Difference between revisions of "Mote-mote radio communication"
m |
m |
||
(11 intermediate revisions by 7 users not shown) | |||
Line 7: | Line 7: | ||
=Introduction= | =Introduction= | ||
− | TinyOS provides a number of ''interfaces'' to abstract the underlying communications services and a number of ''components'' that ''provide'' (implement) these interfaces. All of these interfaces and components use a common message buffer abstraction, called <code>message_t</code>, which is implemented as a nesC struct (similar to a C struct). The <code>message_t</code> abstraction replaces the TinyOS 1.x <code>TOS_Msg</code> abstraction. Unlike TinyOS 1.x, the members of <code>message_t</code> are opaque, and therefore not accessed directly. Rather, <code>message_t</code> is an ''abstract data type'', whose members are read and written using accessor and mutator functions[http://www.tinyos.net/tinyos-2.x/doc/html/ | + | TinyOS provides a number of ''interfaces'' to abstract the underlying communications services and a number of ''components'' that ''provide'' (implement) these interfaces. All of these interfaces and components use a common message buffer abstraction, called <code>message_t</code>, which is implemented as a nesC struct (similar to a C struct). The <code>message_t</code> abstraction replaces the TinyOS 1.x <code>TOS_Msg</code> abstraction. Unlike TinyOS 1.x, the members of <code>message_t</code> are opaque, and therefore not accessed directly. Rather, <code>message_t</code> is an ''abstract data type'', whose members are read and written using accessor and mutator functions <ref name="fn1">[http://www.tinyos.net/tinyos-2.x/doc/html/tep111.html TEP 111: message_t]</ref>. |
==Basic Communications Interfaces== | ==Basic Communications Interfaces== | ||
Line 21: | Line 21: | ||
==Active Message Interfaces== | ==Active Message Interfaces== | ||
− | Since it is very common to have multiple services using the same radio to communicate, TinyOS provides the Active Message (AM) layer to multiplex access to the radio. The term "AM type" refers to the field used for multiplexing. AM types are similar in function to the Ethernet frame type field, IP protocol field, and the UDP port in that all of them are used to multiplex access to a communication service. AM | + | Since it is very common to have multiple services using the same radio to communicate, TinyOS provides the Active Message (AM) layer to multiplex access to the radio. The term "AM type" refers to the field used for multiplexing. AM types are similar in function to the Ethernet frame type field, IP protocol field, and the UDP port in that all of them are used to multiplex access to a communication service. An AM packet also includes a destination field, which stores an "AM address" to address packets to particular motes. Additional interfaces, also located in the <code>tos/interfaces</code> directory, were introduced to support the AM services: |
− | * <code>[http://www.tinyos.net/tinyos-2.x/tos/interfaces/AMPacket.nc AMPacket]</code> - Similar to <code>Packet</code>, provides the basic AM accessors for the <code>message_t</code> abstract data type. This interface provides commands for getting a node's AM address, an AM packet's destination, and an AM packet's type. Commands are also | + | * <code>[http://www.tinyos.net/tinyos-2.x/tos/interfaces/AMPacket.nc AMPacket]</code> - Similar to <code>Packet</code>, provides the basic AM accessors for the <code>message_t</code> abstract data type. This interface provides commands for getting a node's AM address, an AM packet's destination, and an AM packet's type. Commands are also provided for setting an AM packet's destination and type, and checking whether the destination is the local node. |
* <code>[http://www.tinyos.net/tinyos-2.x/tos/interfaces/AMSend.nc AMSend]</code> - Similar to <code>Send</code>, provides the basic Active Message sending interface. The key difference between <code>AMSend</code> and <code>Send</code> is that <code>AMSend</code> takes a destination AM address in its <code>send</code> command. | * <code>[http://www.tinyos.net/tinyos-2.x/tos/interfaces/AMSend.nc AMSend]</code> - Similar to <code>Send</code>, provides the basic Active Message sending interface. The key difference between <code>AMSend</code> and <code>Send</code> is that <code>AMSend</code> takes a destination AM address in its <code>send</code> command. | ||
Line 36: | Line 36: | ||
* <code>[http://www.tinyos.net/tinyos-2.x/tos/system/AMSnooperC.nc AMSnooperC]</code> - Provides <code>Receive</code>, <code>Packet</code>, and <code>AMPacket</code>. | * <code>[http://www.tinyos.net/tinyos-2.x/tos/system/AMSnooperC.nc AMSnooperC]</code> - Provides <code>Receive</code>, <code>Packet</code>, and <code>AMPacket</code>. | ||
* <code>[http://www.tinyos.net/tinyos-2.x/tos/system/AMSnoopingReceiverC.nc AMSnoopingReceiverC]</code> - Provides <code>Receive</code>, <code>Packet</code>, and <code>AMPacket</code>. | * <code>[http://www.tinyos.net/tinyos-2.x/tos/system/AMSnoopingReceiverC.nc AMSnoopingReceiverC]</code> - Provides <code>Receive</code>, <code>Packet</code>, and <code>AMPacket</code>. | ||
− | * <code>[http://www.tinyos.net/tinyos-2.x/tos/system/ActiveMessageAddressC.nc ActiveMessageAddressC]</code> - Provides commands to get and set the node's active message address. This interface is not for general use and changing | + | * <code>[http://www.tinyos.net/tinyos-2.x/tos/system/ActiveMessageAddressC.nc ActiveMessageAddressC]</code> - Provides commands to get and set the node's active message address. This interface is not for general use and changing a node's active message address can break the network stack, so avoid using it unless you know what you are doing. |
==Naming Wrappers== | ==Naming Wrappers== | ||
Line 52: | Line 52: | ||
nx_uint8_t header[sizeof(message_header_t)]; | nx_uint8_t header[sizeof(message_header_t)]; | ||
nx_uint8_t data[TOSH_DATA_LENGTH]; | nx_uint8_t data[TOSH_DATA_LENGTH]; | ||
− | nx_uint8_t footer[sizeof( | + | nx_uint8_t footer[sizeof(message_footer_t)]; |
nx_uint8_t metadata[sizeof(message_metadata_t)]; | nx_uint8_t metadata[sizeof(message_metadata_t)]; | ||
} message_t; | } message_t; | ||
− | '''Note: The header, footer, and metadata fields are all opaque and must not be accessed directly. It is important to access the <code>message_t</code> fields only through <code>Packet</code>, <code>AMPacket</code>, and other such interfaces, as will be demonstrated in this tutorial.''' The rationale for this approach is that it allows the data (payload) to be kept at a fixed offset, avoiding a copy when a message is passed between two link layers. See Section 3 in | + | '''Note: The header, footer, and metadata fields are all opaque and must not be accessed directly. It is important to access the <code>message_t</code> fields only through <code>Packet</code>, <code>AMPacket</code>, and other such interfaces, as will be demonstrated in this tutorial.''' The rationale for this approach is that it allows the data (payload) to be kept at a fixed offset, avoiding a copy when a message is passed between two link layers. See Section 3 in TEP 111<ref name="fn1"/> for more details. |
