
2/85

4.1.2 lPv4 C lient,

IPv6 server

A

general

property

of a dual-stack

host s that IPv6 servers

an

handleboth

IPv4

and

IPv6 clients.

This is done using IPv4

mapped Pv6 addresses.

igure

4.2

shows

an example

of

this.

lPv6 istening

socketbounded

to 0::0,

port

8888

lPv4-mapped

lPv6address

lPvO ddress

zvo.oz.zzo.+z

5f1 :df00:ce3e:e200:

2O:80O:2b37:6426

Type

Dport

0800 8888

J;ffi?ffi'

Fig 4.2 IPv6 server on dual

-stack

host serving IPv4 and

IPv6

clients

From the fig 4.2 :

.

An IPv4 client and an IPv6 client are on the eft. The server

on

the

risht is written

using Pv6 and t is runningon a dual-stack ost .

.

The server has createdan IPv6 listenrngTCP socket hat

is

bound to the

IPv6

wildcardaddress

nd

TCP

port

8888.

.

Assume hat the clients and serverare on the sameEthernet.

.

Routerscould alsoconnect hem,as ong asall the routerssupport Pv4 and Pv6.

.

Assume hat both clients send

SYN sesments o

establish

a

connectionwith

the

server.

IPV4

cl ient

.

The IPv4 client host will

send he SYN in an IPV4 datagrams

.

The TCP

segment

rom

the IPv4 client appears

n

the

wire

as

an Ethernetheader

followed by and IPv4 header, TCP header,

nd the TCP data.

.

The Ethernetheader ontainsa

type field of oxo800,which identrfies he frame

as

an

IPv4

frame.

.

The TCP header ontains he dest inat ion

ort

of 8888.

.

The destination P address n the IPv4 headerwould

be 206.62.226.42.

lHVtl

cl ient

lPv4

cl ient

Enet

nd r

Pv 4

hdr hdr

TC P

oata

no r

Enet

lPv6

no r

TCP

hd r

TC P

oala

4.4

Network rogramming

nd Management


3/85

IPVS client

.

The

IPv6

client's

host

will send

he

SYN

in an

IPv6

datagram.

.

The

TCP

segment

rom

the

IPv6 client

appears

on

the

wire as

an

Ethernet

header

followed

by

and

IPv6

anlPv6

header,

TCP

header,

and

the

TCP

data.

.

The Ethernet

header

contains

a type

field

of

0x86dd,

which

identifies

the

frame as

an IPV6

frame.

.

The TCP

header

n the

IPV6

packet

has

he

same

ormat

as the

TCP

header

n the

IPV4 packet,and contains he destinationport of 8888'

.

The

des t ina t ion

IP

address

in

the

IPv6

header

wou ld

be

5 I b :df00

c

e3e:200

20

800

2b37

6426.

The

receiving

datalink

looks

at the

Ethernet

type

field

and

passes

each

frame

to the

appropriate

P module.

Address

conversion

.

The

IPv4

module,

probably n conjunction

with

the

TCP

module,

detects

hat he

destination

socket

s

an IPv6

socket,

and the

source

PV4

address

n

the IPV4

header

s converted

nto the equivalent

PV4-mapped

PV6

address.

. That mappedaddress s returned o the IPv6 socketas the client's IPv6 address

when accept

eturns

o

the server

with the

IPv4 client

connection'

.

All

remaining

datagrams

or

this connection

are

Pv4

datagrams.

.

When accept

eturns

o

the

server

with

the

IPv6 client

connection,

he

client's

IPv6

address

does

not change

rom

whatever

source

address

appears

n

the IPv6

header.

.

All remaining

datagrams

or this connection

are

Pv6

datagrams.

We can summarize

he

steps hat

allow

an

IPv4

TCP client

to communicate

with an

IPv6 server.

l. The lpv6 serverstarts,createsan IPv6 listeningsocket,and t binds the wildcard address

to the

socket.

2.

The IPv4

client calls

gethostbyname

and finds

an

A record

for

the server.

The

server

host

will

have both

an

A record and

a AAAA

record, since

t supports

both

protocols

but the

IPv4

client asks

for

only an

A record.

3.

The client

calls connect

and

he client's

hostssend

an

IPV4 SYN

to

the server.

4. The server

host eceives

he

IPv4 SYN

directed

o

the

IPV6

listening

socket,

etaa

flag

indicating

hat

this connection

s using

IPV4

-mapped

IPV6

addresses

nd

responds

with an

IPV4 SYN/ACK.

When the connection

s established,

he

address

eturned

to

the server

by accept

s the

IPv4 mapped

Pv6 address.

5.

All

communication

between his client and server akesplace using IPv4 datagrams.

AdvoncedSockefs

4.5


4/85

6.

Unless the

server explicitly

checks

whether

his IPv6 address s and IPv4 mapped

Pv6

address

using

the

IN6_Is_l DDR_V4MAPPED macro),

the server never knows that it

is

communicatingwith

an IPv4 client. The

dual

protocol

stack handles his detail.

