af_decnet.c

linux 内核源代码

/*
 * DECnet       An implementation of the DECnet protocol suite for the LINUX
 *              operating system.  DECnet is implemented using the  BSD Socket
 *              interface as the means of communication with the user level.
 *
 *              DECnet Socket Layer Interface
 *
 * Authors:     Eduardo Marcelo Serrat <[e-mail]>
 *              Patrick Caulfield <[e-mail]>
 *
 * Changes:
 *        Steve Whitehouse: Copied from Eduardo Serrat and Patrick Caulfield's
 *                          version of the code. Original copyright preserved
 *                          below.
 *        Steve Whitehouse: Some bug fixes, cleaning up some code to make it
 *                          compatible with my routing layer.
 *        Steve Whitehouse: Merging changes from Eduardo Serrat and Patrick
 *                          Caulfield.
 *        Steve Whitehouse: Further bug fixes, checking module code still works
 *                          with new routing layer.
 *        Steve Whitehouse: Additional set/get_sockopt() calls.
 *        Steve Whitehouse: Fixed TIOCINQ ioctl to be same as Eduardo's new
 *                          code.
 *        Steve Whitehouse: recvmsg() changed to try and behave in a POSIX like
 *                          way. Didn't manage it entirely, but its better.
 *        Steve Whitehouse: ditto for sendmsg().
 *        Steve Whitehouse: A selection of bug fixes to various things.
 *        Steve Whitehouse: Added TIOCOUTQ ioctl.
 *        Steve Whitehouse: Fixes to username2sockaddr & sockaddr2username.
 *        Steve Whitehouse: Fixes to connect() error returns.
 *       Patrick Caulfield: Fixes to delayed acceptance logic.
 *         David S. Miller: New socket locking
 *        Steve Whitehouse: Socket list hashing/locking
 *         Arnaldo C. Melo: use capable, not suser
 *        Steve Whitehouse: Removed unused code. Fix to use sk->allocation
 *                          when required.
 *       Patrick Caulfield: /proc/net/decnet now has object name/number
 *        Steve Whitehouse: Fixed local port allocation, hashed sk list
 *          Matthew Wilcox: Fixes for dn_ioctl()
 *        Steve Whitehouse: New connect/accept logic to allow timeouts and
 *                          prepare for sendpage etc.
 */


/******************************************************************************
    (c) 1995-1998 E.M. Serrat		[e-mail]

    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation; either version 2 of the License, or
    any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

HISTORY:

Version           Kernel     Date       Author/Comments
-------           ------     ----       ---------------
Version 0.0.1     2.0.30    01-dic-97	Eduardo Marcelo Serrat
					([e-mail])

					First Development of DECnet Socket La-
					yer for Linux. Only supports outgoing
					connections.

Version 0.0.2	  2.1.105   20-jun-98   Patrick J. Caulfield
					([e-mail])

					Port to new kernel development version.

Version 0.0.3     2.1.106   25-jun-98   Eduardo Marcelo Serrat
					([e-mail])
					_
					Added support for incoming connections
					so we can start developing server apps
					on Linux.
					-
					Module Support
Version 0.0.4     2.1.109   21-jul-98   Eduardo Marcelo Serrat
				       ([e-mail])
				       _
					Added support for X11R6.4. Now we can
					use DECnet transport for X on Linux!!!
				       -
Version 0.0.5    2.1.110   01-aug-98   Eduardo Marcelo Serrat
				       ([e-mail])
				       Removed bugs on flow control
				       Removed bugs on incoming accessdata
				       order
				       -
Version 0.0.6    2.1.110   07-aug-98   Eduardo Marcelo Serrat
				       dn_recvmsg fixes

					Patrick J. Caulfield
				       dn_bind fixes
*******************************************************************************/