=Sending a Message over the Radio= | =Sending a Message over the Radio= | ||
− | We will now create a simple application that increments a counter, displays the counter's three least significant bits on the three LEDs, and sends a message with the counter value over the radio. Our implementation will use a single timer and a counter, in a way similar to the <code>BlinkSingle</code> example from [ | + | We will now create a simple application that increments a counter, displays the counter's three least significant bits on the three LEDs, and sends a message with the counter value over the radio. Our implementation will use a single timer and a counter, in a way similar to the <code>BlinkSingle</code> example from [[Modules and the TinyOS Execution Model| lesson 2]]. |
==Reimplementing Blink== | ==Reimplementing Blink== | ||
Line 141: | Line 141: | ||
nx_uint16_t counter; | nx_uint16_t counter; | ||
} BlinkToRadioMsg; | } BlinkToRadioMsg; | ||
− | If this code doesn't look even vaguely familiar, you should spend a few minutes reading up on C structures. If you are familiar with C structures, this syntax should look familar but the <code>nx_</code> prefix on the keywords <code>struct</code> and <code>uint16_t</code> should stand out. The <code>nx_</code> prefix is specific to the nesC language and signifies that the <code>struct</code> and <code>uint16_t</code> are ''external types''[http:// | + | If this code doesn't look even vaguely familiar, you should spend a few minutes reading up on C structures. If you are familiar with C structures, this syntax should look familar but the <code>nx_</code> prefix on the keywords <code>struct</code> and <code>uint16_t</code> should stand out. The <code>nx_</code> prefix is specific to the nesC language and signifies that the <code>struct</code> and <code>uint16_t</code> are ''external types'' <ref name="fn3">'''Programming Hint 15:'''Always use platform independent types when defining message formats. From Phil Levis' [http://csl.stanford.edu/~pal/pubs/tinyos-programming-1-0.pdf ''TinyOS Programming'']</ref><ref name="fn4">'''Programming Hint 16:'''If you have to perform significant computation on a platform independent type or access it many (hundreds or more) times, then temporarily copying it to a native type can be a good idea. From Phil Levis' [http://csl.stanford.edu/~pal/pubs/tinyos-programming-1-0.pdf ''TinyOS Programming'']</ref>. External types have the same representation on all platforms. The nesC compiler generates code that transparently reorders access to <code>nx_</code> data types and eliminates the need to manually address endianness and alignment (extra padding in structs present on some platforms) issues. So what is endianness? Read on... |
− | Different processors represent numbers in different ways in their memory: some processors use a "big endian" representation which means that the most significant byte of a multi-byte (e.g. 16- or 32-bit) number is located at a lower memory address than the least significant byte, while "little endian" stores data in exactly the opposite order. A problem arises when data is serialized and sent over the network because different processors will decode the same set of bytes in different ways, depending on their "endianness." The main difficulty endianness presents is that it requires operations to rearrange byte orders to match the network protocol specification or the processor architecture -- an annoying and error-prone process. The <code>htons</code>, <code>htonl</code>, <code>ntohs</code>, and <code>ntohl</code> calls used with the sockets API are an example of platform-specific calls that convert between network and host byte orders, but you have to remember to use them. The nesC programming language takes a different approach to the problem and defines ''external types'' which allow the programmer to avoid dealing with byte reordering. In particular, the <code>nx_</code> prefix on a type (e.g. <code>nx_uint16_t</code>) indicates the field is to serialized in big endian format. In contrast, the <code>nxle_</code> prefix signifies that the field is serialized in little endian format. | + | Different processors represent numbers in different ways in their memory: some processors use a "big endian" representation which means that the most significant byte of a multi-byte (e.g. 16- or 32-bit) number is located at a lower memory address than the least significant byte, while "little endian" stores data in exactly the opposite order. A problem arises when data is serialized and sent over the network because different processors will decode the same set of bytes in different ways, depending on their "endianness." The main difficulty endianness presents is that it requires operations to rearrange byte orders to match the network protocol specification or the processor architecture -- an annoying and error-prone process. The <code>htons</code>, <code>htonl</code>, <code>ntohs</code>, and <code>ntohl</code> calls used with the sockets API are an example of platform-specific calls that convert between network and host byte orders, but you have to remember to use them. The nesC programming language takes a different approach to the problem and defines ''external types'' which allow the programmer to avoid dealing with byte reordering. In particular, the <code>nx_</code> prefix on a type (e.g. <code>nx_uint16_t</code>) indicates the field is to be serialized in big endian format. In contrast, the <code>nxle_</code> prefix signifies that the field is serialized in little endian format. |
==Sending a Message== | ==Sending a Message== | ||
Line 149: | Line 149: | ||
Now that we have defined a message type for our application, <code>BlinkToRadioMsg</code>, we will next see how to send the message over the radio. Before beginning, let's review the purpose of the application. We want a timer-driven system in which every firing of the timer results in (i) incrementing a counter, (ii) displaying the three lowest bits of the counter on the LEDs, and (iii) transmitting the node's id and counter value over the radio. To implement this program, we follow a number of simple steps, as described in the next paragraph. | Now that we have defined a message type for our application, <code>BlinkToRadioMsg</code>, we will next see how to send the message over the radio. Before beginning, let's review the purpose of the application. We want a timer-driven system in which every firing of the timer results in (i) incrementing a counter, (ii) displaying the three lowest bits of the counter on the LEDs, and (iii) transmitting the node's id and counter value over the radio. To implement this program, we follow a number of simple steps, as described in the next paragraph. | ||