Similarly,

the

IPv4

client has no idea

that

it

is communicating with an IPv6

server.

An

underlying

assumption n this scenario s

that the dual

-

stack server host has

both

an IPv4 address

and an IPv6

address.This will work

until all

the IPv4

addresses re taken.

The

scenario s similar for an IPv6

UDP server,but the

address

ormat

can change or

each datagram.For example, f the IPv6 server receives a datagram rom an IPv4 client, the

address eturned

by recvfrom will

be the client's IPv4

mapped IPv6 address. The

server

responds

o this

client's requestby calling

sendto

with

the IPv4 mapped Pv6

address s he

destination.

his

address ormat tells the kernel

o sendan IPV4

datagram

o

the client. But

the next datagram

eceived for the

server would

be

IPV6

datagram.But the next datagrams

received or

the servercould

be an IP

v6

datagrams,

nd recvfrom will return the IPv6

address.

It the

server esponds,

he kernel will

generate

an IPv6

datagrar4.

Figure

4.3 summarizes

ow a received

Pv4 or IPv6 datagram s

processed,

epending

on the

type

of

the receiving

socket, for

TCP and

UDR assuming a dual stack host.

lPv4sockets

AF-INET

SOCK-DGRAM

sockaddr_in

AF_INET6

SOCK-STREAM

sockaddr_in6

AF ' INET6

SOCK-DGRAM

sockaddrn6

{ , , *

4.6

revo

ocr


5/85

o Ifan

IPv6 datagram

s

received

for an

IPv6 socket,

nothing

special

is done

These are the

two arrows

labeled

IPv6

in the

figure, one

to TCP

and one

to

UDP.

IPv6 datagrams

are exchanged

between

he client

and server.

o But

when an

IPv4 datagram

s

received

for an

IPv6 socket,

the

kernel

returns

the corresponding

IPv4 mapped

Pv6 address

as

the address

eturned

by

accept(TCP)

or recvfrom

(UDP).

These

are the two

dashed

arrows

in the

figure.

This mapping

s

possible

because

n

IPv4 address

an

always

be

representedas an IPv6 address. PvA datagrams re exchangedbetween he

client and server.

The converse

ofthe

previous

bullet

(an

Ipv6

datagram

s received

or

an Ipv4

socket)

is

false:

in

general

an

IPv6 address

connote

be represented

as an

IPv4 address:

herefore

there are no arrows

from the

IPv6

protocol

box to

the two

IPv4 sockets.

Most dual- stack

hostsshould

use he

following

rules

n

dealing

with listening

sockets:

l. A listening Pv4

socketcan

accept

ncoming connections

rom only

IPv4 clients.

2. Ifa

server

has a

listening Pv6

socket hat

hasbound

he

wildcard address,

hat socket

can accept incoming connections from either IPv4 clients or IPv6 clients. For a

connection

rom an Pv4 client

the server's

ocal address

or the

connection

will

be he

corresponding

Pv4 mapped

Pv6 address.

3.

If

a server

has a l istening

IPv6 socket

hat has bound

an IPv6 address

other than an

IPv4 mapped Pv6 address,

hat socket

can accept

ncoming connections

rom

IPv6

clients only.

4.1.3 IPv6 cl ient , IPv4

serve

Consideran

IPv6 TCP client running

on a dual

-stack

host

l. An IPv4 serverstartson an

lPv4-only host and

createsan IPv4

listening socket.

2. The IPv6 client starts,calls

gethostbyname

sking or only

IPv6 addresses

It

enables

the

RES_USE_INET6

option). Since he

IPv4-only server

host has only A records, an

IPV4-mapped IPv6 address

s returned to the client.

3. The

IPv6

client calls

connectwith the lPv4-mapped

Pv6 address n the

IPv6

socket

addressstructure.

The kernel

detects

he mapped addressand

automatically sendsan

IPv4 SYN to the server.

4.

The server esponds

with

an IPv4 SYN/ACK, and he

connection s established

sing

IPv4 datagrams.

We can summarize his scenario n Fieure4. 4

Advonced

Sockels

4.7


6/85

lPv4sockets

AF-INET

SOCK-DGRAM

sockaddr_in

{

{ , * .

AF_INET6

SOCK-STREAM

sockaddr_in6

AF-INET6

SOCK-DGRAM

sockaddr_in6

Pv6sockets

address

returned y

accept r

recvfrom

lPv4datagram lPv6datagram

Fig, 4.4 processing of client rcquests, depending on address type and socket type

.

If

an

IPv4 TCP client

calls

connect

specifyingan IPv4 address,

r if

an

IPv4 UDP

client calls send o specifyingan Pv4 address, othing special s done.Theseare

the two arrows labeled

IPv4

in the figure.

.

If an IPv6 TCP client calls connect

specifyingan

IPv6

address, r if an IPv6 UDP

client calls sendtospecifyingan IPv6

address,

othing

special s doneTheses re

the two

arrow

labeled

IPv6 in

the figure

.

