[lwip-users] How to build a simple TCP server and client which run on the same machine
Antonio de Angelis
I'm using the win32 port and I'm trying to build a simple TCP server and
a TCP client on the same machine as two separate processes (using the
same ethernet adapter). The client connects to the server, sends a
messagge (i.e., "Hello") waits for the ACK and closes its connection.
The server waits for the client to connect, then when it receives the
incoming message, it prints the message on screen and closes its
connection. I'm using the NO_SYS = 1 option and the raw API. I've
created the server, and it seems to function: infact if I connect on the
specified port simply using a web browser(i.e.
http://192.168.1.200:8000) the server process prints correctly the first
string associated with the GET request from the browser. Here it is the
code:
#include "lwip/tcp.h"
static err_t server_accept(void *arg, struct tcp_pcb *pcb, err_t err);
static err_t server_recv(void *arg, struct tcp_pcb *pcb, struct pbuf *p,
err_t err);
static err_t server_poll(void *arg, struct tcp_pcb *pcb);
static err_t server_err(void *arg, err_t err);
void my_server_init(void)
{
struct tcp_pcb *pcb;
pcb = tcp_new();
tcp_bind(pcb, IP_ADDR_ANY, 8000); //server port for incoming connection
pcb = tcp_listen(pcb);
tcp_accept(pcb, server_accept);
}
static void server_close(struct tcp_pcb *pcb)
{
tcp_arg(pcb, NULL);
tcp_sent(pcb, NULL);
tcp_recv(pcb, NULL);
tcp_close(pcb);
printf("\nserver_close(): Closing...\n");
}
static err_t server_accept(void *arg, struct tcp_pcb *pcb, err_t err)
{
LWIP_UNUSED_ARG(arg);
LWIP_UNUSED_ARG(err);
tcp_setprio(pcb, TCP_PRIO_MIN);
tcp_arg(pcb, NULL);
tcp_recv(pcb, server_recv);
tcp_err(pcb, server_err);
tcp_poll(pcb, server_poll, 4); //every two seconds of inactivity of
the TCP connection
tcp_accepted(pcb);
printf("\nserver_accept(): Accepting incoming connection on
server...\n");
return ERR_OK;
}
static err_t server_recv(void *arg, struct tcp_pcb *pcb, struct pbuf *p,
err_t err)
{
char *string;
int length;
LWIP_UNUSED_ARG(arg);
if (err == ERR_OK && p != NULL)
{
tcp_recved(pcb, p->tot_len);
string = p->payload;
length = strlen(string);
printf("\nserver_recv(): Incoming string is %s\n", string);
printf("\nserver_recv(): String length is %d byte\n", length);
printf("server_recv(): pbuf->len is %d byte\n", p->len);
printf("server_recv(): pbuf->tot_len is %d byte\n", p->tot_len);
printf("server_recv(): pbuf->next is %d\n", p->next);
pbuf_free(p);
server_close(pcb);
}
else
{
printf("\nserver_recv(): Errors-> ");
if (err != ERR_OK)
printf("1) Connection is not on ERR_OK state, but in %d
state->\n", err);
if (p == NULL)
printf("2) Pbuf pointer p is a NULL pointer->\n ");
printf("server_recv(): Closing server-side connection...");
pbuf_free(p);
server_close(pcb);
}
return ERR_OK;
}
static err_t server_sent(void *arg, struct tcp_pcb *pcb, u16_t len)
{
LWIP_UNUSED_ARG(len);
LWIP_UNUSED_ARG(arg);
printf("\nserver_sent(): Correctly ACK'ed, closing server-side
connection...\n");
server_close(pcb);
return ERR_OK;
}
static err_t server_poll(void *arg, struct tcp_pcb *pcb)
{
static int counter = 1;
LWIP_UNUSED_ARG(arg);
LWIP_UNUSED_ARG(pcb);
printf("\nserver_poll(): Call number %d\n", counter++);
return ERR_OK;
}
static err_t server_err(void *arg, err_t err)
{
LWIP_UNUSED_ARG(arg);
LWIP_UNUSED_ARG(err);
printf("\nserver_err(): Fatal error, exiting...\n");
return ERR_OK;
}
Now i'm trying to create the client. But the communication between the
processes doesn't work: the client send the SYN segment but never
receives an answer, and the server never receives the connection request
from the client. I've built the server process giving it the address
192.168.1.200 and the client process with 192.168.1.201. They are
separate processes on the same machine but they use the same network
adapter. Here it is the code for the client:
static err_t client_sent(void *arg, struct tcp_pcb *pcb, u16_t len);
static err_t client_connected(void *arg, struct tcp_pcb *pcb, err_t err);
void my_client_init(void)
{
struct tcp_pcb *pcb;
struct ip_addr dest;
err_t ret_val;
IP4_ADDR(&dest, 192, 168, 1, 200);
pcb = tcp_new();
tcp_bind(pcb, IP_ADDR_ANY, 7000); //client port for outcoming connection
tcp_arg(pcb, NULL);