− | First, we need to identify the interfaces (and components) that provide access to the radio and allow us to manipulate the <code>message_t</code> type. Second, we must update the <code>module</code> block in the <code>BlinkToRadioC.nc</code> by adding <code>uses</code> statements for the interfaces we need. Third, we need to declare new variables and add any initialization and start/stop code that is needed by the interfaces and components. Fourth, we must add any calls to the component interfaces we need for our application. Fifth, we need to | + | First, we need to identify the interfaces (and components) that provide access to the radio and allow us to manipulate the <code>message_t</code> type. Second, we must update the <code>module</code> block in the <code>BlinkToRadioC.nc</code> by adding <code>uses</code> statements for the interfaces we need. Third, we need to declare new variables and add any initialization and start/stop code that is needed by the interfaces and components. Fourth, we must add any calls to the component interfaces we need for our application. Fifth, we need to implement any events specified in the interfaces we plan on using. Sixth, the <code>implementation</code> block of the application configuration file, <code>BlinkToRadioApp.c</code>, must be updated by adding a <code>components</code> statement for each component we use that provides one of the interfaces we chose earlier. Finally, we need to wire the interfaces used by the application to the components which provide those interfaces. |
Let's walk through the steps, one-by-one: | Let's walk through the steps, one-by-one: | ||
Line 193: | Line 193: | ||
... | ... | ||
if (!busy) { | if (!busy) { | ||
− | BlinkToRadioMsg* btrpkt = (BlinkToRadioMsg*)(call Packet.getPayload(&pkt, | + | BlinkToRadioMsg* btrpkt = (BlinkToRadioMsg*)(call Packet.getPayload(&pkt, sizeof (BlinkToRadioMsg))); |
btrpkt->nodeid = TOS_NODE_ID; | btrpkt->nodeid = TOS_NODE_ID; | ||
btrpkt->counter = counter; | btrpkt->counter = counter; | ||
Line 202: | Line 202: | ||
} | } | ||
This code performs several operations. First, it ensures that a message transmission is not in progress by checking the busy flag. Then it gets the packet's payload portion and casts it to a pointer to the previously declared <code>BlinkToRadioMsg</code> external type. It can now use this pointer to initialise the packet's fields, and then send the packet by calling <code>AMSend.send</code>. The packet is sent to all nodes in radio range by specyfing <code>AM_BROADCAST_ADDR</code> as the destination address. Finally, the test against SUCCESS verifies that the AM layer accepted the message for transmission. If so, the busy flag is set to true. For the duration of the send attempt, the packet is owned by the radio, and user code must not access it. Note that we could have avoided using the <code>Packet</code> interface, as it's <code>getPayload</code> command is repeated within <code>AMSend</code>. | This code performs several operations. First, it ensures that a message transmission is not in progress by checking the busy flag. Then it gets the packet's payload portion and casts it to a pointer to the previously declared <code>BlinkToRadioMsg</code> external type. It can now use this pointer to initialise the packet's fields, and then send the packet by calling <code>AMSend.send</code>. The packet is sent to all nodes in radio range by specyfing <code>AM_BROADCAST_ADDR</code> as the destination address. Finally, the test against SUCCESS verifies that the AM layer accepted the message for transmission. If so, the busy flag is set to true. For the duration of the send attempt, the packet is owned by the radio, and user code must not access it. Note that we could have avoided using the <code>Packet</code> interface, as it's <code>getPayload</code> command is repeated within <code>AMSend</code>. | ||
− | <li>''' | + | <li>'''Implement any (non-initialization) events specified in the interfaces we plan on using.''' |
<BR>Looking through the <code>Packet</code>, <code>AMPacket</code>, and <code>AMSend</code> interfaces, we see that there is only one <code>event</code> we need to worry about, <code>AMSend.sendDone</code>:</li> | <BR>Looking through the <code>Packet</code>, <code>AMPacket</code>, and <code>AMSend</code> interfaces, we see that there is only one <code>event</code> we need to worry about, <code>AMSend.sendDone</code>:</li> | ||
/** | /** | ||
Line 266: | Line 266: | ||
<BR>We will not require any new variables to receive and process messages from the radio. | <BR>We will not require any new variables to receive and process messages from the radio. | ||
<li>'''Add any program logic and calls to the used interfaces we need for our application.''' | <li>'''Add any program logic and calls to the used interfaces we need for our application.''' | ||
− | <BR>Message reception is an event-driven process so we do not need to call | + | <BR>Message reception is an event-driven process so we do not need to call any commands on the <code>Receive</code>. |
<li>'''Implemement any (non-initialization) events specified in the interfaces we plan on using.''' | <li>'''Implemement any (non-initialization) events specified in the interfaces we plan on using.''' | ||
<BR>We need to implement the <code>Receive.receive</code> event handler:</li> | <BR>We need to implement the <code>Receive.receive</code> event handler:</li> | ||
Line 278: | Line 278: | ||
The <code>receive</code> event handler performs some simple operations. First, we need to ensure that the length of the message is what is expected. Then, the message payload is cast to a structure pointer of type <code>BlinkToRadioMsg*</code> and assigned to a local variable. Then, the counter value in the message is used to set the states of the three LEDs.Note that we can safely manipulate the <code>counter</code> variable outside of an atomic section. The reason is that receive event executes in task context rather than interrupt context (events that have the <code>async</code> keyword can execute in interrupt context). Since the TinyOS execution model allows only one task to execute at a time, if all accesses to a variable occur in task context, then no race conditions will occur for that variable. Since all accesses to <code>counter</code> occur in task context, no critical sections are needed when accessing it. | The <code>receive</code> event handler performs some simple operations. First, we need to ensure that the length of the message is what is expected. Then, the message payload is cast to a structure pointer of type <code>BlinkToRadioMsg*</code> and assigned to a local variable. Then, the counter value in the message is used to set the states of the three LEDs.Note that we can safely manipulate the <code>counter</code> variable outside of an atomic section. The reason is that receive event executes in task context rather than interrupt context (events that have the <code>async</code> keyword can execute in interrupt context). Since the TinyOS execution model allows only one task to execute at a time, if all accesses to a variable occur in task context, then no race conditions will occur for that variable. Since all accesses to <code>counter</code> occur in task context, no critical sections are needed when accessing it. | ||