#include <linux/module.h>
#include <linux/errno.h>
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/socket.h>
#include <linux/in.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/timer.h>
#include <linux/string.h>
#include <linux/sockios.h>
#include <linux/net.h>
#include <linux/netdevice.h>
#include <linux/inet.h>
#include <linux/route.h>
#include <linux/netfilter.h>
#include <linux/seq_file.h>
#include <net/sock.h>
#include <net/tcp_states.h>
#include <net/flow.h>
#include <asm/system.h>
#include <asm/ioctls.h>
#include <linux/capability.h>
#include <linux/mm.h>
#include <linux/interrupt.h>
#include <linux/proc_fs.h>
#include <linux/stat.h>
#include <linux/init.h>
#include <linux/poll.h>
#include <net/net_namespace.h>
#include <net/neighbour.h>
#include <net/dst.h>
#include <net/fib_rules.h>
#include <net/dn.h>
#include <net/dn_nsp.h>
#include <net/dn_dev.h>
#include <net/dn_route.h>
#include <net/dn_fib.h>
#include <net/dn_neigh.h>

struct dn_sock {
	struct sock sk;
	struct dn_scp scp;
};

static void dn_keepalive(struct sock *sk);

#define DN_SK_HASH_SHIFT 8
#define DN_SK_HASH_SIZE (1 << DN_SK_HASH_SHIFT)
#define DN_SK_HASH_MASK (DN_SK_HASH_SIZE - 1)


static const struct proto_ops dn_proto_ops;
static DEFINE_RWLOCK(dn_hash_lock);
static struct hlist_head dn_sk_hash[DN_SK_HASH_SIZE];
static struct hlist_head dn_wild_sk;
static atomic_t decnet_memory_allocated;

static int __dn_setsockopt(struct socket *sock, int level, int optname, char __user *optval, int optlen, int flags);
static int __dn_getsockopt(struct socket *sock, int level, int optname, char __user *optval, int __user *optlen, int flags);

static struct hlist_head *dn_find_list(struct sock *sk)
{
	struct dn_scp *scp = DN_SK(sk);

	if (scp->addr.sdn_flags & SDF_WILD)
		return hlist_empty(&dn_wild_sk) ? &dn_wild_sk : NULL;

	return &dn_sk_hash[dn_ntohs(scp->addrloc) & DN_SK_HASH_MASK];
}

/*
 * Valid ports are those greater than zero and not already in use.
 */
static int check_port(__le16 port)
{
	struct sock *sk;
	struct hlist_node *node;

	if (port == 0)
		return -1;

	sk_for_each(sk, node, &dn_sk_hash[dn_ntohs(port) & DN_SK_HASH_MASK]) {
		struct dn_scp *scp = DN_SK(sk);
		if (scp->addrloc == port)
			return -1;
	}
	return 0;
}

static unsigned short port_alloc(struct sock *sk)
{
	struct dn_scp *scp = DN_SK(sk);
static unsigned short port = 0x2000;
	unsigned short i_port = port;

	while(check_port(dn_htons(++port)) != 0) {
		if (port == i_port)
			return 0;
	}

	scp->addrloc = dn_htons(port);

	return 1;
}

/*
 * Since this is only ever called from user
 * level, we don't need a write_lock() version
 * of this.
 */
static int dn_hash_sock(struct sock *sk)
{
	struct dn_scp *scp = DN_SK(sk);
	struct hlist_head *list;
	int rv = -EUSERS;

	BUG_ON(sk_hashed(sk));

	write_lock_bh(&dn_hash_lock);

	if (!scp->addrloc && !port_alloc(sk))
		goto out;

	rv = -EADDRINUSE;
	if ((list = dn_find_list(sk)) == NULL)
		goto out;

	sk_add_node(sk, list);
	rv = 0;
out:
	write_unlock_bh(&dn_hash_lock);
	return rv;
}

static void dn_unhash_sock(struct sock *sk)
{
	write_lock(&dn_hash_lock);
	sk_del_node_init(sk);
	write_unlock(&dn_hash_lock);
}

static void dn_unhash_sock_bh(struct sock *sk)
{
	write_lock_bh(&dn_hash_lock);
	sk_del_node_init(sk);
	write_unlock_bh(&dn_hash_lock);
}

