problem with htons()
 

i compile using
# gcc -c testing.c
its ok
when i isseu the command
# insmod testing.o
the error appears "unresolved symbol htons"
***********************************
/*************C O D E **************/

#define ETHERNET_DEVICE_NAME "eth0"
#define EAPOL_ETHERNET_TYPE 0x888E

struct net_device* g_poEthernetDevice;
struct packet_type g_oProtocolHandler;

int init_module (void)
{
int nReturnValue = 0;

if((g_poEthernetDevice = dev_get_by_name(ETHERNET_DEVICE_NAME)))
{
g_poEthernetDevice->flags |= IFF_PROMISC;
g_oProtocolHandler.type = htons(EAPOL_ETHERNET_TYPE); // <----- Problem ------
g_oProtocolHandler.dev = g_poEthernetDevice;
g_oProtocolHandler.func = My_EthernetRxPacket;
dev_add_pack(&g_oProtocolHandler);
printk("Ethernet Device allocated\n");
}
else
printk("Ethernet Device allocation ERROR\n");
return(nReturnValue);
}

void cleanup_module (void)
{
dev_remove_pack(&g_oProtocolHandler);
printk("Unregister Success \n");
}

int My_EthernetRxPacket(struct sk_buff* r_poSocketBuffer,struct net_device* r_poNetworkDevice,struct packet_type* r_poPacketType)
{
//here i receive the EAPOL packet
return(0);
}
/****************** E N D *******************/

You need to include a file where the htons macro is defined, netinet/in.h for example.

including that file results in a huge amount of errors
actually there is no error till
gcc -c testing.c
with out including that file

Yeah, what jlliagre said. I always try to compile with the -Wall option. Then gcc would throw up a warning about it being implicitly defined.

Your code sample is incomplete, you do not show the include directives you use.

Jlliagre is correct:

1. htons() is a macro, not a function.

The fact that you get the runtime error "unresolved symbol htons" means that your compiler incorrectly treated it like a function. Which in turn means that your compile (despite the lack of warnings) is broken.

2. The solution to problem 1) is to include the correct header (e.g. "#include "netinet/ih.h>").

3. That fact that you're getting lots of errors when you try to do so means that something ELSE is terribly wrong.

In all likelihood, you're probably missing *OTHER* headers as well.

4. itsme86, too, is absolutely correct: "-Wall" is your friend.

The more warnings you get, the more potential problems you're discovering at precisely the time when it's easiest to fix them.

Good luck .. PSM

It looks like the header you want is linux/byteorder/generic.h

I don't think netinet/in.h exists in the kernel tree so yeah, including it would be a really bad idea.

Since you're building a kernel module you probably need to either build using the kernel's build system (2.6) or #define some things, I think __KERNEL__ and __MODULE__ for 2.4.

You'd #include netinet/in.h if you're compiling a normal sockets program.

On the other hand, you'd #include linux/byteorder/generic.h if you're writing a kernel module (as you appear to be doing).

In either case, the basic problem is that "htons()" is a macro (not a function), and you're failing to #include the right header at compile time.

One other tip:
you can use the special bulletin board tags "" to delimit your source code on linuxquestions.org - that way, you won't lose the formatting.

'Hope that helps .. PSM

Whoops:

That was the vB code stuff in your "Post Reply" page: " [code]" " [\code]"

i have included linux\byteorder\generic.h
now the error is not appearing,
another problem faced by me. when i use i.e a= htons(0x888e) the value returned by the htons() is suppose 32764 ,now removing htons()
i.e a= 0x888e
the output remains 32764

on what architecure are you compiling, intel or sth else ?

If the endianness of the computer your compiling on is the same as the network byte order than htons() doesn't do anything.

Network byte order is simply a means to insure compatibility between computers with different endianness on a network.

Unfortunately, the answer he's getting (32764) is wrong for either arch:

Not sure what's up with that...