ip_eth.c

Minix比较全的源码

/*
generic/ip_eth.c

Ethernet specific part of the IP implementation

Created:	Apr 22, 1993 by Philip Homburg

Copyright 1995 Philip Homburg
*/

#include "inet.h"
#include "type.h"
#include "arp.h"
#include "assert.h"
#include "buf.h"
#include "clock.h"
#include "eth.h"
#include "event.h"
#include "icmp_lib.h"
#include "io.h"
#include "ip.h"
#include "ip_int.h"

THIS_FILE

typedef struct xmit_hdr
{
	time_t xh_time;
	ipaddr_t xh_ipaddr;
} xmit_hdr_t;

PRIVATE ether_addr_t broadcast_ethaddr=
{
	{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }
};
PRIVATE ether_addr_t ipmulticast_ethaddr=
{
	{ 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 }
};

FORWARD void do_eth_read ARGS(( ip_port_t *port ));
FORWARD acc_t *get_eth_data ARGS(( int fd, size_t offset,
	size_t count, int for_ioctl ));
FORWARD int put_eth_data ARGS(( int fd, size_t offset,
	acc_t *data, int for_ioctl ));
FORWARD void ipeth_main ARGS(( ip_port_t *port ));
FORWARD void ipeth_set_ipaddr ARGS(( ip_port_t *port ));
FORWARD void ipeth_restart_send ARGS(( ip_port_t *ip_port ));
FORWARD int ipeth_send ARGS(( struct ip_port *ip_port, ipaddr_t dest, 
	acc_t *pack, int type ));
FORWARD void ipeth_arp_reply ARGS(( int ip_port_nr, ipaddr_t ipaddr,
	ether_addr_t *dst_ether_ptr ));
FORWARD int ipeth_update_ttl ARGS(( time_t enq_time, time_t now,
	acc_t *eth_pack ));
FORWARD void ip_eth_arrived ARGS(( int port, acc_t *pack,
	size_t pack_size ));


PUBLIC int ipeth_init(ip_port)
ip_port_t *ip_port;
{
	assert(BUF_S >= sizeof(xmit_hdr_t));
	assert(BUF_S >= sizeof(eth_hdr_t));

	ip_port->ip_dl.dl_eth.de_fd= eth_open(ip_port->
		ip_dl.dl_eth.de_port, ip_port->ip_port,
		get_eth_data, put_eth_data, ip_eth_arrived,
		0 /* no select_res */);
	if (ip_port->ip_dl.dl_eth.de_fd < 0)
	{
		DBLOCK(1, printf("ip.c: unable to open eth port\n"));
		return -1;
	}
	ip_port->ip_dl.dl_eth.de_state= IES_EMPTY;
	ip_port->ip_dl.dl_eth.de_flags= IEF_EMPTY;
	ip_port->ip_dl.dl_eth.de_q_head= NULL;
	ip_port->ip_dl.dl_eth.de_q_tail= NULL;
	ip_port->ip_dl.dl_eth.de_arp_head= NULL;
	ip_port->ip_dl.dl_eth.de_arp_tail= NULL;
	ip_port->ip_dev_main= ipeth_main;
	ip_port->ip_dev_set_ipaddr= ipeth_set_ipaddr;
	ip_port->ip_dev_send= ipeth_send;
	ip_port->ip_mtu= ETH_MAX_PACK_SIZE-ETH_HDR_SIZE;
	ip_port->ip_mtu_max= ip_port->ip_mtu;
	return 0;
}

PRIVATE void ipeth_main(ip_port)
ip_port_t *ip_port;
{
	int result;

	switch (ip_port->ip_dl.dl_eth.de_state)
	{
	case IES_EMPTY:
		ip_port->ip_dl.dl_eth.de_state= IES_SETPROTO;

		result= eth_ioctl(ip_port->ip_dl.dl_eth.de_fd, NWIOSETHOPT);
		if (result == NW_SUSPEND)
			ip_port->ip_dl.dl_eth.de_flags |= IEF_SUSPEND;
		if (result<0)
		{
			DBLOCK(1, printf("eth_ioctl(..,0x%lx)=%d\n",
				(unsigned long)NWIOSETHOPT, result));
			return;
		}
		if (ip_port->ip_dl.dl_eth.de_state != IES_SETPROTO)
			return;
		/* drops through */
	case IES_SETPROTO:
		result= arp_set_cb(ip_port->ip_dl.dl_eth.de_port,
			ip_port->ip_port,
			ipeth_arp_reply);
		if (result != NW_OK)
		{
			printf("ipeth_main: arp_set_cb failed: %d\n",
				result);
			return;
		}

