Linux Device Drivers for the Radiometrix RPC Radio

Information Sciences Institute
University of Southern California
Marina del Rey, CA 90292 USA

10 July 2000

This work was supported by DARPA under grant No. DABT63-99-1-0011 as part of the SCADDS project, and was also made possible in part due to support from Cisco Systems.

Introduction

This package contains a Linux device driver for the ``RPC'' (Radio Packet Controller) model of radio manufactured by Radiometrix. The RPC is a fairly low-power, self-contained, short-range, plug-on radio. It requires only a simple antenna, 5V power supply, and interface to a byte-wide I/O port on a host microcontroller or bi-directional PC parallel port. The module provides all the RF circuits and processor intensive low level packet formatting and packet recovery functions required to inter-connect an number of microcontrollers in a radio network.

A data packet of 1 to 27 bytes downloaded by a host microcontroller into the RPC's packet buffer is transmitted by the RPC's transceiver and will ``appear'' in the receive buffer of all the RPC's within radio range.

This software package allows control of an RPC under Linux if the RPC is connected to a bi-directional PC parallel port. The driver is a Linux kernel module that implements two types of device drivers:

This Radiometrix driver was written by Jeremy Elson (jelson@circlemud.org) while at the University of Southern California's Information Sciences Institute. This work was supported by DARPA under grant No. DABT63-99-1-0011 as part of the SCADDS project, and was also made possible in part due to support from Cisco Systems. It is freely available under the GNU Public License (GPL). Up-to-date information, including the latest version of the software, is available via the SCADDS home page, or directly from the author's page.

The Physical Interface

The first step in using the Linux Radiometrix RPC Device Driver is establishing the physical connection between the RPC radio and your parallel port.

Note that this driver does not use the RPC development kit sold by Radiometrix. It assumes you have a bare RPC radio hooked directly up to the pins of the parallel port of your PC.

The default pin mappings between the RPC and the parallel port are shown below. However, if necessary, some of these can be modified by changing the #define statements in rpc_lowlevel.c. If you do change the pin mappings, be aware that they are constrained by the capabilities of the PC's parallel port, described in more detail in other documents.)

Pin Label	RPC Pin	Direction	PC (Printer Port) Pin
GND	1	--	any of 18-25
D0	2	[IMAGE png]	2
D1	3	[IMAGE png]	3
D2	4	[IMAGE png]	4
D3	5	[IMAGE png]	5
TXR	6	[IMAGE png]	16
TXA	7	[IMAGE png]	12
RXR	8	[IMAGE png]	13
RXA	9	[IMAGE png]	14
RST	10	[IMAGE png]	17
5V	11	[IMAGE png]	+5 V, 20mA Supply
GND	12	[IMAGE png]	any of 18-25
Interrupt	See note	[IMAGE png]	10

Power and Grounding

The RPC needs a +5V supply in order to operate. According to Radiometrix tech support, the radio can actually handle inputs ranging from +4.5V to +5.5V.

It would be extremely convenient if you could simply supply the needed 5V from one of the output pins of the PC parallel port. However, the original PC parallel port was only spec'd to supply up to 2.5mA, and the RPC draws up to 20mA@5V. Using the parallel port for RPC power may work with modern parallel port controllers (if they exceed the official specs of their predecessors by a factor of 10); I've just never tried it. In our lab we typically use a battery connected to a voltage regulator, or a bench power supply.

Naturally, the PC and RPC must share a common ground. Take care if the RPC's power supply does not share a ground with the PC's power supply--make sure there is not a significant potential between the two grounds before plugging them together.

Interrupts

Using interrupts will significantly improve the performance of the radio but is not strictly required. The two state pins that are relevant for interrupt generation are

Interrupts on the parallel port of the PC are signalled on the rising edge of pin 10. However, RXR and TXA are both normally high. Therefore, the simplest interrupt generator is simply an inverter between RXR and Pin 10. A better one is an XOR gate that combines RXR and TXA--then, when either one of these goes low, there is a low-to-high transition on the interrupt pin.

