -x 3
ifconfig hdlc0 10.0.0.1 pointopoint 10.0.0.2
Configure a Frame-Relay DTE link using GenericHDLC with
-
lmcconfig lmc0 -X 4
sethdlc hdlc0 fr lmi ansi [dce]
sethdlc hdlc0 create 500
ifconfig hdlc0 up
ifconfig pvc0 10.0.0.1 pointopoint 10.0.0.2
Adding
``dce''
after
``ansi''
configures it as a DCE (switch).
GenericHDLC implements ANSI and ITU
link management interfaces (LMIs).
Configure a RAWIP link using GenericHDLC with
-
lmcconfig lmc0 -X 4
sethdlc hdlc0 hdlc
ifconfig hdlc0 10.0.0.1 pointopoint 10.0.0.2
Configure a RAWIP link using only the driver with
-
lmcconfig hdlc0 -X 1
ifconfig hdlc0 10.0.0.1 pointopoint 10.0.0.2
TESTING
Testing with Loopbacks
Testing with loopbacks requires only one card and
can test everything on that card.
Packets can be looped back at many points: in the PCI chip,
in the modem chips, through a loopback plug, in the
local external equipment, or at the far end of a circuit.
All cards can be looped through the PCI chip.
Cards with internal modems can be looped through
the modem framer and the modem line interface.
Cards for external modems can be looped through
the driver/receiver chips.
See
lmcconfig(8)
for details.
Configure the card with
-
ifconfig lmc0 10.0.0.1 10.0.0.1
- HSSI
-
Loopback plugs can be ordered from SBE (and others).
Transmit clock is normally supplied by the external modem.
When an HSSI card is operated with a loopback plug, the PCI bus
clock must be used as the transmit clock, typically 33 MHz.
When testing an HSSI card with a loopback plug,
configure it with
-
lmcconfig lmc0 -a 2
``-a
2
''
selects the PCI bus clock as the transmit clock.
- T3
-
Connect the two BNC jacks with a short coax cable.
- SSI
-
Loopback plugs can be ordered from SBE (only).
Transmit clock is normally supplied by the external modem.
When an SSI card is operated with a loopback plug,
the on-board clock synthesizer must be used.
When testing an SSI card with a loopback plug,
configure it with
-
lmcconfig lmc0 -E -f 10000000
``-E''
puts the card in DCE mode to source a transmit clock.
``-f
10000000
''
sets the internal clock source to 10 Mb/s.
- T1/E1
-
A loopback plug is a modular plug with two wires
connecting pin 1 to pin 4 and pin 2 to pin 5.
One can also test by connecting to a local modem (HSSI and SSI)
or NI (T1 and T3) configured to loop back.
Cards can generate signals to loopback remote equipment
so that complete circuits can be tested; see
lmcconfig(8)
for details.
Testing with a Modem
Testing with a modem requires two cards of different types.
The cards can be in the same machine or different machines.
Configure the two cards with
-
ifconfig lmc0 10.0.0.1 10.0.0.2
ifconfig lmc1 10.0.0.2 10.0.0.1
- T3/HSSI
-
If you have a T3 modem with an HSSI interface
(made by Digital Link, Larscom, Kentrox etc.)
then use an HSSI card and a T3 card.
The coax cables between the card and the modem
must
``cross over''
(see below).
- T1/V.35
-
If you have a T1 (or E1) modem with a V.35, X.21 or EIA530 interface,
then use an SSI card and a T1 card.
Use a T1 null modem cable (see below) between
the external modem and the T1 card.
Testing with a Null Modem Cable
Testing with a null modem cable requires two cards of the same type.
The cards can be in the same machine or different machines.
Configure the two cards with
-
ifconfig lmc0 10.0.0.1 10.0.0.2
ifconfig lmc1 10.0.0.2 10.0.0.1
- HSSI
-
Three-meter HSSI null-modem cables can be ordered from SBE.
In a pinch, a 50-pin SCSI-II cable up to a few meters will
work as a straight HSSI cable (not a null modem cable).
Longer cables should be purpose-built HSSI cables because
the cable impedance is different.
Transmit clock is normally supplied by the external modem.
When an HSSI card is connected by a null modem cable, the PCI bus
clock can be used as the transmit clock, typically 33 MHz.