<li>'''Update the implementation block of the application configuration file by adding a components statement for each component used that provides one of the interfaces chosen earlier.''' | <li>'''Update the implementation block of the application configuration file by adding a components statement for each component used that provides one of the interfaces chosen earlier.''' | ||
− | <BR>The following lines can be added just below the existing <code>components</code> declarations in the implementation block of <code> | + | <BR>The following lines can be added just below the existing <code>components</code> declarations in the implementation block of <code>BlinkToRadioAppC.nc</code>:</li> |
implementation { | implementation { | ||
... | ... | ||
Line 348: | Line 348: | ||
rm -f build/telosb/main.exe.out-2 build/telosb/main.ihex.out-2 | rm -f build/telosb/main.exe.out-2 build/telosb/main.ihex.out-2 | ||
'''<font color="red">At this point, both motes should be blinking their LEDs.</font>''' | '''<font color="red">At this point, both motes should be blinking their LEDs.</font>''' | ||
− | If you press the RESET button on either telosb, then the LEDs on the ''other'' telosb will pause on whatever was being displayed at the moment you pressed RESET. When you release the RESET | + | If you press the RESET button on either telosb, then the LEDs on the ''other'' telosb will pause on whatever was being displayed at the moment you pressed RESET. When you release the RESET button, the paused mote will be reset and then resume counting from one. |
</ol> | </ol> | ||
=Conclusions= | =Conclusions= | ||
− | This lesson has introduced radio communications TinyOS 2.x. | + | This lesson has introduced radio communications in TinyOS 2.x. |
=Related Documentation= | =Related Documentation= | ||
− | + | <references/> | |
− | + | ---- | |
− | |||
− | |||
− | |||
<center> | <center> | ||
− | |||
< '''[[Modules and the TinyOS Execution Model|Previous Lesson]]''' | '''[[Mote-mote radio communication#Introduction|Top]]''' | '''[[Mote-PC serial communication and SerialForwarder|Next Lesson]] >''' | < '''[[Modules and the TinyOS Execution Model|Previous Lesson]]''' | '''[[Mote-mote radio communication#Introduction|Top]]''' | '''[[Mote-PC serial communication and SerialForwarder|Next Lesson]] >''' | ||
</center> | </center> |
Latest revision as of 16:41, 14 July 2011
This lesson introduces radio communications in TinyOS. You will become familar with TinyOS interfaces and components that support communications and you will learn how to:
- Use
message_t
, the TinyOS 2.0 message buffer. - Send a message buffer to the radio.
- Receive a message buffer from the radio.
Contents
Introduction
TinyOS provides a number of interfaces to abstract the underlying communications services and a number of components that provide (implement) these interfaces. All of these interfaces and components use a common message buffer abstraction, called message_t
, which is implemented as a nesC struct (similar to a C struct). The message_t
abstraction replaces the TinyOS 1.x TOS_Msg
abstraction. Unlike TinyOS 1.x, the members of message_t
are opaque, and therefore not accessed directly. Rather, message_t
is an abstract data type, whose members are read and written using accessor and mutator functions <ref name="fn1">TEP 111: message_t</ref>.
Basic Communications Interfaces
There are a number of interfaces and components that use message_t
as the underlying data structure. Let's take a look at some of the interfaces that are in the tos/interfaces
directory to familiarize ourselves with the general functionality of the communications system:
Packet
- Provides the basic accessors for themessage_t
abstract data type. This interface provides commands for clearing a message's contents, getting its payload length, and getting a pointer to its payload area.Send
- Provides the basic address-free message sending interface. This interface provides commands for sending a message and canceling a pending message send. The interface provides an event to indicate whether a message was sent successfully or not. It also provides convenience functions for getting the message's maximum payload as well as a pointer to a message's payload area.Receive
- Provides the basic message reception interface. This interface provides an event for receiving messages. It also provides, for convenience, commands for getting a message's payload length and getting a pointer to a message's payload area.PacketAcknowledgements
- Provides a mechanism for requesting acknowledgements on a per-packet basis.RadioTimeStamping
- Provides time stamping information for radio transmission and reception.
Active Message Interfaces
Since it is very common to have multiple services using the same radio to communicate, TinyOS provides the Active Message (AM) layer to multiplex access to the radio. The term "AM type" refers to the field used for multiplexing. AM types are similar in function to the Ethernet frame type field, IP protocol field, and the UDP port in that all of them are used to multiplex access to a communication service. An AM packet also includes a destination field, which stores an "AM address" to address packets to particular motes. Additional interfaces, also located in the tos/interfaces
directory, were introduced to support the AM services:
AMPacket
- Similar toPacket
, provides the basic AM accessors for themessage_t
abstract data type. This interface provides commands for getting a node's AM address, an AM packet's destination, and an AM packet's type. Commands are also provided for setting an AM packet's destination and type, and checking whether the destination is the local node.AMSend
- Similar toSend
, provides the basic Active Message sending interface. The key difference betweenAMSend
andSend
is thatAMSend
takes a destination AM address in itssend
command.
The AM address of a node can be set at installation time, using the make install.n
or make reinstall.n
commands. It can be changed at runtime using the ActiveMessageAddressC
component (see below).
Components
A number of components implement the basic communications and active message interfaces. Let's take a look at some of the components in the /tos/system
directory. You should be familiar with these components because your code needs to specify both the interfaces your application uses as well as the components which provide (implement) those interfaces:
AMReceiverC
- Provides the following interfaces:Receive
,Packet
, andAMPacket
.AMSenderC
- ProvidesAMSend
,Packet
,AMPacket
, andPacketAcknowledgements
asAcks
.AMSnooperC
- ProvidesReceive
,Packet
, andAMPacket
.AMSnoopingReceiverC
- ProvidesReceive
,Packet
, andAMPacket
.ActiveMessageAddressC
- Provides commands to get and set the node's active message address. This interface is not for general use and changing a node's active message address can break the network stack, so avoid using it unless you know what you are doing.