If

an

IPv6 TCP

client specifiesan IPv4-mapped Pv6

address

o

connector if an

IPv6 UDP client

specifies an IPv4-mapped IPv6 address

o sendto, the

kernel

detects he mappedaddress

nd causes

n

IPv4 datagram o

be sent,

nstead

ofan

IPv6 datagram.Theseare the two dashedarrows in the figure.

.

An IPv 4 client

cannot

specify

an Pv6 address o either

connector sendtobecause

a l6 -byte IPv6 address

oesnot fit in the 4-byte n_addr

structurewithin the IPv4

sockaddr_in structure. Therefore

there are no arrows from the IPv4

clients to the

IPv6

protocol

box

in

the

figure.

In

an IPv4 datagram arriving for

an IPv6 server socket, he

conversion ofthe received

address o the IPv4-mapped Pv6

address s done

by

the kernel

and returned transparently o

the

applicationby accept

or

recvfrom.

In

an IPv4 datagramneeding

o be senton an IPv6

socket he conversionofthe IPv4

address o

the

IPv4-

mapped Pv6 address

s done by the resolverand

the

mapped

addresss

thenpassed ransparently y the application o connector sendto.

Network rogramming

nd

Management

.8


7/85

Summary

of

interoPerabilitY

Figure

4.5 summarizes

his

section

and

the

previous section

and

the

combinations

of

clients

and

servers.

IPv4

server

IPv4-only

host

(A

only)

IPv6server

IPv6-only

host

(AAAA

only)

IPv4

server

dual-

stack

host

(A

and

AAAA)

IP6server

dual

stack

host

(A

and

AAAA)

lPv4

client,

IPv4-

onlY

host

IPv4

(no)

IPv4

IPv4

IPv6

cl

ent,

Pv6

only

host

(no)

IPv6

(no)

IPv6

IPv4

ci ent.

dual-stack

ost

IPv4

(no)

IPv4

IPv4

IPv6

client.

dual-

stack

host

IPv4

IPv6

(no*)

IPv6

Figure

4.5 Summary

of

interoperability

between

Pv4

and

IPv6

clients

and

servers

Each

box

contains IPv4

or

IPv6

if the

combination

s OK,

indicating

which

protocol

is used,

(no) -if

the combination

s invalid.

The

third

column

on

the

final row

is

marked

with

an asterisk

because

nteroperability

depends n the address hosenby the client.

Choosing

he

AAAA

record

and

sending

an

IPV6

datagram

will

not

work.

But

choosing

the

A record,

which

is returned

o the

client

as

an

IPv4-

mapped

Pv6

addresses'

auses

n

IPv4

datagram

o be

sent,

which

will

work'

Although

it appears

hat

one-fourth

of

the table

will

not

interoperate,

n the

real

world

for

the

foreseeable

uture,

most

implementations

f

IPv6

will

be on

dual

-

stack

hosts

and

will

not

be IPv6-

only

implementations.

f

we then

emove

he

second

ow

and

second

olumn,

al l

of the

(no)

entries

disappear

and the

only

problem

is the

entry

with

the

asterisk.

4.1.4 IPv6 Address Testing Macros

IPv6

Address

Macro,

Function,

Option

.

IPv6 address

esting

macros:

N6-IS-ADDR-*

(e.g. V4MAPPED)

.

Protocol

ndependent

ocket

address

unctions:

sock-*

(e.g.

cmp-addr)

.

IPv6_ADDRFORM

socket

option:

change

a socket

type

between

IPv4

and

IPv6,

by

setsockopt

unction

with

IPv6-ADDRFORM

option

There

are

small

classes

of

IPv6 applications

that

must

know

whether

they

are talking

to an

IPv4

peer.These

applications

need to

know

if the

peer's address

s an

IPv4

mapped

IPv6 address.Twelve macros are defined to test an IPv6 address or certain properties.

4.9

Advonced

Sockets


8/85

f,include

f_name);

4.40

Network rogrammingndManagement


39/85

)

pri

ntf

"end-of-f

e on

o/os\n

,

f

ptr->

-n ame)

Close(fd);

fptr->f-flags

F-DoNE;

/*

clears

F-READING

/

return(fpt0;

/*

terminate

hread

/

l

/*

end do_get_read

/

Fig.

4. 2

5

do-g et-rea

d

fu

n ctio

n

From he

ig.4.25,

These wo

decrements

re

placed n the do_get-read

unction

n

which

each

hread

decrement

hese

wo

counters

mmediately

efore

he hread

erminates.

t results

n a

slight

concurrent

rogrammlng rror.

The

problem n the

unction

hat

each hread

xecutes

s

that

hese wo

variables

re

global,

not thread-specific.

f one hread

s in

the middle

of decrementing

variable, hat

thread

s suspended,

nd

another

hread

executes

nd

decrements

he

same

ariable, n

error

can esult.

For

example,

Assume hat

the

C compiler

urns

the

decrement

perator

nto 3

instructions:

.