When testing an HSSI card with a null modem cable,
configure it with
-
lmcconfig lmc0 -a 2
``-a
2
''
selects the PCI bus clock as the transmit clock.
- T3
-
T3 null modem cables are just 75-ohm coax cables with BNC connectors.
TX OUT on one card should be connected to RX IN on the other card.
In a pinch, 50-ohm thin Ethernet cables
usually
work up to a few meters, but they will
not
work for longer runs--75-ohm coax is
required.
- SSI
-
Three-meter SSI null modem cables can be ordered from SBE.
An SSI null modem cable reports a cable type of V.36/EIA449.
Transmit clock is normally supplied by the external modem.
When an SSI card is connected by a null modem cable,
an on-board clock synthesizer is used.
When testing an SSI card with a null modem cable,
configure it with
-
lmcconfig lmc0 -E -f 10000000
``-E''
puts the card in DCE mode to source a transmit clock.
``-f
10000000
''
sets the internal clock source to 10 Mb/s.
- T1/E1
-
A T1 null modem cable has two twisted pairs that connect
pins 1 and 2 on one plug to pins 4 and 5 on the other plug.
Looking into the cable entry hole of a plug,
with the locking tab oriented down,
pin 1 is on the left.
A twisted pair Ethernet cable makes an excellent straight T1 cable.
Alas, Ethernet cross-over cables do not work as T1 null modem cables.
OPERATING NOTES
LEDs
HSSI and SSI cards should be operational if all three green LEDs are
on (the upper-left one should be blinking) and the red LED is off.
RED | upper-right | No Transmit clock
|
GREEN | upper-left | Device driver is alive if blinking
|
GREEN | lower-right | Modem signals are good
|
GREEN | lower-left | Cable is plugged in (SSI only)
|
T1/E1 and T3 cards should be operational if the upper-left
green LED is blinking and all other LEDs are off.
For the T3 card, if other LEDs are on or blinking,
try swapping the coax cables!
RED | upper-right | Received signal is wrong
|
GREEN | upper-left | Device driver is alive if blinking
|
BLUE | lower-right | Alarm Information Signal (AIS)
|
YELLOW | lower-left | Remote Alarm Indication (RAI)
|
RED | blinks if an outward loopback is active.
|
GREEN | blinks if the device driver is alive.
|
BLUE | blinks if sending AIS, on solid if receiving AIS.
|
YELLOW | blinks if sending RAI, on solid if receiving RAI.
|
Packet Lengths
Maximum transmit and receive packet length is unlimited.
Minimum transmit and receive packet length is one byte.
Cleaning up after one packet and setting up for the next
packet involves making several DMA references.
This can take longer than the duration of a short packet,
causing the adapter to fall behind.
For typical PCI bus traffic levels and memory system latencies,
back-to-back packets longer than about 20 bytes will always
work (53 byte cells work), but a burst of several hundred
back-to-back packets shorter than 20 bytes will cause packets
to be dropped.
This usually is not a problem since an IPv4 packet header is
at least 20 bytes long.
The device driver imposes no constraints on packet size.
Most operating systems set the default Maximum Transmission
Unit (MTU) to 1500 bytes; the legal range is usually (72..65535).
This can be changed with
-
ifconfig lmc0 mtu 2000
SPPP enforces an MTU of 1500 bytes for PPP and Cisco-HDLC.
P2P enforces an MTU of 1500 bytes for PPP and Cisco-HDLC,
and 4000 bytes for Frame Relay.
GenericHDLC enforces an MTU range of (68..1500) bytes.
RAWIP sets the default MTU to 4032 bytes,
but allows it to be changed to anything.
ALTQ: Alternate Output Queue Disciplines
The driver has hooks for
altq(9),
the Alternate Queueing package.
To see ALTQ in action, use your favorite traffic generation
program to generate three flows sending down one T3 circuit.
Without ALTQ, the speeds of the three connections will vary chaotically.
Enable ALTQ and two of the connections will run at about 20 Mb/s and
the third will run at about 2 Mb/s.