static struct hlist_head *listen_hash(struct sockaddr_dn *addr)
{
	int i;
	unsigned hash = addr->sdn_objnum;

	if (hash == 0) {
		hash = addr->sdn_objnamel;
		for(i = 0; i < dn_ntohs(addr->sdn_objnamel); i++) {
			hash ^= addr->sdn_objname[i];
			hash ^= (hash << 3);
		}
	}

	return &dn_sk_hash[hash & DN_SK_HASH_MASK];
}

/*
 * Called to transform a socket from bound (i.e. with a local address)
 * into a listening socket (doesn't need a local port number) and rehashes
 * based upon the object name/number.
 */
static void dn_rehash_sock(struct sock *sk)
{
	struct hlist_head *list;
	struct dn_scp *scp = DN_SK(sk);

	if (scp->addr.sdn_flags & SDF_WILD)
		return;

	write_lock_bh(&dn_hash_lock);
	sk_del_node_init(sk);
	DN_SK(sk)->addrloc = 0;
	list = listen_hash(&DN_SK(sk)->addr);
	sk_add_node(sk, list);
	write_unlock_bh(&dn_hash_lock);
}

int dn_sockaddr2username(struct sockaddr_dn *sdn, unsigned char *buf, unsigned char type)
{
	int len = 2;

	*buf++ = type;

	switch(type) {
		case 0:
			*buf++ = sdn->sdn_objnum;
			break;
		case 1:
			*buf++ = 0;
			*buf++ = dn_ntohs(sdn->sdn_objnamel);
			memcpy(buf, sdn->sdn_objname, dn_ntohs(sdn->sdn_objnamel));
			len = 3 + dn_ntohs(sdn->sdn_objnamel);
			break;
		case 2:
			memset(buf, 0, 5);
			buf += 5;
			*buf++ = dn_ntohs(sdn->sdn_objnamel);
			memcpy(buf, sdn->sdn_objname, dn_ntohs(sdn->sdn_objnamel));
			len = 7 + dn_ntohs(sdn->sdn_objnamel);
			break;
	}

	return len;
}

/*
 * On reception of usernames, we handle types 1 and 0 for destination
 * addresses only. Types 2 and 4 are used for source addresses, but the
 * UIC, GIC are ignored and they are both treated the same way. Type 3
 * is never used as I've no idea what its purpose might be or what its
 * format is.
 */
int dn_username2sockaddr(unsigned char *data, int len, struct sockaddr_dn *sdn, unsigned char *fmt)
{
	unsigned char type;
	int size = len;
	int namel = 12;

	sdn->sdn_objnum = 0;
	sdn->sdn_objnamel = dn_htons(0);
	memset(sdn->sdn_objname, 0, DN_MAXOBJL);

	if (len < 2)
		return -1;

	len -= 2;
	*fmt = *data++;
	type = *data++;

	switch(*fmt) {
		case 0:
			sdn->sdn_objnum = type;
			return 2;
		case 1:
			namel = 16;
			break;
		case 2:
			len  -= 4;
			data += 4;
			break;
		case 4:
			len  -= 8;
			data += 8;
			break;
		default:
			return -1;
	}

	len -= 1;

	if (len < 0)
		return -1;

	sdn->sdn_objnamel = dn_htons(*data++);
	len -= dn_ntohs(sdn->sdn_objnamel);

	if ((len < 0) || (dn_ntohs(sdn->sdn_objnamel) > namel))
		return -1;

	memcpy(sdn->sdn_objname, data, dn_ntohs(sdn->sdn_objnamel));

	return size - len;
}

struct sock *dn_sklist_find_listener(struct sockaddr_dn *addr)
{
	struct hlist_head *list = listen_hash(addr);
	struct hlist_node *node;
	struct sock *sk;