Load

rom

Memory

nto a

register.

.

Decrement

he

register.

.

Store

rom

the

register

nto

memory.

Possible

cenarios

Thread

A is

running

and

t loads

hevalue

of nconn

nto a

register.

l. The system

witches

hreads

rom

A to B.A's

egisters

re

saved nd

B's

registers

re

restored.

2. ThreadB executeshe hree nstructionsorrespondingo theC expressionsconn-

,

storing

he

valueof

2.

3.

Sometime

ater

he system

witches

hreads

rom B

to A.

A's

registers

re estored

nd

A

continues

here t left

off, at he

second

machine

nstruction

n the

hree-

nstructions

sequence:

he

valueof

the register

s decrements

rom

3 to

2

and he

value

of 2 is

stored n

nconn

The end

esult

s that

nconn

s 2

when t should

e

l. This

s

wrong.

These

ypes

of

concurrent

rogramming

rors arehard

o find

for numerous

easons.

hey

are

o they

occur

arely.

Nevertheless

t is an

error

and

t

will fail

(Murphy's aw)

.

o the error s hard o duplicate, ince, t depends n thenon-deterministiciming

of many

events.

AdvoncedSockels

4'41


40/85

o

On

some systems

he

hardware

nstructions

might

be atomic;

that is there

exists

a hardware

instruction

to decrement

an integer

in memory

and the

hardware

cannot

be interrupted

during

this

instruction.

But this

is not

guaranteed

y all systems,

so the code works

on one

system

but not

on another.

Threads

programming

s

also called

as concurrentprogramming

or

parallel

programming

since

multiple

hreads an

be

running

concurrently

in

parallel)

accessing

he same ariables.

Fig.4.26

is a

simple

program

hat

creates

wo threads

and then has

each hread

ncrement

a global variable5000 times.

#include

"unpthread.h"

#define

NLOOP

000

int

counter;

/*

this s ncremented

y the

hreads

/

void

*doit(void

);

int

main(int

rgc,

char

*argv)

{

pthread_t

tidA, idB;

Pthread_create(&tidA,

ULL,

&doit,NULL);

Pthread_create(&tidB,

ULL,

&doit,NULL);

/* 4wait or both hreadso terminate

/

Pthreadjoin(tidA,

ULL)

Pthreadjoin(tidB,

ULL)

exit(0);

)

void

doit(void

vptr)

{

int

i , val;

t ,

*

Each

hread etches,

r ints,

nd ncrements

he

counterNLOOPimes.

*

The

valueof the counter

hould ncrease

monotonically.

*l

for

( i

=

0;

int

pthread_mutex_lock

pthread-mutex-t

mptr)

Int pthread_mutex_trylockpthread-mulex-t

mplr)

int

pthread_mutex_unlock

pthread_mutex-t

mptr)

Both

eturn:

if OK

positive

xxx

alue

on error

If

we

try to

lock

a mutex

hat s already

ockedby

some

other

hread,

we

are

blocked

until the

mutex

s unlocked.

o The

pthread_mutex_lock0

outine

s used

by a

thread

o acquire

lock

on the

specified

mutex

ariable.

f the

mutex

s already

ocked

by another

hread,

his

call

will

block

he

calling hread

ntil he

mutex

s unlocked.

o

pthread_mutex_trylock0

ill attempt

o

lock a

mutex.

However,

f the

mutex s

already

ocked,

he routine

will return

mmediately

ith

a

"busy"

error

code.

This

routine

maybe useful

n

preventing eadlock

onditions,

s n a priority-

inversion

ituation.

o

pthread_mutex_unlock0

ill unlock

a

mutex f

called

by the

owning

thread.

Calling

his

routine

s

required

fter

a thread

ascompleted

ts use

of

protected

data

f other

hreadsare

o acquire

he

mutex

or their

work

with the

protected

data.

An error

will be returned

f:

o

If the mutex

wasalready

unlocked

o

If the

mutex

s owned

bv another

hread

Usage

o

Mutex

variables

must be

declared

with type

pthread-mutex-t,

nd

must

be

initialized

before

hey can

be used.

There

are two

ways to

initialize

a mutex

variable:

l. Statically,

hen t is

declared.

or

example:

pthread-mutex-tmymutex

=

PTHREAD-

MUTEX-INITIALIZER;

2.

Dynamically,

with the

pthread_mutex_init0

outine.

This

method

permits

setting

mutexobject

attributes,

/lr.

Fig.4.27

s a corrected

ersion f

Fig.4.26

hat

uses

single

mutex

o ock

hecounter

betweenhe wo threads.