FreeBSD ALTQ example
Enable
pf(4)
and add the following lines to
/etc/pf.conf
:
-
altq on lmc0 bandwidth 44Mb hfsc queue{ a b c}
queue a on lmc0 bandwidth 48%
pass in on lmc0 queue a from 10.0.0.2 port 12345 to 10.0.0.1
pass out on lmc0 queue a from 10.0.0.1 to 10.0.0.2 port 12345
queue b on lmc0 bandwidth 48%
pass in on lmc0 queue b from 10.0.0.2 port 12346 to 10.0.0.1
pass out on lmc0 queue b from 10.0.0.1 to 10.0.0.2 port 12346
queue c on lmc0 bandwidth 4% hfsc(default)
pass in on lmc0 queue c from 10.0.0.2 port 12347 to 10.0.0.1
pass out on lmc0 queue c from 10.0.0.1 to 10.0.0.2 port 12347
NetBSD and OpenBSD ALTQ example
Enable
altqd(8)
and add the following lines to
/etc/altq.conf
:
-
interface lmc0 bandwidth 44M hfsc
class hfsc lmc0 a root pshare 48
filter lmc0 a 10.0.0.2 12345 10.0.0.1 0 6
filter lmc0 a 10.0.0.1 0 10.0.0.2 12345 6
class hfsc lmc0 b root pshare 48
filter lmc0 b 10.0.0.2 12346 10.0.0.1 0 6
filter lmc0 b 10.0.0.1 0 10.0.0.2 12346 6
class hfsc lmc0 c root pshare 4 default
filter lmc0 c 10.0.0.2 12347 10.0.0.1 0 6
filter lmc0 c 10.0.0.1 0 10.0.0.2 12347 6
The example above requires
the Packet Filter
pf(4)
and
the
altq(4)
Hierarchical Fair Service Curve
queue discipline to be configured in
conf/YOURKERNEL
:
-
device pf
option ALTQ
option ALTQ_HFSC.
options ALTQ
options ALTQ_HFSC.
BPF: Berkeley Packet Filter
The driver has hooks for
bpf(4),
the Berkeley Packet Filter, a protocol-independent
raw interface to data link layers.
To test the BPF kernel interface,
bring up a link between two machines, then run
ping(8)
and
tcpdump(1):
-
ping 10.0.0.1
and in a different window:
-
tcpdump -i lmc0
The output from tcpdump should look like this:
-
03:54:35.979965 10.0.0.2 > 10.0.0.1: icmp: echo request
03:54:35.981423 10.0.0.1 > 10.0.0.2: icmp: echo reply
Line protocol control packets may appear among the
ping packets occasionally.
The kernel must be configured with
-
device bpf
options bpfilter
option bpfilter
pseudo-device bpfilter
Device Polling
A T3 receiver can generate over 100K interrupts per second,
This can cause a system to
``live-lock'':
spend all of its time servicing interrupts.
Linux's
polling mechanism disables a card's interrupt when it interrupts,
calls the card's interrupt service routine with kernel interrupts
enabled, and then reenables the card's interrupt.
The driver is permitted to process a limited number of packets
each time it is called by the kernel.
Card interrupts are left disabled if more packets arrive than are
permitted to be processed, which in extreme cases will result in
packets being dropped in hardware at no cost to software.
Polling is enabled using
menuconfig
by selecting
-
LanMedia (and some SBE) WAN card support
[*] Use NAPI to prevent livelock
Device Polling
A T3 receiver can generate over 100K interrupts per second,
This can cause a system to
``live-lock'':
spend all of its time servicing interrupts.
FreeBSD's
polling(4)
mechanism permanently disables interrupts from the card
and instead the card's interrupt service routine is polled each
time the kernel is entered (syscall, timer interrupt, etc.)
and from the kernel idle loop; this adds some latency.
The driver is permitted to process a limited number of packets
each time it is called by the kernel.
Polling is enabled with
-
ifconfig lmc0 polling
The kernel must be configured with
-
options DEVICE_POLLING
SNMP: Simple Network Management Protocol
The driver is aware of what is required to be a Network Interface
Object managed by an Agent of the Simple Network Management Protocol.
The driver exports SNMP-formatted configuration and status
information sufficient for an SNMP Agent to create MIBs for:
- RFC-2233
-
- RFC-2496
-
- RFC-2495
-
- RFC-1659
-
An SNMP Agent is a user program, not a kernel function.
Agents can retrieve configuration and status information
by using
Netgraph control messages or
ioctl(2)
system calls.