Naming Wrappers
Since TinyOS supports multiple platforms, each of which might have their own implementation of the radio drivers, an additional, platform-specific, naming wrapper called ActiveMessageC
is used to bridge these interfaces to their underlying, platform-specific implementations. ActiveMessageC
provides most of the communication interfaces presented above. Platform-specific versions of ActiveMessageC
, as well the underlying implementations which may be shared by multiple platforms (e.g. Telos and MicaZ) include:
ActiveMessageC
for the eyesIFX platform is implemented byTda5250ActiveMessageC
.ActiveMessageC
for the intelmote2, micaz, telosa, and telosb are all implemented byCC2420ActiveMessageC
.ActiveMessageC
for the mica2 platform is implemented byCC1000ActiveMessageC
.
The TinyOS 2.0 Message Buffer
TinyOS 2.0 introduces a new message buffer abstraction called message_t
. If you are familiar with earlier versions of TinyOS, you need to know that message_t
replaces TOS_Msg
. The message_t
structure is defined in tos/types/message.h
.
typedef nx_struct message_t { nx_uint8_t header[sizeof(message_header_t)]; nx_uint8_t data[TOSH_DATA_LENGTH]; nx_uint8_t footer[sizeof(message_footer_t)]; nx_uint8_t metadata[sizeof(message_metadata_t)]; } message_t;
Note: The header, footer, and metadata fields are all opaque and must not be accessed directly. It is important to access the message_t
fields only through Packet
, AMPacket
, and other such interfaces, as will be demonstrated in this tutorial. The rationale for this approach is that it allows the data (payload) to be kept at a fixed offset, avoiding a copy when a message is passed between two link layers. See Section 3 in TEP 111<ref name="fn1"/> for more details.
Sending a Message over the Radio
We will now create a simple application that increments a counter, displays the counter's three least significant bits on the three LEDs, and sends a message with the counter value over the radio. Our implementation will use a single timer and a counter, in a way similar to the BlinkSingle
example from lesson 2.
Reimplementing Blink
As a first step, we can reimplement Blink
using a single timer and counter. Create a new directory in apps named BlinkToRadio:
$ cd tinyos-2.x/apps $ mkdir BlinkToRadio
Inside this directory, create a file BlinkToRadioC.nc
, which has this code:
#include <Timer.h> #include "BlinkToRadio.h" module BlinkToRadioC { uses interface Boot; uses interface Leds; uses interface Timer<TMilli> as Timer0; } implementation { uint16_t counter = 0; event void Boot.booted() { call Timer0.startPeriodic(TIMER_PERIOD_MILLI); } event void Timer0.fired() { counter++; call Leds.set(counter); } }
Let's look at a few specific lines in this program. First, notice the C preprocessor include
directive on the first line. This directive tells the compiler to simply replace the directive with the entire contents of Timer.h
. The compiler looks for Timer.h
in the standard places. In this case, standard means the TinyOS system directories that are located in tos
or its subdirectories. It is possible to tell the compiler to look beyond these standard directories by using the -I
flag in the Makefile, for example, as is common when including contributed libraries located in contrib
directory tree.
The second line of this program is also an include
directive, but note that it uses quotes around the filename rather than angle brackets. The quotes tell the preprocessor to look in the current directory before searching through the standard directories for the particular file. In this case, the BlinkToRadio.h
file is located in the same directory and defines some constants that are used in this program. We will take a look at BlinkToRadio.h
in just a bit.
Next, the call to Leds.set
directly sets the three LEDs to the three low-order bits of the counter.
Finally, note the call Timer0.startPeriodic(TIMER_PERIOD_MILLI)
line in the Boot.booted
function. The value of TIMER_PERIOD_MILLI
is defined in the BlinkToRadio.h
header file:
#ifndef BLINKTORADIO_H #define BLINKTORADIO_H enum { TIMER_PERIOD_MILLI = 250 }; #endif
BlinkToRadio.h
is a pretty standard header file but there are two things to note here. First, notice the use of the ifndef
, define
, and endif
directives. These directives are used to ensure that the definitions in each header file is not included multiple times because the compiler would complain about multiply-defined objects. By convention, the literal used for these directives is an all-caps version of the filename with any periods converted to underscores. The other important thing to note is the use of an enum
declaration for defining the constant TIMER_PERIOD_MILLI
. Using enum
for defining constants is preferred over using define
because enum
does not indiscriminantly replace every occurence of the define
d literal, regardless of where it appears in the source. As a result, enum
s provide better scoping as well.
BlinkToRadioC.nc
provides the implementation logic of the program and BlinkToRadio.h
defines constants and/or data structures. A third file is needed to wire the interfaces that the implementation uses
to the actual components which provide
these interfaces. The BlinkToRadioAppC.nc
provides the needed wiring:
#include <Timer.h> #include "BlinkToRadio.h" configuration BlinkToRadioAppC { } implementation { components MainC; components LedsC; components BlinkToRadioC as App; components new TimerMilliC() as Timer0; App.Boot -> MainC; App.Leds -> LedsC; App.Timer0 -> Timer0; }
The BlinkToRadioAppC
should look familiar to you since it is essentially a subset of the Blink
application/configuration from an earlier lesson.
These three files constitute all of the application code: the only other thing it needs is a Makefile. Create a file named Makefile
. For an application as simple as this one, the Makefile is very short:
COMPONENT=BlinkToRadioAppC include $(MAKERULES)
The first line tells the TinyOS make system that the top-level application component is BlinkToRadioAppC. The second line loads in the TinyOS build system, which has all of the rules for building and installing on different platforms.