Advonced Sockefs

4.43


42/85

#include

"unpthread.h"

#define

NLOOP 000

int

counter;

/*

this

s ncremented

y the hreads

/

pthread_mutex_t

counter_mutex PTHREAD_MUTEX_lNlTlALlZER;

void

*doit(void );

in t

main(int

rgc,char

*argv)

t

pthread_t

tidA, idB;

Pthread_create(&tidA,ULL,&doit,NULL);

Pthread_create(&tidB,ULL,

&doit,

NULL);

l*

4wail or

both threads o terminate

/

Pthreadjoin(tidA, ULL)

Pthreadjoin(tidB, ULL)

exit(0);

)

void

*

doit(void

vptr)

{

int i , val;

l*

* Each hread etches, rints, nd ncrementshe counterNLOOPimes.

*rThe

value

of

the counter hould ncreasemonotonically.

for

( i

=

0;

sasend->sa_family,OCK_RAW,

r->icmpproto);

setuid(getuid0);

/*

don't

need

pecial

ermissions

nymore

/

size

=

60

*

1024;

/*

OK f setsockoptails

/

setsockopt(sockfd,

OL_SOCKET,O_RCVBUE

size, izeof(size));

sig_alrm(SIGALRM);

*

send

irst

packet

/

f o r ( ; ; ) {

1sn

pr->salen;

n

=

recvfrom(sockfd,ecvbuf,

izeof(recvbuf),

,

pr->sarecv,

len);

i l ( n < 0 ) {

i f

(errno

=

EINTR)

continue;

else

err_sys("recvfromrror")

)

Gettimeofday(&tval,ULL)

(*pr->fproc)(recvbuf,

, &tval);

)

)

Fig. 4.

3

4 read

oop

unct

ion

Create

socket

o

{ raw socketofthe appropriate

rotocol

s created.

. The call to setuid setsour eff ective user D to our real user D.

.

The

program

must have

super

user

privileges

o

create he raw socket,but now

that

the

socket

s

created,and the extra

privileges

are

given

up .

.

It is

always

best

o

give

up this extra

privilege

when it is no longer needed.

Set socket receive buffer

.

The socket

eceive buffer

size

s

set to 61,440 bytes(60X1024)

which should be

larger than the default.

.

The

user

pings

either the

IPv4

broadcast

address

or a relticast addresseither of

which

can

generate

lots of replies.

.

By making the buffer large4 thereis a smaller chance hat the socketreceive buffer

will overflow.

Advonced

Sockets

4.65


64/85

Send

irst

packet

.

Signal

handler

s called

which sends

packetand

schedules

SIGALRM

for one

second

n future.

.

It is

not common

o see

a signal

handler

called

directly

but

it is

ok .

o

d

s igna l

hand le r

is

jus t

a

C func t ion ,

eventhough

t

is norma l ly

ca l led

asynchronously

y the

kernel.

Infinite

loop

reading

all ICMP

messages

.

The

main oop

of

the

program s

an nfinite

loop

that

reads

all

packets

eturned

on

the

RAW

ICMP

sockets.

.

Gettimeofday

s called

to

record

the time

that

the

packet was

received

and

then

call

the

appropriate

protocol function(proc-v4

or

proc-v6) to

process

ICMP

messages.

Fig .4.35

shows

the

proc_v4

function

which

processes

all

received

CMPv4

messages.

.

It

is also realizedthat

when the

ICMPv4

message

s received

by

the

process n the

RAW

socket,

he kernel

has already

verified

that

the basic

fields

in the

IPv4

header

and

n the

ICMPv4

header

are

valid.

# inc lude

"p ing .h"

void

proc_v4(char

ptr,

ssize-t

en,struct

imeval

tvrecv)

{

in t h len l ,

cmp len ;

double

rtt ;

struct

p

* ip;

struct

cmp

*icmp;

struct imeval

tvsend:

ip = (structp *) ptr; /* startof lP header'/

hlenl

=

ip->ip_hl

icmp_type

=

ICMP_ECHOREPLY)

if

( icmp->icmp_id

=

Pid)

return;

/"

not a respon se

o our

ECHO-REQUEST

/

if

( icmplen

16)

err_quit(" icmplen

%d)

sasend ai->ai_addr; /* contains estinationddress /

4.76

Network

Programming

nd

Management


75/85

pr->sarecv

Cal loc(1,

i ->ai_addr len);

pr->salast

=

Cal loc(1

ai ->ai_addr len);

pr->sabind

=

Cal loc(1, i ->ai_addr len);

pr->salen

=

ai ->ai_addr len;

t raceloop0;

exi t (0) ;

Fig.4,45 Main

function

for

traceroute

prograrfl'

Define

proto

structures

The two

proto

structures

redefined,

one

or IPv4

and

one

or IPv6,

although

he

pointer

o

thesocket

ddress tructures

renot allocated

ntil

the end

of

this

unction.

Set

Defaults

.

The

maximum

TTL

or Hop limit

that the

program

uses

defaults

230.

.

The

-m

commandline

option

is

provided o let

the user

change

his'

.

For each

TTL three

probe packetsare sent

but

this could

be

changed

with

anothercommand ine option.

.

The intial

destination

port

is 32768

+

666 and

his

will be

incremented

y

one

each ime

a

UDP

datagram

s sent.

.

