Agostinho L S Castro [email protected] Telecommunications and Multimedia Unit BSD TCP/IP Protocol Suite An Overview of the Implementation.

Agostinho L S [email protected]

Telecommunications and Multimedia Unit

BSD TCP/IP Protocol SuiteAn Overview of the Implementation

BSD TCP/IP Protocol Suite – The Implementation



•Network Implementation

4.4 BSD supports 4 distinct communication protocol families

1. TCP/IP

2. XNS (Xerox Network Systems)

3. OSI protocols

4. Unix domain protocols

•The TCP/IP protocol suite allows computers of all sizes, from different computer vendor, running totally different operating systems, to communicate with each other.

Application Telnet, FTP, e-mail, etc.

Transport TCP, UDP

Network IP, ICMP, IGMP

Link Device driver and Interface card

•TCP/IP is normally considered to be a 4 layer system




•General organization of networking code in 4.4 BSD




•Mbufs (Memory Buffers)

A fundamental concept in the design of the Berkeley Networking is the memory buffer, called an mbuf, used throughout the networking code to hold various pieces of information.

mbuf structure, mbuff macros and definitionssys/mbuf.h

kern/uipc_mbuf.cmbuf.h functions

linux/include/linux/skbuff.h







mbuf.h/* header at beginning of each mbuf */struct m_hdr {

struct mbuf *mh_next ; /* next buffer in chain */ struct mbuf *mh_nextpkt; /* next chain in queue record */

…………………………………………………………………………………… short mh_type; /* type of data */ short mh_flags; /* flags */………………………………………..…………………………………………..#define m_next m.hdr.mh_next#define m_len m.hdr.mh_len……………………………………………………………………..…………….. };

mbuf structuresm_flags Description

M_BROADCAST

sent/receive as link level broadcast

m_type Description

MT_data Dynamic data allocation







•Remarks on mbuf

1. The size of the mbuf structure is always 128 bytes

2. In each of the mbuf we show the m_data member pointing to the beginning of the corresponding buffer. (this pointer can point anywhere in the corresponding buff – not necessarily in the front)

3. The m_next pointer links together the mbuf forming a single packet into an mbuf chain

4. The m_next pointer links multiple packets together to form a queue of mbuf




Linked list of mbuf chains with head pointer




•Interface layer

4.4 BSD interface layer attempts to provide a hardware-independent programming interface between the network protocols and the drivers for the network devices connected to a system

The interface layer supports provides for all devices

i. A well-defined set of interface functionsii. A standard set of statistics and control flagsiii. A device-independent method of storing protocol addressi. A standard queueing method for outgoing packets




sys/socket.h address structure definitions

net/if.h interface structure definitions

net/if_dl.h link-level structure definitions

linux/init/main.c system and interface initialisations

net/init_main.c system and interface initialisations

drivers/net/skeleton.c generic interface code

net/if.c generic interface code

drivers/net/loopback.c loopback device driver

net/if_loop.c loopback device driver




•Interface layer

•Ifnet struture

•Ifaddr struture

•sockaddr struture




•ifnet structure

It contains information common to all interfaces.

During system initialization, a separate ifnet structure is allocated for each network device.

Every ifnet structure has a list of one or more protocol address associated with it.




ifnet description•Implementation information

List constructionCommon interface information (name, such as: eth0, sl0,...)

•Flags (if_flags specifies the operational state and

properties of the interface)If_flags Kernel

OnlyDescriptions

IFF_BROADCAST x The interface is for a broadcast network

IFF_MULTICAST x The interface supports multicasting

IFF_PROMISC The interface receives all network packets




ifnet description•Interface timer•BSD packet filter

Through BPF, a process can receive copies of packets transmitted or received by interface

•Hardware informationInterface characteristics (hardware address type supported by the interface – Ethernet , SLIP, loopback interface)mtu, metric, baudrate




ifnet description•Interfaces statistics•Functions pointers (interface procedures)

It contains pointers to the standard interface-layer

functions which isolate device-specific details from

network layer Function

Description

if_init Initialize the inteface

if_output

Queue outgoing packets for transmission

if_start Initiate transmission of packets




ifnet description•Output queue

Each interface has its own ifnet structure and therefore it own output queue

ifaddr structure •Each interface maintains a linked list of ifaddr structures because some data links, such as Ethernet, support more then one protocol•A separate ifaddr structure describes each address assigned to the interface, usually one address per protocol




if.hstruct ifaddr { struct ifaddr *ifa_next; /*next address for interface */ struct ifnet *ifa_ifp; /*back-pointer to interface */……………………………………………………………..…..};




sockaddr structure•Addressing information for an interface consists of more then single host address. 4.4 BSD maintains host, broadcast and network mask in structures derived from generic sockaddr structuresocket.h

struct sockaddr { u_char sa_len; /* total length */ u_char sa_family; /* address family*/.......................................................................};




sa_family Protocol

AF_INET Internet

AF_ISO, AF_OSI OSI

AF_UNIX Unix

AF_ROUTE Rpouting table

AF_LINK Data link

sa_family constants




Ifnet and ifaddr specialization

•The ifnet and ifaddr structures contain general information applicable to all network interface and protocol address. •To accommodate additional device and protocol-specific information, each driver defines and each protocol allocates a specialized version of the ifnet and ifaddr structures.