ret_val = tcp_connect(pcb, &dest, 8000, client_connected); //server
port for incoming connection
if (ret_val != ERR_OK)
printf("\tcp_connect(): Errors on return value, returned value
is %d\n", ret_val);
}
static void client_close(struct tcp_pcb *pcb)
{
tcp_arg(pcb, NULL);
tcp_sent(pcb, NULL);
tcp_close(pcb);
printf("\nclient_close(): Closing...\n");
}
static err_t client_connected(void *arg, struct tcp_pcb *pcb, err_t err)
{
char *string = "Hello!";
LWIP_UNUSED_ARG(arg);
if (err != ERR_OK)
printf("\nclient_connected(): err argument not set to ERR_OK,
but is value is %d\n", err);
else
{
tcp_sent(pcb, client_sent);
tcp_write(pcb, string, sizeof(string), 0);
}
return err;
}
static err_t client_sent(void *arg, struct tcp_pcb *pcb, u16_t len)
{
LWIP_UNUSED_ARG(arg);
printf("\nclient_sent(): Number of bytes ACK'ed is %d", len);
client_close(pcb);
return ERR_OK;
}
I'm using the test.c file "model" to build my applications: I've
removed from the test application all the code which I don't need and
then I've introduced the source code written for my applications. I use
the same test.c for both the client and server, compiling it with
my_client_init(); or my_server_init() for the client or the server. So
the code for test.c file (containing the main()) is this:
/* C runtime includes */
#include <stdio.h>
#include <stdarg.h>
#include <time.h>
#include <string.h>
#include <conio.h>
/* lwIP core includes */
#include "lwip/opt.h"
#include "lwip/init.h"
#include "lwip/tcp.h"
/* lwIP netif includes */
#include "netif/etharp.h"
/* applications includes */
/*
#include "server.h"
#include "client.h"
*/
void my_server_init(void);
void my_client_init(void);
/* NO_SYS == 1: we need information about the timer intervals: */
#include "lwip/ip_frag.h"
#include "lwip/igmp.h"
/* ethernet layer include */
#include "pktif.h"
/* include the port-dependent configuration */
#include "lwipcfg_msvc.h"
/* port-defined functions used for timer execution, defined in sys_arch.c */
void sys_init_timing();
u32_t sys_now();
/* globales variables for netifs */
/* THE ethernet interface */
struct netif netif;
/* special functions used for NO_SYS==1 only */
typedef struct _timers_infos
{
int timer;
int timer_interval;
void (*timer_func)(void);
} timers_infos;
static timers_infos timers_table[] =
{
{ 0, TCP_FAST_INTERVAL, tcp_fasttmr},
{ 0, TCP_SLOW_INTERVAL, tcp_slowtmr},
{ 0, ARP_TMR_INTERVAL, etharp_tmr},
{ 0, IGMP_TMR_INTERVAL, igmp_tmr},
{ 0, IP_TMR_INTERVAL, ip_reass_tmr},
};
/* get the current time and see if any timer has expired */
static void timers_update()
{
/* static variables for timer execution, initialized to zero! */
static int last_time;
int cur_time, time_diff, idxtimer;
cur_time = sys_now();
time_diff = cur_time - last_time;
/* the '> 0' is an easy wrap-around check: the big gap at
* the wraparound step is simply ignored... */
if (time_diff > 0)
{
last_time = cur_time;
for( idxtimer=0;
idxtimer<(sizeof(timers_table)/sizeof(timers_infos)); idxtimer++)
{
timers_table[idxtimer].timer += time_diff;
if (timers_table[idxtimer].timer >
timers_table[idxtimer].timer_interval)
{
timers_table[idxtimer].timer_func();
timers_table[idxtimer].timer -=
timers_table[idxtimer].timer_interval;
}
}
}
}
void status_callback(struct netif *netif)
{
if (netif_is_up(netif))
{
printf("status_callback==UP, local interface IP is %s\n",
inet_ntoa(*(struct in_addr*)&(netif->ip_addr)));
}
else
{
printf("status_callback==DOWN\n");
}
}
void link_callback(struct netif *netif)
{
if (netif_is_link_up(netif))
{
printf("link_callback==UP\n");
}
else
{
printf("link_callback==DOWN\n");
}
}
/* This function initializes all network interfaces */
static void msvc_netif_init()
{
struct ip_addr ipaddr, netmask, gw;
LWIP_PORT_INIT_GW(&gw);
LWIP_PORT_INIT_IPADDR(&ipaddr);
LWIP_PORT_INIT_NETMASK(&netmask);
printf("Starting lwIP, local interface IP is %s\n",
inet_ntoa(*(struct in_addr*)&ipaddr));
netif_set_default(netif_add(&netif, &ipaddr, &netmask, &gw, NULL,
ethernetif_init, ethernet_input));
netif_set_status_callback(&netif, status_callback);
netif_set_link_callback(&netif, link_callback);
netif_set_up(&netif);
}
void main_loop()
{
/* initialize lwIP stack, network interfaces and application */
sys_init_timing();
lwip_init();