Note that if the RPC interface hardware generates interrupts, the PC parallel port itself must also be configured to generate interrupts; see Section 3.1 for details.

Important Note: If any transmitter in your testbed generates interrupts on TXA, make sure all receivers generate interrupts on RXR (or both RXR and TXA). This is required because a transmitter that uses TXA interrupts will transmit too quickly for a non-interrupt-enabled receiver.

Compiling and Installing the Driver

Unpacking the distribution and typing make should generate an object file called krpc.o. (Other files are also created while krpc.o is being built; these files can be ignored). A Linux kernel v2.2 or higher is required.

The parapin parallel port pin control library, which is required by the krpc module, is included in the RPC driver distribution. It can be used separately for other applications or drivers that use the PC parallel port as a generic digital I/O interface. For more details, see the Parapin home page.

After the module has been compiled, it can be installed much as any other Linux kernel module; e.g. insmod krpc.o. Note that it depends on the parport and parport_pc modules; these might have to be installed manually before you install krpc. For example:

Configuring the Parallel Port for Interrupts

If your RPC interface is capable of generating interrupts, special care must be taken to ensure that those interrupts are delivered all the way up to the RPC application. First, make sure the parallel port hardware is configured to generate interrupts. On modern motherboards with integrated parallel ports, this is selected from the BIOS setup screen. Older systems may use a DIP switch or jumper on the motherboard or the expansion board that drives the parallel port.

Next, the Linux kernel itself must be configured to handle parallel port interrupts. Unlike most other hardware devices, the kernel does not detect or claim the parallel port's interrupts by default. It is possible to manually enable kernel IRQ handling for the parallel port by writing the interrupt number into the special file /proc/parport/[IMAGE png]/irq, where [IMAGE png] is the parallel port number. For example, the following command tells the kernel that parport0 is using IRQ 7:

The actual interrupt number used by the kernel (7 in the examples above) must, of course, match the interrupt line being used by the hardware. The IRQ used by the parallel port hardware is usually configured in the BIOS setup screen on modern motherboards that have built-in parallel ports. Older motherboards or stand-alone ISA cards usually have jumpers or DIP switches for configuring the interrupt number. The typical assignment of interrupts to parallel ports is as follows:

These are reasonable defaults if the actual hardware configuration is not known.

When the krpc module is inserted, it will report its status in the kernel log that appears on the console and/or in the system log files. Make sure these messages indicate the module is using interrupts. If the driver reports that it is in ``polling mode'', the Linux kernel is not properly configured as described above. Note that the Linux kernel must be configured to use interrupts before the krpc module is inserted.

Use the irq_debug module option (desribed below) to make sure that interrupts are actually being delivered. At least one ``interrupt received'' message should appear in the system log each time a packet is received (if your RPC interface generates RXR interrupts) or transmitted (if your RPC interface generates TXA interrupts).

Module Options

The krpc module takes a number of options that can be set by specifying arguments with insmod; for example:

Using the Device Driver

Once the krpc module has been successfully compiled and installed, it actually creates two different ``personalities.'' The first is a character device (e.g., /dev/rpc), described in Section 5. The character device can be used to transmit arbitrary data from one station to another using a simple file interface; e.g. ``echo hello > /dev/rpc'' will transmit the string hello.

The second personality is a network interface (rpc0), described in Section 6. The network interface can be used to create a peer-to-peer IP network using the RPC radios. This interface can be configured in the normal way (i.e., using ifconfig).

Only one of these personalities can be used at a time. Multiple processes are allowed to access the device simultaneously; however, all accesses must be using the same personality at the same time. For example, if the network interface is configured and up, any attempt to use the character interface will return an error of EBUSY or EAGAIN.