		/* Wait until the interface is configured up. */
		ip_port->ip_dl.dl_eth.de_state= IES_GETIPADDR;
		if (!(ip_port->ip_flags & IPF_IPADDRSET))
		{
			ip_port->ip_dl.dl_eth.de_flags |= IEF_SUSPEND;
			return;
		}

		/* fall through */
	case IES_GETIPADDR:
		ip_port->ip_dl.dl_eth.de_state= IES_MAIN;
		do_eth_read(ip_port);
		return;
	default:
		ip_panic(( "unknown state: %d",
			ip_port->ip_dl.dl_eth.de_state));
	}
}

PRIVATE acc_t *get_eth_data (fd, offset, count, for_ioctl)
int fd;
size_t offset;
size_t count;
int for_ioctl;
{
	ip_port_t *ip_port;
	acc_t *data;
	int result;

	ip_port= &ip_port_table[fd];

	switch (ip_port->ip_dl.dl_eth.de_state)
	{
	case IES_SETPROTO:
		if (!count)
		{
			result= (int)offset;
			if (result<0)
			{
				ip_port->ip_dl.dl_eth.de_state= IES_ERROR;
				break;
			}
			if (ip_port->ip_dl.dl_eth.de_flags & IEF_SUSPEND)
				ipeth_main(ip_port);
			return NW_OK;
		}
		assert ((!offset) && (count == sizeof(struct nwio_ethopt)));
		{
			struct nwio_ethopt *ethopt;
			acc_t *acc;

			acc= bf_memreq(sizeof(*ethopt));
			ethopt= (struct nwio_ethopt *)ptr2acc_data(acc);
			ethopt->nweo_flags= NWEO_COPY|NWEO_EN_BROAD|
				NWEO_EN_MULTI|NWEO_TYPESPEC;
			ethopt->nweo_type= HTONS(ETH_IP_PROTO);
			return acc;
		}

	case IES_MAIN:
		if (!count)
		{
			result= (int)offset;
			if (result<0)
				ip_warning(( "error on write: %d\n", result ));
			bf_afree (ip_port->ip_dl.dl_eth.de_frame);
			ip_port->ip_dl.dl_eth.de_frame= 0;

			if (ip_port->ip_dl.dl_eth.de_flags & IEF_WRITE_SP)
			{
				ip_port->ip_dl.dl_eth.de_flags &=
					~IEF_WRITE_SP;
				ipeth_restart_send(ip_port);
			}
			return NW_OK;
		}
		data= bf_cut (ip_port->ip_dl.dl_eth.de_frame, offset, count);
		assert (data);
		return data;
	default:
		printf(
		"get_eth_data(%d, 0x%d, 0x%d) called but ip_state=0x%x\n",
			fd, offset, count, ip_port->ip_dl.dl_eth.de_state);
		break;
	}
	return 0;
}

PRIVATE int put_eth_data (port, offset, data, for_ioctl)
int port;
size_t offset;
acc_t *data;
int for_ioctl;
{
	ip_port_t *ip_port;
	int result;

	ip_port= &ip_port_table[port];

	assert(0);

	if (ip_port->ip_dl.dl_eth.de_flags & IEF_READ_IP)
	{
		if (!data)
		{
			result= (int)offset;
			if (result<0)
			{
				DBLOCK(1, printf(
				"ip.c: put_eth_data(..,%d,..)\n", result));
				return NW_OK;
			}
			if (ip_port->ip_dl.dl_eth.de_flags & IEF_READ_SP)
			{
				ip_port->ip_dl.dl_eth.de_flags &= 
						~(IEF_READ_IP|IEF_READ_SP);
				do_eth_read(ip_port);
			}
			else
				ip_port->ip_dl.dl_eth.de_flags &= ~IEF_READ_IP;
			return NW_OK;
		}
		assert (!offset);
		/* Warning: the above assertion is illegal; puts and
		   gets of data can be brokenup in any piece the server
		   likes. However we assume that the server is eth.c
		   and it transfers only whole packets. */
		ip_eth_arrived(port, data, bf_bufsize(data));
		return NW_OK;
	}
	printf("ip_port->ip_dl.dl_eth.de_state= 0x%x",
		ip_port->ip_dl.dl_eth.de_state);
	ip_panic (( "strange status" ));
}