	read_lock(&dn_hash_lock);
	sk_for_each(sk, node, list) {
		struct dn_scp *scp = DN_SK(sk);
		if (sk->sk_state != TCP_LISTEN)
			continue;
		if (scp->addr.sdn_objnum) {
			if (scp->addr.sdn_objnum != addr->sdn_objnum)
				continue;
		} else {
			if (addr->sdn_objnum)
				continue;
			if (scp->addr.sdn_objnamel != addr->sdn_objnamel)
				continue;
			if (memcmp(scp->addr.sdn_objname, addr->sdn_objname, dn_ntohs(addr->sdn_objnamel)) != 0)
				continue;
		}
		sock_hold(sk);
		read_unlock(&dn_hash_lock);
		return sk;
	}

	sk = sk_head(&dn_wild_sk);
	if (sk) {
		if (sk->sk_state == TCP_LISTEN)
			sock_hold(sk);
		else
			sk = NULL;
	}

	read_unlock(&dn_hash_lock);
	return sk;
}

struct sock *dn_find_by_skb(struct sk_buff *skb)
{
	struct dn_skb_cb *cb = DN_SKB_CB(skb);
	struct sock *sk;
	struct hlist_node *node;
	struct dn_scp *scp;

	read_lock(&dn_hash_lock);
	sk_for_each(sk, node, &dn_sk_hash[dn_ntohs(cb->dst_port) & DN_SK_HASH_MASK]) {
		scp = DN_SK(sk);
		if (cb->src != dn_saddr2dn(&scp->peer))
			continue;
		if (cb->dst_port != scp->addrloc)
			continue;
		if (scp->addrrem && (cb->src_port != scp->addrrem))
			continue;
		sock_hold(sk);
		goto found;
	}
	sk = NULL;
found:
	read_unlock(&dn_hash_lock);
	return sk;
}



static void dn_destruct(struct sock *sk)
{
	struct dn_scp *scp = DN_SK(sk);

	skb_queue_purge(&scp->data_xmit_queue);
	skb_queue_purge(&scp->other_xmit_queue);
	skb_queue_purge(&scp->other_receive_queue);

	dst_release(xchg(&sk->sk_dst_cache, NULL));
}

static int dn_memory_pressure;

static void dn_enter_memory_pressure(void)
{
	if (!dn_memory_pressure) {
		dn_memory_pressure = 1;
	}
}

static struct proto dn_proto = {
	.name			= "NSP",
	.owner			= THIS_MODULE,
	.enter_memory_pressure	= dn_enter_memory_pressure,
	.memory_pressure	= &dn_memory_pressure,
	.memory_allocated	= &decnet_memory_allocated,
	.sysctl_mem		= sysctl_decnet_mem,
	.sysctl_wmem		= sysctl_decnet_wmem,
	.sysctl_rmem		= sysctl_decnet_rmem,
	.max_header		= DN_MAX_NSP_DATA_HEADER + 64,
	.obj_size		= sizeof(struct dn_sock),
};

static struct sock *dn_alloc_sock(struct net *net, struct socket *sock, gfp_t gfp)
{
	struct dn_scp *scp;
	struct sock *sk = sk_alloc(net, PF_DECnet, gfp, &dn_proto);

	if  (!sk)
		goto out;

	if (sock)
		sock->ops = &dn_proto_ops;
	sock_init_data(sock, sk);

	sk->sk_backlog_rcv = dn_nsp_backlog_rcv;
	sk->sk_destruct    = dn_destruct;
	sk->sk_no_check    = 1;
	sk->sk_family      = PF_DECnet;
	sk->sk_protocol    = 0;
	sk->sk_allocation  = gfp;
	sk->sk_sndbuf	   = sysctl_decnet_wmem[1];
	sk->sk_rcvbuf	   = sysctl_decnet_rmem[1];

