ip_queue.c

Linux内核源代码 为压缩文件 是<<Linux内核>>一书中的源代码

/*
 * This is a module which is used for queueing IPv4 packets and
 * communicating with userspace via netlink.
 *
 * (C) 2000 James Morris, this code is GPL.
 *
 * 2000-03-27: Simplified code (thanks to Andi Kleen for clues).
 * 2000-05-20: Fixed notifier problems (following Miguel Freitas' report).
 * 2000-06-19: Fixed so nfmark is copied to metadata (reported by Sebastian 
 *             Zander).
 * 2000-08-01: Added Nick Williams' MAC support.
 *
 */
#include <linux/module.h>
#include <linux/skbuff.h>
#include <linux/init.h>
#include <linux/ip.h>
#include <linux/notifier.h>
#include <linux/netdevice.h>
#include <linux/netfilter.h>
#include <linux/netlink.h>
#include <linux/spinlock.h>
#include <linux/rtnetlink.h>
#include <linux/sysctl.h>
#include <linux/proc_fs.h>
#include <net/sock.h>

#include <linux/netfilter_ipv4/ip_queue.h>

#define IPQ_QMAX_DEFAULT 1024
#define IPQ_PROC_FS_NAME "ip_queue"
#define NET_IPQ_QMAX 2088
#define NET_IPQ_QMAX_NAME "ip_queue_maxlen"

typedef struct ipq_queue_element {
	struct list_head list;		/* Links element into queue */
	int verdict;			/* Current verdict */
	struct nf_info *info;		/* Extra info from netfilter */
	struct sk_buff *skb;		/* Packet inside */
} ipq_queue_element_t;

typedef int (*ipq_send_cb_t)(ipq_queue_element_t *e);

typedef struct ipq_peer {
	pid_t pid;			/* PID of userland peer */
	unsigned char died;		/* We think the peer died */
	unsigned char copy_mode;	/* Copy packet as well as metadata? */
	size_t copy_range;		/* Range past metadata to copy */
	ipq_send_cb_t send;		/* Callback for sending data to peer */
} ipq_peer_t;

typedef struct ipq_queue {
 	int len;			/* Current queue len */
 	int *maxlen;			/* Maximum queue len, via sysctl */
 	unsigned char flushing;		/* If queue is being flushed */
 	unsigned char terminate;	/* If the queue is being terminated */
 	struct list_head list;		/* Head of packet queue */
 	spinlock_t lock;		/* Queue spinlock */
 	ipq_peer_t peer;		/* Userland peer */
} ipq_queue_t;

/****************************************************************************
 *
 * Packet queue
 *
 ****************************************************************************/

/* Dequeue a packet if matched by cmp, or the next available if cmp is NULL */
static ipq_queue_element_t *
ipq_dequeue(ipq_queue_t *q,
            int (*cmp)(ipq_queue_element_t *, unsigned long),
            unsigned long data)
{
	struct list_head *i;

	spin_lock_bh(&q->lock);
	for (i = q->list.prev; i != &q->list; i = i->prev) {
		ipq_queue_element_t *e = (ipq_queue_element_t *)i;
		
		if (!cmp || cmp(e, data)) {
			list_del(&e->list);
			q->len--;
			spin_unlock_bh(&q->lock);
			return e;
		}
	}
	spin_unlock_bh(&q->lock);
	return NULL;
}

/* Flush all packets */
static void ipq_flush(ipq_queue_t *q)
{
	ipq_queue_element_t *e;
	
	spin_lock_bh(&q->lock);
	q->flushing = 1;
	spin_unlock_bh(&q->lock);
	while ((e = ipq_dequeue(q, NULL, 0))) {
		e->verdict = NF_DROP;
		nf_reinject(e->skb, e->info, e->verdict);
		kfree(e);
	}
	spin_lock_bh(&q->lock);
	q->flushing = 0;
	spin_unlock_bh(&q->lock);
}

static ipq_queue_t *ipq_create_queue(nf_queue_outfn_t outfn,
                                     ipq_send_cb_t send_cb,
                                     int *errp, int *sysctl_qmax)
{
	int status;
	ipq_queue_t *q;

	*errp = 0;
	q = kmalloc(sizeof(ipq_queue_t), GFP_KERNEL);
	if (q == NULL) {
		*errp = -ENOMEM;
		return NULL;
	}
	q->peer.pid = 0;
	q->peer.died = 0;
	q->peer.copy_mode = IPQ_COPY_NONE;
	q->peer.copy_range = 0;
	q->peer.send = send_cb;
	q->len = 0;
	q->maxlen = sysctl_qmax;
	q->flushing = 0;
	q->terminate = 0;
	INIT_LIST_HEAD(&q->list);
	spin_lock_init(&q->lock);
	status = nf_register_queue_handler(PF_INET, outfn, q);
	if (status < 0) {
		*errp = -EBUSY;
		kfree(q);
		return NULL;
	}
	return q;
}

static int ipq_enqueue(ipq_queue_t *q,
                       struct sk_buff *skb, struct nf_info *info)
{
	ipq_queue_element_t *e;
	int status;
	