/* init network interfaces */
msvc_netif_init();
/* init application */
//my_server_init(); /* My Server */
my_client_init(); /* My Client */
/* MAIN LOOP for driver update (and timers if NO_SYS) */
while (!_kbhit())
{
/* handle timers (already done in tcpip.c when NO_SYS=0) */
timers_update();
/* check for packets and link status*/
ethernetif_poll(&netif);
/* check for loopback packets on all netifs */
netif_poll_all();
}
/* release the pcap library... */
ethernetif_shutdown(&netif);
}
int main(void)
{
/* no stdio-buffering, please! */
setvbuf(stdout, NULL,_IONBF, 0);
main_loop();
return 0;
}
I hope that someone can help me, I don't know why the communication
doesn't work, I can't understand! Thank you Bye!
Antonio de Angelis
_______________________________________________
lwip-users mailing list
lwip-...@nongnu.org
http://lists.nongnu.org/mailman/listinfo/lwip-users
gold...@gmx.de
> [lengthy post...]
>
First of all, your mail is very difficult to read if you mix such long
code passages with the actual questions you want to ask. Chances you get
a response are higher if the problem is expalined shortly and code is
put at the end of the mail.
As to your problem: I don't really know why this doesn't work, I already
did such a setup on my PC. Did you try your client application to
connect to some other (non lwIP) server? Maybe even outside your PC? If
you still have problems, you can try to use wireshark to monitor the
ethernet packets sent on the network. Maybe that helps to find the problem.
Simon
Antonio de Angelis
I'm sorry, but it's the first time that I write code in a mailing list
:-) I've tried to connect to a server different from the server process
that I've created, a common server on Internet (www.telecom.it) and the
client works! I can't understand why it doesn't work when I try to
connect using the client to my server process!!! I've launched Wireshark
and observed that when I try to connect to my server with my client,
there are only ARP requests of the type "Who has 192.168.1.200? Tell
192.168.1.201" and the reply that specifies the correct MAC address
associated to 192.168.1.200. I can't see the SYN segment sent in
Wireshark, but if I debug the code, after calling tcp_connect, the state
in the pcb structure associated with the connection is SYN SENT...
Please can someone help me? Thank you! :-)
Antonio de Angelis
gold...@gmx.de ha scritto:
Antonio de Angelis
my home network, and it seems to work. But I want to make it work on the
same machine... I attach the wireshark dump of the communication... is
it working in the exact way?
Antonio de Angelis
gold...@gmx.de ha scritto:
Kieran Mansley
> UPDATE: I've tried to put the server process on a different machine on
> my home network, and it seems to work. But I want to make it work on the
> same machine... I attach the wireshark dump of the communication... is
> it working in the exact way?
The problem here is likely to be the routing table. When sending a
packet lwIP searches through the list of netifs and compares the IP
address of the interface with the destination address of the packet. If
none of them match it will use the default, which is usually to send it
out onto the network.
I'm guessing that you have two different processes, and so two separate
instances of lwIP for your client and server. Do they share a driver?
Do they share a network card? Assuming they are structured like this,
it would mean that they would each have a separate routing table and not
know that the other is local. They would therefore send packets that
should go to the other one out on to the network.
You need something, either in the network, or a software bridge, that
will know that they are both on the same machine and route the packets
back again. On Linux it is relatively easy to set up a software bridge
to do this, but I'm not sure how you would do it on Windows.
Kieran
gold...@gmx.de
driver (which uses the winpcap library). So as far as I know, this
should be no problem. I'll try to re-check it it here...
Simon
Bill Auerbach
> will know that they are both on the same machine and route the packets
> back again. On Linux it is relatively easy to set up a software bridge
> to do this, but I'm not sure how you would do it on Windows.