Defining a Message Structure
Now that Blink
has been reimplemented using a single timer and counter, we can now turn our attention to defining a message format to send data over the radio. Our message will send both the node id and the counter value over the radio. Rather than directly writing and reading the payload area of the message_t
with this data, we will use a structure to hold them and then use structure assignment to copy the data into the message payload area. Using a structure allows reading and writing the message payload more conveniently when your message has multiple fields or multi-byte fields (like uint16_t or uint32_t) because you can avoid reading and writing bytes from/to the payload using indices and then shifting and adding (e.g. uint16_t x = data[0] << 8 + data[1]
). Even for a message with a single field, you should get used to using a structure because if you ever add more fields to your message or move any of the fields around, you will need to manually update all of the payload position indices if you read and write the payload at a byte level. Using structures is straightforward. To define a message structure with a uint16_t
node id and a uint16_t
counter in the payload, we add the following lines to BlinkToRadio.h
, just before the endif
directive:
typedef nx_struct BlinkToRadioMsg { nx_uint16_t nodeid; nx_uint16_t counter; } BlinkToRadioMsg;
If this code doesn't look even vaguely familiar, you should spend a few minutes reading up on C structures. If you are familiar with C structures, this syntax should look familar but the nx_
prefix on the keywords struct
and uint16_t
should stand out. The nx_
prefix is specific to the nesC language and signifies that the struct
and uint16_t
are external types <ref name="fn3">Programming Hint 15:Always use platform independent types when defining message formats. From Phil Levis' TinyOS Programming</ref><ref name="fn4">Programming Hint 16:If you have to perform significant computation on a platform independent type or access it many (hundreds or more) times, then temporarily copying it to a native type can be a good idea. From Phil Levis' TinyOS Programming</ref>. External types have the same representation on all platforms. The nesC compiler generates code that transparently reorders access to nx_
data types and eliminates the need to manually address endianness and alignment (extra padding in structs present on some platforms) issues. So what is endianness? Read on...
Different processors represent numbers in different ways in their memory: some processors use a "big endian" representation which means that the most significant byte of a multi-byte (e.g. 16- or 32-bit) number is located at a lower memory address than the least significant byte, while "little endian" stores data in exactly the opposite order. A problem arises when data is serialized and sent over the network because different processors will decode the same set of bytes in different ways, depending on their "endianness." The main difficulty endianness presents is that it requires operations to rearrange byte orders to match the network protocol specification or the processor architecture -- an annoying and error-prone process. The htons
, htonl
, ntohs
, and ntohl
calls used with the sockets API are an example of platform-specific calls that convert between network and host byte orders, but you have to remember to use them. The nesC programming language takes a different approach to the problem and defines external types which allow the programmer to avoid dealing with byte reordering. In particular, the nx_
prefix on a type (e.g. nx_uint16_t
) indicates the field is to be serialized in big endian format. In contrast, the nxle_
prefix signifies that the field is serialized in little endian format.
Sending a Message
Now that we have defined a message type for our application, BlinkToRadioMsg
, we will next see how to send the message over the radio. Before beginning, let's review the purpose of the application. We want a timer-driven system in which every firing of the timer results in (i) incrementing a counter, (ii) displaying the three lowest bits of the counter on the LEDs, and (iii) transmitting the node's id and counter value over the radio. To implement this program, we follow a number of simple steps, as described in the next paragraph.
First, we need to identify the interfaces (and components) that provide access to the radio and allow us to manipulate the message_t
type. Second, we must update the module
block in the BlinkToRadioC.nc
by adding uses
statements for the interfaces we need. Third, we need to declare new variables and add any initialization and start/stop code that is needed by the interfaces and components. Fourth, we must add any calls to the component interfaces we need for our application. Fifth, we need to implement any events specified in the interfaces we plan on using. Sixth, the implementation
block of the application configuration file, BlinkToRadioApp.c
, must be updated by adding a components
statement for each component we use that provides one of the interfaces we chose earlier. Finally, we need to wire the interfaces used by the application to the components which provide those interfaces.
Let's walk through the steps, one-by-one:
- Identify the interfaces (and components) that provide access to the radio and allow us to manipulate the
message_t
type.
We will use theAMSend
interface to send packets as well as thePacket
andAMPacket
interfaces to access themessage_t
abstract data type. Although it is possible to wire directly to theActiveMessageC
component, we will instead use theAMSenderC
component. However, we still need to start the radio using theActiveMessageC.SplitControl
interface.The reason for usingAMSenderC
is because it provides a virtualized abstraction. Earlier versions of TinyOS did not virtualize access to the radio, so it was possible for two components that were sharing the radio to interfere with each other. It was not at all uncommon for one component to discover the radio was busy because some other component, unknown to the first component, was accessing the active message layer. Radio virtualization was introduced in TinyOS 2.0 to address this interference andAMSenderC
was written to provide this virtualization. Every user ofAMSenderC
is provided with a 1-deep queue and the queues of all users are serviced in a fair manner. - Update the
module
block in theBlinkToRadioC.nc
by addinguses
statements for the interfaces we need: - Declare any new variables and add any needed initialization code.
First, we need to declare some new module-scope variables. We need amessage_t
to hold our data for transmission. We also need a flag to keep track of when the radio is busy sending. These declarations need to be added in theimplementation
block ofBlinkToRadioC.nc
: - Add any program logic and calls to the used interfaces we need for our application.
Since we want to transmit the node's id and counter value every time the timer fires, we need to add some code to theTimer0.fired
event handler: - Implement any (non-initialization) events specified in the interfaces we plan on using.
Looking through thePacket
,AMPacket
, andAMSend
interfaces, we see that there is only oneevent
we need to worry about,AMSend.sendDone
: - Update the
implementation
block of the application configuration file by adding acomponents
statement for each component used that provides one of the interfaces chosen earlier.
The following lines can be added just below the existingcomponents
declarations in theimplementation
block ofBlinkToRadioAppC.nc
: - Wire the the interfaces used by the application to the components which provide those interfaces.