User programs should poll
sc->cfg.ticks
which increments once per second after the SNMP state has been updated.
E1 Framing
Phone companies usually insist that customers put a
Frame Alignment Signal
(FAS) in time slot 0.
A Cyclic Redundancy Checksum (CRC) can also ride in time slot 0.
Channel Associated Signalling
(CAS) uses Time Slot 16.
In telco-speak
signalling
is on/off hook, ringing, busy, etc.
Signalling is not needed here and consumes 64 Kb/s.
Only use E1-CAS formats if the other end insists on it!
Use E1-FAS+CRC framing format on a public circuit.
Depending on the equipment installed in a private circuit,
it may be possible to use all 32 time slots for data (E1-NONE).
T3 Framing
M13 is a technique for multiplexing 28 T1s into a T3.
Muxes use the C-bits for speed-matching the tributaries.
Muxing is not needed here and usurps the FEBE and FEAC bits.
Only use T3-M13 format if the other end insists on it!
Use T3-CParity framing format if possible.
Loop Timing, Fractional T3, and HDLC packets in
the Facility Data Link are
not
supported.
T1 & T3 Frame Overhead Functions
Performance Report Messages (PRMs) are enabled in T1-ESF.
Bit Oriented Protocol (BOP) messages are enabled in T1-ESF.
In-band loopback control (framed or not) is enabled in T1-SF.
Far End Alarm and Control (FEAC) msgs are enabled in T3-CPar.
Far End Block Error (FEBE) reports are enabled in T3-CPar.
Remote Alarm Indication (RAI) is enabled in T3-Any.
Loopbacks initiated remotely time out after 300 seconds.
T1/E1 'Fractional' 64 kb/s Time Slots
T1 uses time slots 24..1; E1 uses time slots 31..0.
E1 uses TS0 for FAS overhead and TS16 for CAS overhead.
E1-NONE has
no
overhead, so all 32 TSs are available for data.
Enable/disable time slots by setting 32 1s/0s in a config param.
Enabling an E1 overhead time slot,
or enabling TS0 or TS25-TS31 for T1,
is ignored by the driver, which knows better.
The default TS param, 0xFFFFFFFF, enables the maximum number
of time slots for whatever frame format is selected.
56 Kb/s time slots are
not
supported.
SEE ALSO
tcpdump(1),
tcpdump(1),
tcpdump(1),
tcpdump(1),
ioctl(2),
bpf(4),
bpf(4),
bpf(4),
bpf(4),
de(4),
de(4),
de(4),
de(4),
kld(4),
netgraph(4),
p2p(4),
pf(4),
polling(4),
sppp(4),
sppp(4),
sppp(4),
altq.conf(5),
altq.conf(5),
loader.conf(5),
pf.conf(5),
altqd(8),
altqd(8),
frconfig(8),
ifconfig(8),
init(8),
init(8),
init(8),
kldload(8),
lkm(8),
lkm(8),
lmcconfig(8),
modload(8),
modload(8),
modprobe(8),
ngctl(8),
ping(8),
ppp(8),
tcpdump(8),
altq(9),
altq(9),
altq(9),
ifnet(9)
ifnet(9)
ifnet(9)
ifnet(9)
http://www.sbei.net/
http://www.kernel.org:/pub/linux/utils/net/hdlc
file://usr/src/linux/Documentation/networking/generic-hdlc.txt
HISTORY
Ron Crane
had the idea to use a Fast Ethernet chip as a PCI interface
and add an Ethernet-to-HDLC gate array to make a WAN card.
David Boggs
designed the Ethernet-to-HDLC gate array and PC cards.
We did this at our company,
LAN Media Corporation (LMC).
SBE Corporation
aquired
LMC
and continues to make the cards.
Since the cards use Tulip Ethernet chips, we started with
Matt Thomas'
ubiquitous
de(4)
driver.
Michael Graff
stripped out the Ethernet stuff and added HSSI stuff.
Basil Gunn
ported it to
Solaris
(lost) and
Rob Braun
ported it to
Linux.
Andrew Stanley-Jones
added support for three more cards.
David Boggs
rewrote everything and now feels responsible for it.
AUTHORS
David Boggs <boggs@boggs.palo-alto.ca.us>