These

ports

are not in use on

the destination

ost

when the

datagrams

inally

reach he

destination. ut there

s

no

guarantee.

Process

Command

Line

Argument

The

-v

command

ine option causes

most

received

CMP

messages

o be

printed.

Process ostname

or IP addressargument

and f inish

init ia l ization

.

The destination

host nameor

IP address

s

processed y our

host-serv

unction

returning

a

pointer

to an addrinfo

structure

.

Depending

on the

type of returned

address

Pv4

or IPv6,

initializing

the

proto

structure,store

he

pointer

n the

pr

global

and

allocate

additional

socket

address

structureofthe correct

size

are finished.

.

The function

traceloop

shown

in fig.4.46

sends

he datagrams

and

reads

he

returned CMP messages. his is the main loop of the program.

4.77

dvonced

Sockefs


76/85

#include

" t race.h"

void

traceloop(void)

t

in t

seq, code,

done;

double

r t t ;

struct rec

*rec:

struct imeval tvrecv;

recvfd

=

Socket(pr- sasend-

>

sa_f mily,

SOCK_RAW,

r-

>

cmpproto)

setuid(getuid0);

/ *

don' t

need

specia l

permissions

ny

more

*/

sendfd

=

Socket (pr->sasend->sa_fami ly,

OCK_DGRAM,

);

pr-

>

sabind-

>

sa_family

=

pr-

sosnd- sa_f

mi y;

sport

=

(getpid0

& oxff f f )

0x8000;

/*

our source UDP

porl#

*l

sock_set_port (pr->sabind,

r->salen,

tons(sport ) ) ;

Bind(sendld,

r->sabind,pr->salen);

s ig_al rm(SIGALRM);

seq

=

0'

done

=

0;

f o r

( t t l

=

1 ; t t l

tt l level,pr->tt loptname,

&ttl, sizeof(int))

bzero(pr- salast,

pr-

>

salen)

pr int f ( "%2d

,

t t l ) ;

f f ush(stdout ) ;

for

(probe

=

0;

probe rec_seq ++seq;

rec->rec_ttl

=

t t l ;

Gettimeofday(&rec- rec_tv,NULL)

sock_set_port(pr->sasend,

r->salen,

htons(dport+ seq));

Sendto(sendfd,

endbuf ,datalen, ,

pr->sasend, r ->salen);

if

( (code

=

(*pr->recv)(seq,

&tvrecv))

==

-3 ;

pr int f ( "

" ) '

/ *

t imeout ,

no reply

* /

e lse

{

char st r INI_MAXHOST] ;

if

(sock_cmp_addr(pr->sarecv,

r->salast,pr->salen)

: 0)

{

if (getnameinfo(pr->sarecv, r->salen,str,sizeof(str),

N U L L , 0 , 0 )

=

g ;

printf("

%s

(o/os)",

t r,

Sock_ntop_host(pr-

sarecv,

pr-

>

salen))

e lse

printf("

7os",

Sock_ntop_host(pr->sarecv,

r->salen));

memcpy(pr->salast ,

r ->sarecv, r ->salen);

)

tv_sub(&tvrecv,

&rec-> rec_tv)

rt t

=

tvrecv.tv_s ec 1000.0 +

tvrecv.tv_us ec 1000.0;

printf("

%.31ms", rt t );

if

(code

==

-1)

/*

port

unreachable;

t destination

/

done++:

4.78

Network rogrammingnd Management


77/85

pr->sarecv

Cal loc( ' | ,

i ->ai_addr len);

pr->salast

Cal loc(1,

i ->ai_addr len);

pr->sabind

=

Cal loc(1,

i ->ai_addr len);

pr->salen

=

ai ->ai_addr len;

t raceloop0;

exi t (0) ;

Fig.4.45

Main

function

or

traceroute

rograu.

Define

proto

structures

The two

proto

structures

are

defined,

one

for IPv4

and

one

for

IPv6, although

the

pointer

o the socket

address

tructures

re not allocated

until

the

end of

this

function.

Set Defaults

'

The

maximum

TTL or

Hop limit

that the

program usesdefaults

230.

.

The

-m

commandline

option

is

provided o let

the user

change

his.

.

For each

TTL three

probe packetsare sent

but

this could

be changed

with

anothercommand ine option.

.

The intial

destination

port

is 32768

+

666 and

this

will be

incremented

y one

each

ime a UDP

datagram

s sent.

.

These

ports

are

not in use on

the destination

host

when

the

datagrams

inally

reach

he

destination, ut there

s

no

guarantee.

ProcessCommand

Line Argument

The

-v

command

ine option causes

most

received

CMP

messages

o

be

printed.

Process

ostname

or

IP addressargument

and

finish

init ia l ization

.

The destination

host

name or

IP address

s

processed y

our host-serv

unction

returning

a

pointer

to an addrinfo

structure

.

Depending

on the type

of returned

address

Pv4

or

IPv6,

initializing

the

proto

structure,

tore

he

pointer

n the

pr

global

and

allocate

additional

socket

address

structure

ofthe correct

size

are

finished.