The following lines will wire the used interfaces to the providing components. These lines should be added to the bottom of theimplementation
block ofBlinkToRadioAppC.nc
:
module BlinkToRadioC { ... uses interface Packet; uses interface AMPacket; uses interface AMSend; uses interface SplitControl as AMControl; }
Note that SplitControl
has been renamed to AMControl
using the as
keyword. nesC allows interfaces to be renamed in this way for several reasons. First, it often happens that two or more components that are needed in the same module provide the same interface. The as
keyword allows one or more such names to be changed to distinct names so that they can each be addressed individually. Second, interfaces are sometimes renamed to something more meaningful. In our case, SplitControl
is a general interface used for starting and stopping components, but the name AMControl
is a mnemonic to remind us that the particular instance of SplitControl
is used to control the ActiveMessageC
component.
implementation { bool busy = FALSE; message_t pkt; ... }
Next, we need to handle the initialization of the radio. The radio needs to be started when the system is booted so we must call AMControl.start
inside Boot.booted
. The only complication is that in our current implementation, we start a timer inside Boot.booted
and we are planning to use this timer to send messages over the radio but the radio can't be used until it has completed starting up. The radio signals that it has completed starting through the AMControl.startDone
event. To ensure that we do not start using the radio before it is ready, we need to postpone starting the timer until after the radio has completed starting. We can accomplish this by moving the call to start the timer, which is now inside Boot.booted
, to AMControl.startDone
, giving us a new Boot.booted
with the following body:
event void Boot.booted() { call AMControl.start(); }
We also need to implement the AMControl.startDone
and AMControl.stopDone
event handlers, which have the following bodies:
event void AMControl.startDone(error_t err) { if (err == SUCCESS) { call Timer0.startPeriodic(TIMER_PERIOD_MILLI); } else { call AMControl.start(); } } event void AMControl.stopDone(error_t err) { }
If the radio is started successfully, AMControl.startDone
will be called with the error_t
parameter set to a value of SUCCESS
. If the radio starts successfully, then it is appropriate to start the timer. If, however, the radio does not start successfully, then it obviously cannot be used so we try again to start it. This process continues until the radio starts, and ensures that the node software doesn't run until the key components have started successfully. If the radio doesn't start at all, a human operator might notice that the LEDs are not blinking as they are supposed to, and might try to debug the problem.
event void Timer0.fired() { ... if (!busy) { BlinkToRadioMsg* btrpkt = (BlinkToRadioMsg*)(call Packet.getPayload(&pkt, sizeof (BlinkToRadioMsg))); btrpkt->nodeid = TOS_NODE_ID; btrpkt->counter = counter; if (call AMSend.send(AM_BROADCAST_ADDR, &pkt, sizeof(BlinkToRadioMsg)) == SUCCESS) { busy = TRUE; } } }
This code performs several operations. First, it ensures that a message transmission is not in progress by checking the busy flag. Then it gets the packet's payload portion and casts it to a pointer to the previously declared BlinkToRadioMsg
external type. It can now use this pointer to initialise the packet's fields, and then send the packet by calling AMSend.send
. The packet is sent to all nodes in radio range by specyfing AM_BROADCAST_ADDR
as the destination address. Finally, the test against SUCCESS verifies that the AM layer accepted the message for transmission. If so, the busy flag is set to true. For the duration of the send attempt, the packet is owned by the radio, and user code must not access it. Note that we could have avoided using the Packet
interface, as it's getPayload
command is repeated within AMSend
.
/** * Signaled in response to an accepted send request. msg is * the message buffer sent, and error indicates whether * the send was successful. * * @param msg the packet which was submitted as a send request * @param error SUCCESS if it was sent successfully, FAIL if it was not, * ECANCEL if it was cancelled * @see send * @see cancel */ event void sendDone(message_t* msg, error_t error);
This event is signaled after a message transmission attempt. In addition to signaling whether the message was transmitted successfully or not, the event also returns ownership of msg
from AMSend
back to the component that originally called the AMSend.send
command. Therefore sendDone
handler needs to clear the busy
flag to indicate that the message buffer can be reused:
event void AMSend.sendDone(message_t* msg, error_t error) { if (&pkt == msg) { busy = FALSE; } }
Note the check to ensure the message buffer that was signaled is the same as the local message buffer. This test is needed because if two components wire to the same AMSend
, both will receive a sendDone
event after either component issues a send
command. Since a component writer has no way to enforce that her component will not be used in this manner, a defensive style of programming that verifies that the sent message is the same one that is being signaled is required.
implementation { ... components ActiveMessageC; components new AMSenderC(AM_BLINKTORADIO); ... }
These statements indicate that two components, ActiveMessageC
and AMSenderC
, will provide the needed interfaces. However, note the slight difference in their syntax. ActiveMessageC
is a singleton component that is defined once for each type of hardware platform. AMSenderC
is a generic, parameterized component. The new
keyword indicates that a new instance of AMSenderC
will be created. The AM_BLINKTORADIO
parameter indicates the AM type of the AMSenderC
. We can extend the enum
in the BlinkToRadio.h
header file to incorporate the value of AM_BLINKTORADIO
:
... enum { AM_BLINKTORADIO = 6, TIMER_PERIOD_MILLI = 250 }; ...
implementation { ... App.Packet -> AMSenderC; App.AMPacket -> AMSenderC; App.AMSend -> AMSenderC; App.AMControl -> ActiveMessageC; }
Receiving a Message over the Radio
Now that we have an application that is transmitting messages, we can add some code to receive and process the messages. Let's write code that, upon receiving a message, sets the LEDs to the three least significant bits of the counter in the message. To make this application interesting, we will want to remove the line
from the call Leds.set(counter);Timer0.fired
event handler. Otherwise, both the timer events and packet receptions will update the LEDs and the resulting effect will be bizarre.
If two motes are programmed with our modified application, then each will display the other mote's counter value. If the motes go out of radio range, then the LEDs will stop changing. You can even investigate link asymmetry by trying to get one mote's LEDs to keep blinking while the other mote's LEDs stop blinking. This would indicate that the link from the non-blinking mote to blinking mote was available but that the reverse channel was no longer available.
- Identify the interfaces (and components) that provide access to the radio and allow us to manipulate the
message_t
type.
We will use theReceive
interface to receive packets. - Update the module block in the BlinkToRadioC.nc by adding uses statements for the interfaces we need:
- Declare any new variables and add any needed initialization code.