	/* Initialization of DECnet Session Control Port		*/
	scp = DN_SK(sk);
	scp->state	= DN_O;		/* Open			*/
	scp->numdat	= 1;		/* Next data seg to tx	*/
	scp->numoth	= 1;		/* Next oth data to tx  */
	scp->ackxmt_dat = 0;		/* Last data seg ack'ed */
	scp->ackxmt_oth = 0;		/* Last oth data ack'ed */
	scp->ackrcv_dat = 0;		/* Highest data ack recv*/
	scp->ackrcv_oth = 0;		/* Last oth data ack rec*/
	scp->flowrem_sw = DN_SEND;
	scp->flowloc_sw = DN_SEND;
	scp->flowrem_dat = 0;
	scp->flowrem_oth = 1;
	scp->flowloc_dat = 0;
	scp->flowloc_oth = 1;
	scp->services_rem = 0;
	scp->services_loc = 1 | NSP_FC_NONE;
	scp->info_rem = 0;
	scp->info_loc = 0x03; /* NSP version 4.1 */
	scp->segsize_rem = 230 - DN_MAX_NSP_DATA_HEADER; /* Default: Updated by remote segsize */
	scp->nonagle = 0;
	scp->multi_ireq = 1;
	scp->accept_mode = ACC_IMMED;
	scp->addr.sdn_family    = AF_DECnet;
	scp->peer.sdn_family    = AF_DECnet;
	scp->accessdata.acc_accl = 5;
	memcpy(scp->accessdata.acc_acc, "LINUX", 5);

	scp->max_window   = NSP_MAX_WINDOW;
	scp->snd_window   = NSP_MIN_WINDOW;
	scp->nsp_srtt     = NSP_INITIAL_SRTT;
	scp->nsp_rttvar   = NSP_INITIAL_RTTVAR;
	scp->nsp_rxtshift = 0;

	skb_queue_head_init(&scp->data_xmit_queue);
	skb_queue_head_init(&scp->other_xmit_queue);
	skb_queue_head_init(&scp->other_receive_queue);

	scp->persist = 0;
	scp->persist_fxn = NULL;
	scp->keepalive = 10 * HZ;
	scp->keepalive_fxn = dn_keepalive;

	init_timer(&scp->delack_timer);
	scp->delack_pending = 0;
	scp->delack_fxn = dn_nsp_delayed_ack;

	dn_start_slow_timer(sk);
out:
	return sk;
}

/*
 * Keepalive timer.
 * FIXME: Should respond to SO_KEEPALIVE etc.
 */
static void dn_keepalive(struct sock *sk)
{
	struct dn_scp *scp = DN_SK(sk);

	/*
	 * By checking the other_data transmit queue is empty
	 * we are double checking that we are not sending too
	 * many of these keepalive frames.
	 */
	if (skb_queue_empty(&scp->other_xmit_queue))
		dn_nsp_send_link(sk, DN_NOCHANGE, 0);
}


/*
 * Timer for shutdown/destroyed sockets.
 * When socket is dead & no packets have been sent for a
 * certain amount of time, they are removed by this
 * routine. Also takes care of sending out DI & DC
 * frames at correct times.
 */
int dn_destroy_timer(struct sock *sk)
{
	struct dn_scp *scp = DN_SK(sk);

	scp->persist = dn_nsp_persist(sk);

	switch(scp->state) {
		case DN_DI:
			dn_nsp_send_disc(sk, NSP_DISCINIT, 0, GFP_ATOMIC);
			if (scp->nsp_rxtshift >= decnet_di_count)
				scp->state = DN_CN;
			return 0;

		case DN_DR:
			dn_nsp_send_disc(sk, NSP_DISCINIT, 0, GFP_ATOMIC);
			if (scp->nsp_rxtshift >= decnet_dr_count)
				scp->state = DN_DRC;
			return 0;

		case DN_DN:
			if (scp->nsp_rxtshift < decnet_dn_count) {
				/* printk(KERN_DEBUG "dn_destroy_timer: DN\n"); */
				dn_nsp_send_disc(sk, NSP_DISCCONF, NSP_REASON_DC, GFP_ATOMIC);
				return 0;
			}
	}

	scp->persist = (HZ * decnet_time_wait);

	if (sk->sk_socket)
		return 0;

	if ((jiffies - scp->stamp) >= (HZ * decnet_time_wait)) {
		dn_unhash_sock(sk);
		sock_put(sk);
		return 1;
	}

	return 0;
}