.

The function

traceloop

shown

in fig.4.46

sends

he datagrams

and reads

he

returned CMP messages. his is the main loop of the program.

4.77

Advonced

Sockets


78/85

else

f

(code

>

=

0)

pr intf(" ICMP

%s)",

*pr->icmpcode)(code)) ;

)

fflush(stdout);

)

pr intf" \n") ;

i

Fig.4.46 Traceloop

unction

: main

processing loop.

Create wo

sockets

.

Two sockets

re

needed:

raw socket n

which

all returned

CMP

are ead

and

a

Udp socket

on

which a

probe

packets

resent

with the

ncreasing

TLs.

.

After

creating

raw socket,

ureffective

ser

D to

our

ealuser

D is

reset,

ince

superuser

rivileges renot onger

equired.

Bind source

ort

of UDP

sockets

.

Bind a source

ort

o theUDP

sockethat

s used

or sending,

sing

he ow

order

l5 bits

of our

process

D

with thehigh order

bit

set o

I .

.

Since t

is

possible

or multiple opies

f the

race

outeprogramo be unning t

any

given

ime

it

sneededo determine

f a

received

CMP

message

as

generated

in response

o oneof our datagrams,

r in response

o a dtagram

ent

by another

copy

of the

program.

.

Thesource

ort

nt theUDPheader

s used o

dentify

he

sending

rocess ecause

the CMP

message lways

eturns

he UDP header

rom

the

datagram

hat

caused

the CMP

error.

Establish

ignal

handler

or SIGALRM

Sig_alrm

unction

s established

s he signal

handler

or

SIGALRM

because

ach

timea UDPdatagrams sentand3 secondss requiredo wait for an CMPmessageefore

sendinghe

next

probe.

Main oop;set TL

or hop imit and

send

probes

.

The

main

oop of the unction

s a

double ested

or

loop'

.

The

outer oop

stars he

TTL

at I and

ncreases

y it to

l,while

the

nner oop

sends

probes

o the destination.

.

Each ime, he

TTL changes

etsockopt

s called

o set he

new

valueusing

either

IP_TTL

or

IPV6_UNICAST-HOPS

ocket

option.

. Each imearound heouter oop, he socketaddress tructures initialized o 0.

Advonced

Sockcls

4.79


79/85

.

This structure

will be compared o the

socketaddress

tructure

eturnedby

recvfrom

when he ICMP

message

s

read,and

f

the

2 structures

re different,

he IP address

from the

new structure s

printed.

.

Using this technique he IP address

orresponding

o

the first

probe

for each

TT L

is

pr inted

.

If the

IP

addresschanges

or a

given

value of

TTL, the new

IP

address

s then

printed.

Read CMP message

The

functions

recv_v4 0r recv_v6 calls recvfrom

to read

and

process

he ICMP

messages

that are eturned.

These

2 functions return

-3

if

a timeout

occurs,

2

if an ICMP

"time

exceeded

in transit"

error is received, I

if

an

ICMP

"port

unreachable"

rror

is received,or the

no n

negative CMP

code

f

some

other

ICMP

destination

error

is received.

Print

reply

.

If

this

is

a

first reply for

a

given

TTL or if

the IP addresss

f the node sending

he

ICMP message aschanged or the

TTL,the hostname

and P addresss re

printed.

.

The RTT is calculatedas the time difference rom when the probe is sent o the

time the ICMP m essages returnedand

printed.

Our

recv_v4 function is shown nthe

fig.4.47

#include

"trace.h"

*

Return:

3

on imeout

*

-2

on ICMP

imeexceededn ransit

cal ler

eeps

oing)

*

-1

on ICMP

ort

unreachable

cal ler

s

done)

int

recv_v4(int

eq,struct imeval

tv)

{

in t

h len1 , len2 ,cmplen ;

socklen_t

len;

ssize_t

n;

structp

* ip,

*hip;

structcmp

*icmp;

struct

dphdr

*udp;

alarm(3);

f o r ( ; ; ) {

;gn

=

pr->salen;

n

=

recvfrom(recvfd,

ecvbuf,

izeof(recvbuf),,

pr->sarecv,

len);

i f ( n < 0 )

i f

(errno

=

EINTR)

return(-3); l* alarm xpired

/

else

4.80

Network rogramming

nd

Management


80/85

err_sys("ecvf

om error")

I

Gettimeofday(tv,

ULL);

/*

get

time

of

packet

arrival

/

ip

=

(struct

p

*)

recvbuf;

*

start of

lP header

*/

h lenl

=

ip->ip_hl

icmp_type

=

ICMP-TIMXCEED

&

icmp->icmp_code

=

lcMP_TIMXCEED_INTRANS)

i f ( i c m p l e n < 8 + 2 0 + 8 )

err_qui t (" icmplen%d) < 8 + 20 + 8" , cmplen);

hip

=

(struct

p

*)