In Settings -> Network Connections select 2 (or more) connections (with the
Ctrl key) and right click the selection and choose Bridge Connections.
Bill
gold...@gmx.de
be that your 2 instances of lwIP are running with the same MAC address?
This is not covered by the netif used with the win32 port: you have to
change the MAC address for pktif.c at line 131 to prevent using the same
MAC address.
Simon
> ------------------------------------------------------------------------
zhou yi
Follow the suggestion, using the wordpad.exe to fix all the dsw and dsp, I can open test.dsw in vc6. I have one ethernet card, so I change lwipcfg_msvc.h as following
#define PACKET_LIB_ADAPTER_NR 1
#define LWIP_PORT_INIT_IPADDR(addr) IP4_ADDR((addr), 192,168,0,200)
#define LWIP_PORT_INIT_GW(addr) IP4_ADDR((addr), 192,168,0,128)
#define LWIP_PORT_INIT_NETMASK(addr) IP4_ADDR((addr), 255,255,255,0)
After compiling the whole project, the output information as following:
Starting lwIP, loopback interface IP is 127.0.0.1
Starting lwIP, local interface IP is 192.168.0.200
0: \Device\NPF_GenericDialupAdapter
1: \Device\NPF_{4988064C-B617-452A-9DEF-4EFC0DDD034A}
USE: Realtek RTL8169/8110 Family Gigabit Ethernet NIC
(Microsoft's Packet Scheduler)
MAC: 00:C0:9F:9D:F7:61
link_callback==UP
status_callback==UP, local interface IP is 192.168.0.200
ping: send 192.168.0.128
ping: recv - 1000 ms - timeout
ping: send 192.168.0.128
sntp_request: not recvfrom==0
ping: recv - 1000 ms - timeout
ping: send 192.168.0.128
ping: recv - 1000 ms - timeout
ping: send 192.168.0.128
3com.com: <not found>
ping: recv - 1000 ms - timeout
ping: send 192.168.0.128
ping: recv 192.168.0.128 1 ms
We can see the ping command sometimes will fail, and I try to connect to the httpd server, open the IE and input http://192.168.0.200, but can not show the right answer.
So my question is:
1. Why ping command sometimes will failed?
2. How to confige the project, so that I can connect the httpd server.
Thanks.
Antonio de Angelis
with the same MAC address because I have only one Ethernet adapter... can
you explain what should I change in pktif.c exactly? I'm a beginner in lwip,
I understand that now the problem is that both instances have the same MAC
address but I don't know how to fix this issue... thank you and thanks to
Kieran and Bill, all of your mails help me very much understanding the
problem!!! :-)
Antonio de Angelis
-----Messaggio originale-----
Da: lwip-users-bounces+finalvengeance=tisca...@nongnu.org
[mailto:lwip-users-bounces+finalvengeance=tisca...@nongnu.org] Per conto
di gold...@gmx.de
Inviato: mercoledì 9 luglio 2008 21.00
A: Mailing list for lwIP users
Oggetto: Re: [lwip-users] How to build a simple TCP [...]
Simon Goldschmidt
This issue is completely handled in pktif.c: the netif's MAC is assigned there by simply adding 1 + lwIP's netif index to that last octet of the mac (which is done in line 131).
If your PC's MAC address is 01:02:03:04:05:06, lwIP will take 01:02:03:04:05:07 for its first pktif driver running on that ethernet card (or :08 if you also use the loopif). However, the other instances of lwIP (running in their own process) do the same, so all instances of lwIP use the same MAC address. What you would have to do is to make the first instance of lwIP use a different MAC than the second. But since this is difficult to do, this has not been done in CVS (yet?). For example, you could take the process ID and add it as an offset...
Simon
-------- Original-Nachricht --------
> Datum: Thu, 10 Jul 2008 09:57:34 +0200
> Von: "Antonio de Angelis" <finalve...@tiscali.it>
> An: "\'Mailing list for lwIP users\'" <lwip-...@nongnu.org>
> Betreff: Re: [lwip-users] How to build a simple TCP [...]
--
Ist Ihr Browser Vista-kompatibel? Jetzt die neuesten
Browser-Versionen downloaden: http://www.gmx.net/de/go/browser
Kieran Mansley
> If your PC's MAC address is 01:02:03:04:05:06, lwIP will take 01:02:03:04:05:07
Yikes! We should document that more obviously, as it's something that
people will likely want to change to be more robust.