PRIVATE void ipeth_set_ipaddr(ip_port)
ip_port_t *ip_port;
{
	arp_set_ipaddr (ip_port->ip_dl.dl_eth.de_port, ip_port->ip_ipaddr);
	if (ip_port->ip_dl.dl_eth.de_state == IES_GETIPADDR)
		ipeth_main(ip_port);
}

PRIVATE int ipeth_send(ip_port, dest, pack, type)
struct ip_port *ip_port;
ipaddr_t dest;
acc_t *pack;
int type;
{
	int i, r;
	acc_t *eth_pack, *tail;
	size_t pack_size;
	eth_hdr_t *eth_hdr;
	xmit_hdr_t *xmit_hdr;
	ipaddr_t hostpart, tmpaddr;
	time_t t;
	u32_t *p;

	/* Start optimistic: the arp will succeed without blocking and the
	 * ethernet packet can be sent without blocking also. Start with
	 * the allocation of the ethernet header.
	 */
	eth_pack= bf_memreq(sizeof(*eth_hdr));
	assert(eth_pack->acc_next == NULL);
	eth_pack->acc_next= pack;
	pack_size= bf_bufsize(eth_pack);
	if (pack_size<ETH_MIN_PACK_SIZE)
	{
		tail= bf_memreq(ETH_MIN_PACK_SIZE-pack_size);

		/* Clear padding */
		for (i= (ETH_MIN_PACK_SIZE-pack_size)/sizeof(*p),
			p= (u32_t *)ptr2acc_data(tail);
			i >= 0; i--, p++)
		{
			*p= 0xdeadbeef;
		}

		eth_pack= bf_append(eth_pack, tail);
	}
	eth_hdr= (eth_hdr_t *)ptr2acc_data(eth_pack);

	/* Lookup the ethernet address */
	if (type != IP_LT_NORMAL)
	{
		if (type == IP_LT_BROADCAST)
			eth_hdr->eh_dst= broadcast_ethaddr;
		else
		{
			tmpaddr= ntohl(dest);
			eth_hdr->eh_dst= ipmulticast_ethaddr;
			eth_hdr->eh_dst.ea_addr[5]= tmpaddr & 0xff;
			eth_hdr->eh_dst.ea_addr[4]= (tmpaddr >> 8) & 0xff;
			eth_hdr->eh_dst.ea_addr[3]= (tmpaddr >> 16) & 0x7f;
		}
	}
	else
	{
		if ((dest ^ ip_port->ip_ipaddr) & ip_port->ip_subnetmask)
		{
			ip_panic(( "invalid destination" ));
		}

		hostpart= (dest & ~ip_port->ip_subnetmask);
		assert(dest != ip_port->ip_ipaddr);

		r= arp_ip_eth(ip_port->ip_dl.dl_eth.de_port,
			dest, &eth_hdr->eh_dst);
		if (r == NW_SUSPEND)
		{
			/* Unfortunately, the arp takes some time, use
			 * the ethernet header to store the next hop
			 * ip address and the current time.
			 */
			xmit_hdr= (xmit_hdr_t *)eth_hdr;
			xmit_hdr->xh_time= get_time();
			xmit_hdr->xh_ipaddr= dest;
			eth_pack->acc_ext_link= NULL;
			if (ip_port->ip_dl.dl_eth.de_arp_head == NULL)
				ip_port->ip_dl.dl_eth.de_arp_head= eth_pack;
			else
			{
				ip_port->ip_dl.dl_eth.de_arp_tail->
					acc_ext_link= eth_pack;
			}
			ip_port->ip_dl.dl_eth.de_arp_tail= eth_pack;
			return NW_OK;
		}
		if (r == EDSTNOTRCH)
		{
			bf_afree(eth_pack);
			return EDSTNOTRCH;
		}
		assert(r == NW_OK);
	}

	/* If we have no write in progress, we can try to send the ethernet
	 * packet using eth_send. If the IP packet is larger than mtu,
	 * enqueue the packet and let ipeth_restart_send deal with it. 
	 */
	pack_size= bf_bufsize(eth_pack);
	if (ip_port->ip_dl.dl_eth.de_frame == NULL && pack_size <=