nf_conntrack_reasm.c

linux-2.6.15.6

/*
 * IPv6 fragment reassembly for connection tracking
 *
 * Copyright (C)2004 USAGI/WIDE Project
 *
 * Author:
 *	Yasuyuki Kozakai @USAGI <[e-mail]>
 *
 * Based on: net/ipv6/reassembly.c
 *
 * 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 (at your option) any later version.
 */

#include <linux/config.h>
#include <linux/errno.h>
#include <linux/types.h>
#include <linux/string.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/jiffies.h>
#include <linux/net.h>
#include <linux/list.h>
#include <linux/netdevice.h>
#include <linux/in6.h>
#include <linux/ipv6.h>
#include <linux/icmpv6.h>
#include <linux/random.h>
#include <linux/jhash.h>

#include <net/sock.h>
#include <net/snmp.h>

#include <net/ipv6.h>
#include <net/protocol.h>
#include <net/transp_v6.h>
#include <net/rawv6.h>
#include <net/ndisc.h>
#include <net/addrconf.h>
#include <linux/sysctl.h>
#include <linux/netfilter.h>
#include <linux/netfilter_ipv6.h>
#include <linux/kernel.h>
#include <linux/module.h>

#if 0
#define DEBUGP printk
#else
#define DEBUGP(format, args...)
#endif

#define NF_CT_FRAG6_HIGH_THRESH 262144 /* == 256*1024 */
#define NF_CT_FRAG6_LOW_THRESH 196608  /* == 192*1024 */
#define NF_CT_FRAG6_TIMEOUT IPV6_FRAG_TIMEOUT

unsigned int nf_ct_frag6_high_thresh = 256*1024;
unsigned int nf_ct_frag6_low_thresh = 192*1024;
unsigned long nf_ct_frag6_timeout = IPV6_FRAG_TIMEOUT;

struct nf_ct_frag6_skb_cb
{
	struct inet6_skb_parm	h;
	int			offset;
	struct sk_buff		*orig;
};

#define NFCT_FRAG6_CB(skb)	((struct nf_ct_frag6_skb_cb*)((skb)->cb))

struct nf_ct_frag6_queue
{
	struct nf_ct_frag6_queue	*next;
	struct list_head lru_list;		/* lru list member	*/

	__u32			id;		/* fragment id		*/
	struct in6_addr		saddr;
	struct in6_addr		daddr;

	spinlock_t		lock;
	atomic_t		refcnt;
	struct timer_list	timer;		/* expire timer		*/
	struct sk_buff		*fragments;
	int			len;
	int			meat;
	struct timeval		stamp;
	unsigned int		csum;
	__u8			last_in;	/* has first/last segment arrived? */
#define COMPLETE		4
#define FIRST_IN		2
#define LAST_IN			1
	__u16			nhoffset;
	struct nf_ct_frag6_queue	**pprev;
};

/* Hash table. */

#define FRAG6Q_HASHSZ	64

static struct nf_ct_frag6_queue *nf_ct_frag6_hash[FRAG6Q_HASHSZ];
static rwlock_t nf_ct_frag6_lock = RW_LOCK_UNLOCKED;
static u32 nf_ct_frag6_hash_rnd;
static LIST_HEAD(nf_ct_frag6_lru_list);
int nf_ct_frag6_nqueues = 0;

static __inline__ void __fq_unlink(struct nf_ct_frag6_queue *fq)
{
	if (fq->next)
		fq->next->pprev = fq->pprev;
	*fq->pprev = fq->next;
	list_del(&fq->lru_list);
	nf_ct_frag6_nqueues--;
}

static __inline__ void fq_unlink(struct nf_ct_frag6_queue *fq)
{
	write_lock(&nf_ct_frag6_lock);
	__fq_unlink(fq);
	write_unlock(&nf_ct_frag6_lock);
}

static unsigned int ip6qhashfn(u32 id, struct in6_addr *saddr,
			       struct in6_addr *daddr)
{
	u32 a, b, c;

	a = saddr->s6_addr32[0];
	b = saddr->s6_addr32[1];
	c = saddr->s6_addr32[2];

	a += JHASH_GOLDEN_RATIO;
	b += JHASH_GOLDEN_RATIO;
	c += nf_ct_frag6_hash_rnd;
	__jhash_mix(a, b, c);

	a += saddr->s6_addr32[3];
	b += daddr->s6_addr32[0];
	c += daddr->s6_addr32[1];
	__jhash_mix(a, b, c);

	a += daddr->s6_addr32[2];
	b += daddr->s6_addr32[3];
	c += id;
	__jhash_mix(a, b, c);

	return c & (FRAG6Q_HASHSZ - 1);
}

static struct timer_list nf_ct_frag6_secret_timer;
int nf_ct_frag6_secret_interval = 10 * 60 * HZ;

static void nf_ct_frag6_secret_rebuild(unsigned long dummy)
{
	unsigned long now = jiffies;
	int i;

	write_lock(&nf_ct_frag6_lock);
	get_random_bytes(&nf_ct_frag6_hash_rnd, sizeof(u32));
	for (i = 0; i < FRAG6Q_HASHSZ; i++) {
		struct nf_ct_frag6_queue *q;

		q = nf_ct_frag6_hash[i];
		while (q) {
			struct nf_ct_frag6_queue *next = q->next;
			unsigned int hval = ip6qhashfn(q->id,
						       &q->saddr,
						       &q->daddr);

			if (hval != i) {
				/* Unlink. */
				if (q->next)
					q->next->pprev = q->pprev;
				*q->pprev = q->next;