We will not require any new variables to receive and process messages from the radio. - Add any program logic and calls to the used interfaces we need for our application.
Message reception is an event-driven process so we do not need to call any commands on theReceive
. - Implemement any (non-initialization) events specified in the interfaces we plan on using.
We need to implement theReceive.receive
event handler: - Update the implementation block of the application configuration file by adding a components statement for each component used that provides one of the interfaces chosen earlier.
The following lines can be added just below the existingcomponents
declarations in the implementation block ofBlinkToRadioAppC.nc
: - Wire the the interfaces used by the application to the components which provide those interfaces.
Update the wiring by insert the following line just before the closing brace of theimplementation
block in BlinkToRadioAppC: - Test your application!
Testing your application is easy. Get two motes. They can be mica2, micaz, telosa, telosb, or tmote. For this exercise, let's assume that the motes are telosb (if not, skip past the motelist part and program the mote using whatever the appropriate programmer parameters are for your hardware). Assuming you are using a telosb, first open a Cygwin or Linux shell and cd to theapps/tutorials/BlinkToRadio
directory. Then, insert the first telosb mote into an available USB port on the PC and typemotelist
the at the Cygwin or Linux prompt ($). You should see exactly one mote listed. For example:
module BlinkToRadioC { ... uses interface Receive; }
event message_t* Receive.receive(message_t* msg, void* payload, uint8_t len) { if (len == sizeof(BlinkToRadioMsg)) { BlinkToRadioMsg* btrpkt = (BlinkToRadioMsg*)payload; call Leds.set(btrpkt->counter); } return msg; }
The receive
event handler performs some simple operations. First, we need to ensure that the length of the message is what is expected. Then, the message payload is cast to a structure pointer of type BlinkToRadioMsg*
and assigned to a local variable. Then, the counter value in the message is used to set the states of the three LEDs.Note that we can safely manipulate the counter
variable outside of an atomic section. The reason is that receive event executes in task context rather than interrupt context (events that have the async
keyword can execute in interrupt context). Since the TinyOS execution model allows only one task to execute at a time, if all accesses to a variable occur in task context, then no race conditions will occur for that variable. Since all accesses to counter
occur in task context, no critical sections are needed when accessing it.
implementation { ... components new AMReceiverC(AM_BLINKTORADIO); ... }
This statement means that a new instance of AMReceiverC
will be created. AMReceiver
is a generic, parameterized component. The new
keyword indicates that a new instance of AMReceiverC
will be created. The AM_BLINKTORADIO
parameter indicates the AM type of the AMReceiverC
and is chosen to be the same as that used for the AMSenderC
used earlier, which ensures that the same AM type is being used for both transmissions and receptions. AM_BLINKTORADIO
is defined in the BlinkToRadio.h
header file.
implementation { ... App.Receive -> AMReceiverC; }
$ motelist Reference CommPort Description ---------- ---------- ---------------------------------------- UCC89MXV COM17 Telos (Rev B 2004-09-27)
Now, assuming you are in the apps/tutorials/BlinkToRadio
directory, type make telosb install,1
. You should see a lot text scroll by that looks something like:
$ make telosb install,1 mkdir -p build/telosb compiling BlinkToRadioAppC to a telosb binary ncc -o build/telosb/main.exe -Os -O -mdisable-hwmul -Wall -Wshadow -DDEF_TOS_AM_GROUP=0x7d -Wnesc-all -target=telosb -fnesc-cfile=build/telosb/app.c -board= BlinkToRadioAppC.nc -lm compiled BlinkToRadioAppC to build/telosb/main.exe 9040 bytes in ROM 246 bytes in RAM msp430-objcopy --output-target=ihex build/telosb/main.exe build/telosb/main.ihex writing TOS image tos-set-symbols --objcopy msp430-objcopy --objdump msp430-objdump --target ihex build/telosb/main.ihex build/telosb/main.ihex.out-1 TOS_NODE_ID=1 ActiveMessageAddressC$addr=1 found mote on COM17 (using bsl,auto) installing telosb binary using bsl tos-bsl --telosb -c 16 -r -e -I -p build/telosb/main.ihex.out-1 MSP430 Bootstrap Loader Version: 1.39-telos-8 Mass Erase... Transmit default password ... Invoking BSL... Transmit default password ... Current bootstrap loader version: 1.61 (Device ID: f16c) Changing baudrate to 38400 ... Program ... 9072 bytes programmed. Reset device ... rm -f build/telosb/main.exe.out-1 build/telosb/main.ihex.out-1
Now, remove the first telosb from the USB port, insert the batteries, and set it aside. Insert the second telos into the USB port and once again type motelist
. You should again see something like:
$ motelist Reference CommPort Description ---------- ---------- ---------------------------------------- UC9VN03I COM14 Telos (Rev B 2004-09-27)
Finally, type make telosb reinstall,2
and you should once again see something like the following scroll by:
$ make telosb reinstall,2 tos-set-symbols --objcopy msp430-objcopy --objdump msp430-objdump --target ihex build/telosb/main.ihex build/telosb/main.ihex.out-2 TOS_NODE_ID=2 ActiveMessageAddressC$addr=2 found mote on COM14 (using bsl,auto) installing telosb binary using bsl tos-bsl --telosb -c 13 -r -e -I -p build/telosb/main.ihex.out-2 MSP430 Bootstrap Loader Version: 1.39-telos-8 Mass Erase... Transmit default password ... Invoking BSL... Transmit default password ... Current bootstrap loader version: 1.61 (Device ID: f16c) Changing baudrate to 38400 ... Program ... 9072 bytes programmed. Reset device ... rm -f build/telosb/main.exe.out-2 build/telosb/main.ihex.out-2
At this point, both motes should be blinking their LEDs. If you press the RESET button on either telosb, then the LEDs on the other telosb will pause on whatever was being displayed at the moment you pressed RESET. When you release the RESET button, the paused mote will be reset and then resume counting from one.
Conclusions
This lesson has introduced radio communications in TinyOS 2.x.
Related Documentation
<references/>
< Previous Lesson | Top | Next Lesson >