	e = kmalloc(sizeof(*e), GFP_ATOMIC);
	if (e == NULL) {
		printk(KERN_ERR "ip_queue: OOM in enqueue\n");
		return -ENOMEM;
	}
	e->verdict = NF_DROP;
	e->info = info;
	e->skb = skb;
	spin_lock_bh(&q->lock);
	if (q->len >= *q->maxlen) {
		spin_unlock_bh(&q->lock);
		if (net_ratelimit()) 
			printk(KERN_WARNING "ip_queue: full at %d entries, "
			       "dropping packet(s).\n", q->len);
		goto free_drop;
	}
	if (q->flushing || q->peer.copy_mode == IPQ_COPY_NONE
	    || q->peer.pid == 0 || q->peer.died || q->terminate) {
		spin_unlock_bh(&q->lock);
		goto free_drop;
	}
	status = q->peer.send(e);
	if (status > 0) {
		list_add(&e->list, &q->list);
		q->len++;
		spin_unlock_bh(&q->lock);
		return status;
	}
	spin_unlock_bh(&q->lock);
	if (status == -ECONNREFUSED) {
		printk(KERN_INFO "ip_queue: peer %d died, "
		       "resetting state and flushing queue\n", q->peer.pid);
			q->peer.died = 1;
			q->peer.pid = 0;
			q->peer.copy_mode = IPQ_COPY_NONE;
			q->peer.copy_range = 0;
			ipq_flush(q);
	}
free_drop:
	kfree(e);
	return -EBUSY;
}

static void ipq_destroy_queue(ipq_queue_t *q)
{
	nf_unregister_queue_handler(PF_INET);
	spin_lock_bh(&q->lock);
	q->terminate = 1;
	spin_unlock_bh(&q->lock);
	ipq_flush(q);
	kfree(q);
}

static int ipq_mangle_ipv4(ipq_verdict_msg_t *v, ipq_queue_element_t *e)
{
	int diff;
	struct iphdr *user_iph = (struct iphdr *)v->payload;

	if (v->data_len < sizeof(*user_iph))
		return 0;
	diff = v->data_len - e->skb->len;
	if (diff < 0)
		skb_trim(e->skb, v->data_len);
	else if (diff > 0) {
		if (v->data_len > 0xFFFF)
			return -EINVAL;
		if (diff > skb_tailroom(e->skb)) {
			struct sk_buff *newskb;
			
			newskb = skb_copy_expand(e->skb,
			                         skb_headroom(e->skb),
			                         diff,
			                         GFP_ATOMIC);
			if (newskb == NULL) {
				printk(KERN_WARNING "ip_queue: OOM "
				      "in mangle, dropping packet\n");
				return -ENOMEM;
			}
			kfree_skb(e->skb);
			e->skb = newskb;
		}
		skb_put(e->skb, diff);
	}
	memcpy(e->skb->data, v->payload, v->data_len);
	e->skb->nfcache |= NFC_ALTERED;
	return 0;
}

static inline int id_cmp(ipq_queue_element_t *e, unsigned long id)
{
	return (id == (unsigned long )e);
}

static int ipq_set_verdict(ipq_queue_t *q,
                           ipq_verdict_msg_t *v, unsigned int len)
{
	ipq_queue_element_t *e;

	if (v->value > NF_MAX_VERDICT)
		return -EINVAL;
	e = ipq_dequeue(q, id_cmp, v->id);
	if (e == NULL)
		return -ENOENT;
	else {
		e->verdict = v->value;
		if (v->data_len && v->data_len == len)
			if (ipq_mangle_ipv4(v, e) < 0)
				e->verdict = NF_DROP;
		nf_reinject(e->skb, e->info, e->verdict);
		kfree(e);
		return 0;
	}
}

static int ipq_receive_peer(ipq_queue_t *q, ipq_peer_msg_t *m,
                            unsigned char type, unsigned int len)
{

	int status = 0;
	int busy;
		
	spin_lock_bh(&q->lock);
	busy = (q->terminate || q->flushing);
	spin_unlock_bh(&q->lock);
	if (busy)
		return -EBUSY;
	if (len < sizeof(ipq_peer_msg_t))
		return -EINVAL;
	switch (type) {
		case IPQM_MODE:
			switch (m->msg.mode.value) {
				case IPQ_COPY_META:
					q->peer.copy_mode = IPQ_COPY_META;
					q->peer.copy_range = 0;
					break;
				case IPQ_COPY_PACKET:
					q->peer.copy_mode = IPQ_COPY_PACKET;
					q->peer.copy_range = m->msg.mode.range;
					if (q->peer.copy_range > 0xFFFF)
						q->peer.copy_range = 0xFFFF;
					break;
				default:
					status = -EINVAL;
			}
			break;
		case IPQM_VERDICT:
			if (m->msg.verdict.value > NF_MAX_VERDICT)
				status = -EINVAL;
			else
				status = ipq_set_verdict(q,
				                         &m->msg.verdict,
				                         len - sizeof(*m));
			break;
		default:
			 status = -EINVAL;
	}
	return status;
}