Arpcom structure is common to all Ethernet drivers and contains information for the Address Resolution Protocol (ARP)




Arragement of ifnet structure within device-dependent structures




Ethernet Initialization•cpu_start (init_main.c)

the kernel locates any attached network devicesle,sl,loop xxxattach – initialization function

•each device driver for a network interface initialize a specialized ifnet structure•if_attach function completes the initialization and inserts the structure into linked list the interface (ifnet structure specialized, arpcom structure)

After this initialization, the interfaces are configured only with link-level address







•Interface: Ethernet

File Description

net/if_ether.h Ethernet structures

net/if.h Ioctl command definitions

net/if_ethersubr.c

Generic Ethernet frames

dev/if_le.c Ethernet driver (LANCE)




if_le.cleintr Function

•leintr Function examines the hardware and if a frame has arrived. Calls lread function to transfer the frame from the interface to a chain of mbufs

leread Function•ether_header construction•chain of mbufs construction•it passes the incoming frames to BPF




ether_input Function•ether_input function examines the ether_header structure to determine the type of data has been received and then queues the received packet for processing•broadcast and multicast recognition•link-level demultiplexing•queue the packet




ether_output Function•output frames of Ethernet frames•network-level protocol (IP) call if_output function (ifnet structure)•if_output ~ ether_output (accept and queue frame for begin transmission of frame - verification

- protocol-specific processing- frame construction- interface queueing




lestart Function•interface must be initialized•dequeue frame from output queue•transmit frame and pass to BPF•repeat if device is ready for more frames•mark device as busy




•IP: Internet Protocol

File Description

net/route.h Route entries

netinet/ip.h IP Header

netinet/ip_input.c IP Processing

netinet/ip_output.c

IP output Processing

Netinet/in_cksum.c

Internet Checksum algorithm




IP datagram, including ip structure namesip structure - ip.h file




input processingipintr Function – ip_input.c

•verification of incoming packetsdequeueing packets from ipintrqverifies their contentsdamaged or erroneous packets are silently discardedIP version (ip_v = 4)IP checksumByte orderingPacket length




ipintr Function – ip_input.c •Determines whether or not the packet has reached its final destination

NO – forward (if systems is configured as a router)YES – transport-level protocol

•Reassembly and demultipexing•The protocol specified in the datagram is mapped by ip_p.




ip_forward Function

A packet arriving at a system other than its final destination needs to be forward. ipintr calls the function ip_forward which implements the forwarding algorithm




ip_output Function – ip_output.c

two sources: ip_forward and the transport protocols

•Header initialization The first section of ip_output merge options into the outgoing packets and completes the IP header for packets that passed from the transport protocolsIP header constructionPacket already including header (for a forward packet)




ip_output Function – ip_output.c•Route selection

After completing the IP header, the next task for ip_output is to locate a route to the destination

•Source address selection and fragmentationThe final section of ip_output ensure that the IP header has a valid source address and the interface associated with the route




•TCP: Transmission Control Protocol

File Description

netinet/tcp.h tcphdr structure definition

netinet/tcp_input.c

tcp_input

netinet/tcp_output tcp_outputtcp_debug.h, tcp_seq.h, tcp_timer.h, tcp_var.h, tcp_tcpip.h tcp_debug.c, tcp_subr.c, tcp_timer.c, tcp_usrreq.c

28 functions and almost 4,500 lines of C code




TCP Headertcp.h file




TCP Output – tcp_output.c

tcp_output is called whenever a segment to be sent on a connection

tcp_output determines whether a segment can be sent, and if so, what values to set all the TCP header fields




tcp_output description• Send a segment•it fills in all the fields in the TCP header and passes the segment to IP for output Calculate amount of data to sendBuild MSS (maximum segment size) option Build window scale optionBuild timestamp optionCheck if options have overflowed segmentUpdate statisticsAllocate an mbuf for IP and TCP headersCopy data into mbuf




tcp_output descriptionSet PSH (push flag) flagUpdate statisticsGet mbuf for IP and TCP headersCopy IP and TCP header templates into mbufSet sequence number filed of segmentSet acknowledgment field of segmentSet header length if option presentDon´t advertise less than one full-size segment




tcp_output descriptionObserve upper limit for advertised window on thisconnectionDo not shrink windowSet urgent offsetSet retransmission timerPersist stateAdd trace record for socket debuggingSet IP length, TTL, and TOSPass datagram to IP




TCP Input – tcp_input.c

•The function tcp_input is about 1100 lines code.

•The function tcp_input is called by ipintr when a datagram is received with a protocol field of TCP.





Preliminary processingGet IP and TCP in first mbufVerify TCP checksumVerify TCP offset fieldGet headers plus option into first mbufProcess timestamp option quicklySave input flags and convert fields to host byte orderLocate Internet PCB (protocol control block)




Preliminary processingDrop segment and generate RSTSilently drop segmentUnscale advertised windowSave connection state and TCP/IP headers if socket debug option enabledCreate new socket if segment arrives for listening socketComputer window scale factorReset idle time and keep alive timerProcess TCP options if not in LISTEN state



Bibliography

1. Wright, G. R., Stevens, W. R.,"TCP/IP Illustrated – The Implementaion",Volume 2.,Addison-Wesley, 1995 .

2. Wright, G. R., Stevens, W. R.,"TCP/IP Illustrated – The Protocol",Volume 1.,Addison-Wesley, 1994 .

3. Rubini, A., “LINUX Device Drivers”, O´Reilly& Associates, Inc.,1998.

4. Rusling, David A., “The Linux Kernel”, LDP – www.linuxdoc.org

Agostinho L S Castro [email protected] Telecommunications and Multimedia Unit BSD TCP/IP Protocol Suite An Overview of the Implementation.

Documents

bsd bsd tcpip protocol

multimedia unit mbuf

h bsd tcpip protocol

implementation agostinho

implementation slide

s castro

multimedia unit slide

queue of mbuf slide