(recvbuf

+ hlenl

+ 8);

h l en2

=

h ip -> i p_h l

iP-P

==

IPPROTO-UDP

&

udp->uh_sPort

:

htons(sPort )

&

udp->uh_dport

==

htons(dport

seq))

return(-2);

/*

we

hit

an

intermediate

outer

*/

)

e lse

f

( icmp->icmp_type

=

ICMP-UNREACH)

i f ( i c m p l e n < 8 + 2 0 + 8 )

err_qui t (" icmplen

%d) i p_h l

iP-P= = IPPROTO-UDP &

udp->uh_sport

=

htons(sport ) &

udp->uh_dport

==

htons(dport

seq))

i l

( icmp->icmp_code

=

ICMP-UNREACH-PORT)

return(-1) ;

"

have eacheddest inat ion

/

else

re t u rn ( i cm p-> i cm p_code ) ;

*

0 ,

1 , 2 , . . . ' l

)

)

else f

(verbose)

printf("

(from

%s: type

:

%d, code

=

o/od)\n",

Sock_ntop_host(pr->sarecv,

r->salen),

icmp-

cmp_type,cmp->

cmp_code)

)

/*

Some

other

ICMP error,

ecvfrom$again

*/

) )

Fig.4.47

recv-v4

unclion:

readand

process

CMPv4

nrcssages

Set alarm

and read each

ICMP message

An alarm

s set

or

3

secondsn

the futureand

he

functionenters

loop hat calls

recvfrom, eading

ach CMPv4

message

eturned n he

aw socket.

Get

pointer

o

ICMP header

ip

points

o

the beginning f

the Pv4

header nd

cmp

points

o thebeginning

f the

ICMPheader.

Fig.4.48 hows hevarious eaders,ointers, nd engths sed y thecode'

Advonced

Sockels

4.81


81/85

n

h len l

icmplen

hlen2

lPv4

header

lPv4

options

ICMP4

header

lPv4

header

lPv4

options

UDP

header

o-40

t

8

I

udp

lPv4 atagramhat

generated

he

CMP

error

Fig,4.48 Headers,

pointers

and lengths in

processing ICMPv4 erron

Process

CMP time exceeded

n

transit

message

If

the ICMP

messages a

"

time exceeded n transi t" message,

t is

possibilily

a

reply

to one

of our

probes.

hip

points

to

the IPv4 header hat s returned n t he ICMP

message,

following

the

8

byte

ICMP header.

dp

points

o the UDP

header hat follows. If the

ICMP

message as

generated

y a UDP datagram nd f the sourceand destination

ports

of that

datgramare the values hat are sent, hen his is a reply to out probe from an intermediate

router.

Process

CMP

port

unreachablemessage

If t he ICMP messages a

"destination

unreachable", hen the UDP

header s returned

in the ICMP

message

o see f

the messages a response

o our

probe.

f the ICMP code

s

"port

unreachable"

1

is returned

and

the

final destinati on s reached. f t he ICMP me ssage

is from

one of ourprobes, but

it

is not

a

"port

unreachable" hen that ICMP

port

value s

returned.

Handle other ICMP messages

All

other ICMP

messages

re

printed

f the

-v

recv_v6

s shown infig.4.49

#include

"trace.h"

flag was

specified.The

next function

*

Return:

3

on imeout

*

-2

on ICMP

ime

exceededn

ransit

cal ler

eeps

oing)

*

-'l

on

ICMP

ort

unreachable

caller

s

done)

int

fzo

ovtes o-40

I

8

ip

icmp

t z o

hip

F--

4.82

Network rogramming

ndManagement


82/85

recv_v6(int

eq,struct

imeval

tv)

t

#ifdef lPV6

int

hlen' | ,

len2,cmP6len;

ssize_t

n;

socklen_t

len;

struct

p6_hdr

* ip6,

hiP6;

structicmp6_hdr

icmP6;

struct dphdr

*udp;

alarm(3);

f o r ( ; ; ) {

len

=

pr->salen;

n

=

recvfrom(recvfd,

ecvbuf,

izeof(recvbuf),

,

pr->sarecv,

len);

i f ( n < 0 ) {

i f

(errno

=

EINTR)

return(-3);

/*

alarm

exPired

/

else

err_sys("

ecvfom

error")

)

Gettimeofday(tv,

ULL);

/*

get

ime

of

packet

rrival

/

ip6

=

(struct

p6-hdr

)

recvbuf;

/r

start

of

lPv6header

/

hlenl

=

sizeof(struct

p6_hd0;

icmp6

=

(struct

cmp6-hdr

)

(recvbuf hlenl) ;

*

ICMP

dr

*/

i l

( ( icmp6len n

-

hlenl)

icmp6-type

=

ICMP6-TIME-EXCEEDED

&

icmp6->icmp6-code

=

ICMP6-TIME-EXCEED-TRANSIT)

i f ( i c m p 6 l e n < 8 + 4 0 + 8 )

err_quit(" icmp6len

%d)

Documents

Network Programmimg Unit 4