				/* Relink to new hash chain. */
				if ((q->next = nf_ct_frag6_hash[hval]) != NULL)
					q->next->pprev = &q->next;
				nf_ct_frag6_hash[hval] = q;
				q->pprev = &nf_ct_frag6_hash[hval];
			}

			q = next;
		}
	}
	write_unlock(&nf_ct_frag6_lock);

	mod_timer(&nf_ct_frag6_secret_timer, now + nf_ct_frag6_secret_interval);
}

atomic_t nf_ct_frag6_mem = ATOMIC_INIT(0);

/* Memory Tracking Functions. */
static inline void frag_kfree_skb(struct sk_buff *skb, unsigned int *work)
{
	if (work)
		*work -= skb->truesize;
	atomic_sub(skb->truesize, &nf_ct_frag6_mem);
	if (NFCT_FRAG6_CB(skb)->orig)
		kfree_skb(NFCT_FRAG6_CB(skb)->orig);

	kfree_skb(skb);
}

static inline void frag_free_queue(struct nf_ct_frag6_queue *fq,
				   unsigned int *work)
{
	if (work)
		*work -= sizeof(struct nf_ct_frag6_queue);
	atomic_sub(sizeof(struct nf_ct_frag6_queue), &nf_ct_frag6_mem);
	kfree(fq);
}

static inline struct nf_ct_frag6_queue *frag_alloc_queue(void)
{
	struct nf_ct_frag6_queue *fq = kmalloc(sizeof(struct nf_ct_frag6_queue), GFP_ATOMIC);

	if (!fq)
		return NULL;
	atomic_add(sizeof(struct nf_ct_frag6_queue), &nf_ct_frag6_mem);
	return fq;
}

/* Destruction primitives. */

/* Complete destruction of fq. */
static void nf_ct_frag6_destroy(struct nf_ct_frag6_queue *fq,
				unsigned int *work)
{
	struct sk_buff *fp;

	BUG_TRAP(fq->last_in&COMPLETE);
	BUG_TRAP(del_timer(&fq->timer) == 0);

	/* Release all fragment data. */
	fp = fq->fragments;
	while (fp) {
		struct sk_buff *xp = fp->next;

		frag_kfree_skb(fp, work);
		fp = xp;
	}

	frag_free_queue(fq, work);
}

static __inline__ void fq_put(struct nf_ct_frag6_queue *fq, unsigned int *work)
{
	if (atomic_dec_and_test(&fq->refcnt))
		nf_ct_frag6_destroy(fq, work);
}

/* Kill fq entry. It is not destroyed immediately,
 * because caller (and someone more) holds reference count.
 */
static __inline__ void fq_kill(struct nf_ct_frag6_queue *fq)
{
	if (del_timer(&fq->timer))
		atomic_dec(&fq->refcnt);

	if (!(fq->last_in & COMPLETE)) {
		fq_unlink(fq);
		atomic_dec(&fq->refcnt);
		fq->last_in |= COMPLETE;
	}
}

static void nf_ct_frag6_evictor(void)
{
	struct nf_ct_frag6_queue *fq;
	struct list_head *tmp;
	unsigned int work;

	work = atomic_read(&nf_ct_frag6_mem);
	if (work <= nf_ct_frag6_low_thresh)
		return;

	work -= nf_ct_frag6_low_thresh;
	while (work > 0) {
		read_lock(&nf_ct_frag6_lock);
		if (list_empty(&nf_ct_frag6_lru_list)) {
			read_unlock(&nf_ct_frag6_lock);
			return;
		}
		tmp = nf_ct_frag6_lru_list.next;
		BUG_ON(tmp == NULL);
		fq = list_entry(tmp, struct nf_ct_frag6_queue, lru_list);
		atomic_inc(&fq->refcnt);
		read_unlock(&nf_ct_frag6_lock);

		spin_lock(&fq->lock);
		if (!(fq->last_in&COMPLETE))
			fq_kill(fq);
		spin_unlock(&fq->lock);

		fq_put(fq, &work);
	}
}

static void nf_ct_frag6_expire(unsigned long data)
{
	struct nf_ct_frag6_queue *fq = (struct nf_ct_frag6_queue *) data;

	spin_lock(&fq->lock);

	if (fq->last_in & COMPLETE)
		goto out;

	fq_kill(fq);