Kieran
Simon Goldschmidt
Simon
-------- Original-Nachricht --------
> Datum: Thu, 10 Jul 2008 09:52:14 +0100
> Von: Kieran Mansley <kie...@recoil.org>
> An: Mailing list for lwIP users <lwip-...@nongnu.org>
> Betreff: Re: [lwip-users] How to build a simple TCP [...]
> On Thu, 2008-07-10 at 10:41 +0200, Simon Goldschmidt wrote:
--
GMX startet ShortView.de. Hier findest Du Leute mit Deinen Interessen!
Jetzt dabei sein: http://www.shortview.de/wasistshortview.php?mc=sv_ext_mf@gmx
Kieran Mansley
> Is this something to be afraid of? I didn't think the win32 port would be used in production environments where this matters... But of course we could change this to be defined somwhere instead of calculating it from the real ethernet card's MAC.
I think that would be better. It would make it more obvious that the
MAC address being used is just made up, and so needs to be carefully
chosen.
gold...@gmx.de
port-specific configuration is your problem. Maybe some other
configuration options (look at lwipopts.h) are set so low that you run
out of memory somewhere? Or you have a routing problem? I would take a
look at 'lwip_stats' in the debugger to see if you ran out of memory and
if not take a look at the actual network traffic using wireshark.
Simon.
> ------------------------------------------------------------------------
gold...@gmx.de
> On Thu, 2008-07-10 at 12:49 +0200, Simon Goldschmidt wrote:
>
>> Is this something to be afraid of? I didn't think the win32 port would be used in production environments where this matters... But of course we could change this to be defined somwhere instead of calculating it from the real ethernet card's MAC.
>>
>
> I think that would be better. It would make it more obvious that the
> MAC address being used is just made up, and so needs to be carefully
> chosen.
>
lwipcfg_msvc.h:
#define LWIP_MAC_ADDR_BASE {0x00,0x01,0x02,0x03,0x04,0x05}
is the (bad) default.
Simon
Bob McConnell
> On Thu, 2008-07-10 at 12:49 +0200, Simon Goldschmidt wrote:
>> Is this something to be afraid of? I didn't think the win32 port
>> would be used in production environments where this matters... But
>> of course we could change this to be defined somwhere instead of
>> calculating it from the real ethernet card's MAC.
>
> I think that would be better. It would make it more obvious that the
> MAC address being used is just made up, and so needs to be carefully
> chosen.
The specs also say that there are some bits in the first two bytes of
the MAC that need to be set when it is a virtual NIC instead of a real
card. I have implemented that at work for emulations using tap devices
on FreeBSD, but don't have the reference here at home.
Bob McConnell
N2SPP
David Empson
Bob McConnell wrote:
> Kieran Mansley wrote:
>> Simon Goldschmidt wrote:
>>> Is this something to be afraid of? I didn't think the win32 port
>>> would be used in production environments where this matters... But
>>> of course we could change this to be defined somwhere instead of
>>> calculating it from the real ethernet card's MAC.
>>
>> I think that would be better. It would make it more obvious that the
>> MAC address being used is just made up, and so needs to be carefully
>> chosen.
>
> The specs also say that there are some bits in the first two bytes of
> the MAC that need to be set when it is a virtual NIC instead of a real
> card. I have implemented that at work for emulations using tap devices
> on FreeBSD, but don't have the reference here at home.
The second bit transmitted in the first octet of the MAC address identifies
the address as "locally administered".
This corresponds to a value of 02 in the first octet when it is written in
the usual MSB first order, so an example of a locally administered address
would be 02-01-02-03-04-05.
The first octet can have any even value (bit 0 must not be set, because that
would identify it as a group address, i.e. broadcast/multicast). The other
five octets can be anything.
http://standards.ieee.org/regauth/groupmac/tutorial.html
http://en.wikipedia.org/wiki/MAC_address
A locally administered address cannot be allocated by a manufacturer to any
production Ethernet hardware, but can be used by network administrators to
assign arbitrary MAC addresses for use within their organisation.
Since locally administered MAC addresses are potentially being reused by
multiple people, if LWIP picks an arbitrary one it may have a greater chance
of encountering a conflict. If LWIP "borrows" an arbitrary universally
administered MAC address, there can be only one conflicting device in the
whole world.
I suggest using a locally administered address anyway, but it should be well
commented to explain the potential for a conflict, and summarise the rules
for allocating a different MAC address, in particular not to use an odd
value for the first octet (broadcast/multicast).
zhou yi
#define LWIP_PORT_INIT_GW(addr) IP4_ADDR((addr), 192,168,0,128)
#define LWIP_PORT_INIT_NETMASK(addr) IP4_ADDR((addr), 255,255,255,0)