The Character Device

When krpc is installed, it registers a major device number for the RPC device. By default, the major number is 240 (this number is reserved for local/experimental use in the kernel). To change the major number to something different, use the krpc_major_number module parameter described in Section 3.2. Type cat /proc/devices to see the bindings of major numbers to device drivers.

The driver also has a number of different ``sub-personalities'' that can be selected using the minor number of the device file. Specifically:

Minor Number	Mode Description
0	Raw, direct access to the RPC radio. Messages of more than 27 bytes are not allowed due to the RPC hardware's message-size limitation.
1	Cooked radio interface; implements fragmentation so that larger packets, up to 8K, may be transmitted and/or received.
2	Packets are delivered to applications with the `rpc_augmented` structure prepended (i.e., before the data). This structure contains meta-data such as the time of reception of the packet. See the header file `krpc.h` for the definition of this structure.
3	Both 1 and 2.

Remember, only one of these personalities may be used at a time. If one process is reading from /dev/rpc, another process that attempts to open /dev/rpcc will receive EAGAIN.

Creating Device Files

After you decide on a major number to use for krpc, device files using that major number must be created. For example, using major number 240, the following commands would create the appropriate device files:

mknod /dev/rpc c 240 0 minor number 0: this creates the standard interface, max of 27 bytes per write] mknod /dev/rpcc c 240 1 minor number 1: this creates the "cooked" interface, which does fragmentation, and allows 8K writes mknod /dev/rpca c 240 2 minor number 2: the "augmented" interface mknod /dev/rpcca c 240 3 minor number 3: both 1 and 2

Using the Character Device

The RPC character device driver implements blocking reads, so to wait for incoming packets on a receiver you can just type

Minor number 0 is a raw interface straight to the radio, so it will not accept writes longer than 27 bytes. This is due to the RPC packet controller's 27 byte message-size limitation.

To use the cooked interface that includes fragmentation, allowing you to send longer messages, use minor number 1:

cat /dev/rpcc (on one side) echo "this is a really long message that requires fragmentation" > /dev/rpcc (on the other side)

Make sure to set up the receiver before sending; when you start receiving, the queue is cleared of old packets. Just like a socket, you can't receive data that was sent before you were listening for it.

In many ways, the semantics of the driver are the same as any other character device. For example:

However, there are important differences between this driver and other character devices. In particular, be aware that the sequence of data returned when you issue a read() acts like a datagram interface. All reads will deliver data starting from the beginning of a packet. If a 20-byte packet comes in, and you only issue a read for 15 bytes, the remaining 5 bytes are lost forever. The next read does not start where the previous one left off; it gives you the beginning of the next complete packet. Therefore, it is important to issue reads with buffers large enough to consume entire packets.

The Network Device

When the krpc kernel module is loaded, it also creates a network device, rpc0. This lets you run IP over the RPC radio, similar to a wireless Ethernet card in ``peer-to-peer'' mode. All IP protocols (e.g. TCP, UDP, ICMP) work as usual--just a lot slower. In fact, the interface is so slow and unreliable that it is not especially useful for interactive applications such as telnet or X terminals. However, it may be useful for sending occasional debug or control information.

Known Bugs and Limitations

Currently, the RPC driver can only control one RPC device at a time. Configurations with multiple RPC radios are not supported. (The underlying Parapin library currently has the same limitation.)

When using the fragmentation interface, the memory used by krpc will grow slowly over time (up to a certain limit). The driver currently does not implement garbage collection; it will in the future.

When providing an IP interface, the RPC driver really should do link-layer retransmissions. TCP performance is rather bad otherwise.

The implementation of fragmentation can be made somewhat more efficient; the first ``introduction'' fragment currently does not carry any data.

About this document ...

This document was generated using the LaTeX2HTML translator Version 99.2beta6 (1.42)

The command line arguments were:
latex2html -split 0 -ascii_mode -noauto_link -rootdir /tmp/rpc_driverstemp rpc_drivers

Port	Interrupt
LPT1	7
LPT2	5