out:
	spin_unlock(&fq->lock);
	fq_put(fq, NULL);
}

/* Creation primitives. */


static struct nf_ct_frag6_queue *nf_ct_frag6_intern(unsigned int hash,
					  struct nf_ct_frag6_queue *fq_in)
{
	struct nf_ct_frag6_queue *fq;

	write_lock(&nf_ct_frag6_lock);
#ifdef CONFIG_SMP
	for (fq = nf_ct_frag6_hash[hash]; fq; fq = fq->next) {
		if (fq->id == fq_in->id && 
		    !ipv6_addr_cmp(&fq_in->saddr, &fq->saddr) &&
		    !ipv6_addr_cmp(&fq_in->daddr, &fq->daddr)) {
			atomic_inc(&fq->refcnt);
			write_unlock(&nf_ct_frag6_lock);
			fq_in->last_in |= COMPLETE;
			fq_put(fq_in, NULL);
			return fq;
		}
	}
#endif
	fq = fq_in;

	if (!mod_timer(&fq->timer, jiffies + nf_ct_frag6_timeout))
		atomic_inc(&fq->refcnt);

	atomic_inc(&fq->refcnt);
	if ((fq->next = nf_ct_frag6_hash[hash]) != NULL)
		fq->next->pprev = &fq->next;
	nf_ct_frag6_hash[hash] = fq;
	fq->pprev = &nf_ct_frag6_hash[hash];
	INIT_LIST_HEAD(&fq->lru_list);
	list_add_tail(&fq->lru_list, &nf_ct_frag6_lru_list);
	nf_ct_frag6_nqueues++;
	write_unlock(&nf_ct_frag6_lock);
	return fq;
}


static struct nf_ct_frag6_queue *
nf_ct_frag6_create(unsigned int hash, u32 id, struct in6_addr *src,				   struct in6_addr *dst)
{
	struct nf_ct_frag6_queue *fq;

	if ((fq = frag_alloc_queue()) == NULL) {
		DEBUGP("Can't alloc new queue\n");
		goto oom;
	}

	memset(fq, 0, sizeof(struct nf_ct_frag6_queue));

	fq->id = id;
	ipv6_addr_copy(&fq->saddr, src);
	ipv6_addr_copy(&fq->daddr, dst);

	init_timer(&fq->timer);
	fq->timer.function = nf_ct_frag6_expire;
	fq->timer.data = (long) fq;
	fq->lock = SPIN_LOCK_UNLOCKED;
	atomic_set(&fq->refcnt, 1);

	return nf_ct_frag6_intern(hash, fq);

oom:
	return NULL;
}

static __inline__ struct nf_ct_frag6_queue *
fq_find(u32 id, struct in6_addr *src, struct in6_addr *dst)
{
	struct nf_ct_frag6_queue *fq;
	unsigned int hash = ip6qhashfn(id, src, dst);

	read_lock(&nf_ct_frag6_lock);
	for (fq = nf_ct_frag6_hash[hash]; fq; fq = fq->next) {
		if (fq->id == id && 
		    !ipv6_addr_cmp(src, &fq->saddr) &&
		    !ipv6_addr_cmp(dst, &fq->daddr)) {
			atomic_inc(&fq->refcnt);
			read_unlock(&nf_ct_frag6_lock);
			return fq;
		}
	}
	read_unlock(&nf_ct_frag6_lock);

	return nf_ct_frag6_create(hash, id, src, dst);
}


static int nf_ct_frag6_queue(struct nf_ct_frag6_queue *fq, struct sk_buff *skb, 
			     struct frag_hdr *fhdr, int nhoff)
{
	struct sk_buff *prev, *next;
	int offset, end;

	if (fq->last_in & COMPLETE) {
		DEBUGP("Allready completed\n");
		goto err;
	}

	offset = ntohs(fhdr->frag_off) & ~0x7;
	end = offset + (ntohs(skb->nh.ipv6h->payload_len) -
			((u8 *) (fhdr + 1) - (u8 *) (skb->nh.ipv6h + 1)));

	if ((unsigned int)end > IPV6_MAXPLEN) {
		DEBUGP("offset is too large.\n");
 		return -1;
	}

 	if (skb->ip_summed == CHECKSUM_HW)
 		skb->csum = csum_sub(skb->csum,
 				     csum_partial(skb->nh.raw,
						  (u8*)(fhdr + 1) - skb->nh.raw,
						  0));

	/* Is this the final fragment? */
	if (!(fhdr->frag_off & htons(IP6_MF))) {
		/* If we already have some bits beyond end
		 * or have different end, the segment is corrupted.
		 */
		if (end < fq->len ||
		    ((fq->last_in & LAST_IN) && end != fq->len)) {
			DEBUGP("already received last fragment\n");
			goto err;
		}
		fq->last_in |= LAST_IN;
		fq->len = end;
	} else {
		/* Check if the fragment is rounded to 8 bytes.
		 * Required by the RFC.
		 */
		if (end & 0x7) {
			/* RFC2460 says always send parameter problem in
			 * this case. -DaveM