[mirror_ubuntu-zesty-kernel.git] / net / ipv6 / reassembly.c

/*
 *	IPv6 fragment reassembly
 *	Linux INET6 implementation 
 *
 *	Authors:
 *	Pedro Roque		<roque@di.fc.ul.pt>	
 *
 *	$Id: reassembly.c,v 1.26 2001/03/07 22:00:57 davem Exp $
 *
 *	Based on: net/ipv4/ip_fragment.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.
 */

/* 
 *	Fixes:	
 *	Andi Kleen	Make it work with multiple hosts.
 *			More RFC compliance.
 *
 *      Horst von Brand Add missing #include <linux/string.h>
 *	Alexey Kuznetsov	SMP races, threading, cleanup.
 *	Patrick McHardy		LRU queue of frag heads for evictor.
 *	Mitsuru KANDA @USAGI	Register inet6_protocol{}.
 *	David Stevens and
 *	YOSHIFUJI,H. @USAGI	Always remove fragment header to
 *				calculate ICV correctly.
 */
#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>

int sysctl_ip6frag_high_thresh = 256*1024;
int sysctl_ip6frag_low_thresh = 192*1024;

int sysctl_ip6frag_time = IPV6_FRAG_TIMEOUT;

struct ip6frag_skb_cb
{
	struct inet6_skb_parm	h;
	int			offset;
};

#define FRAG6_CB(skb)	((struct ip6frag_skb_cb*)((skb)->cb))


/*
 *	Equivalent of ipv4 struct ipq
 */

struct frag_queue
{
	struct hlist_node	list;
	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;
	int			iif;
	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;
};

/* Hash table. */

#define IP6Q_HASHSZ	64

static struct hlist_head ip6_frag_hash[IP6Q_HASHSZ];
static DEFINE_RWLOCK(ip6_frag_lock);
static u32 ip6_frag_hash_rnd;
static LIST_HEAD(ip6_frag_lru_list);
int ip6_frag_nqueues = 0;

static __inline__ void __fq_unlink(struct frag_queue *fq)
{
	hlist_del(&fq->list);
	list_del(&fq->lru_list);
	ip6_frag_nqueues--;
}

static __inline__ void fq_unlink(struct frag_queue *fq)
{
	write_lock(&ip6_frag_lock);
	__fq_unlink(fq);
	write_unlock(&ip6_frag_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 += ip6_frag_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 & (IP6Q_HASHSZ - 1);
}

static struct timer_list ip6_frag_secret_timer;
int sysctl_ip6frag_secret_interval = 10 * 60 * HZ;

static void ip6_frag_secret_rebuild(unsigned long dummy)
{
	unsigned long now = jiffies;
	int i;

	write_lock(&ip6_frag_lock);
	get_random_bytes(&ip6_frag_hash_rnd, sizeof(u32));
	for (i = 0; i < IP6Q_HASHSZ; i++) {
		struct frag_queue *q;
		struct hlist_node *p, *n;

		hlist_for_each_entry_safe(q, p, n, &ip6_frag_hash[i], list) {
			unsigned int hval = ip6qhashfn(q->id,
						       &q->saddr,
						       &q->daddr);

			if (hval != i) {
				hlist_del(&q->list);

				/* Relink to new hash chain. */
				hlist_add_head(&q->list,
					       &ip6_frag_hash[hval]);

			}
		}
	}
	write_unlock(&ip6_frag_lock);

	mod_timer(&ip6_frag_secret_timer, now + sysctl_ip6frag_secret_interval);
}

atomic_t ip6_frag_mem = ATOMIC_INIT(0);

/* Memory Tracking Functions. */
static inline void frag_kfree_skb(struct sk_buff *skb, int *work)
{
	if (work)
		*work -= skb->truesize;
	atomic_sub(skb->truesize, &ip6_frag_mem);
	kfree_skb(skb);
}

static inline void frag_free_queue(struct frag_queue *fq, int *work)
{
	if (work)
		*work -= sizeof(struct frag_queue);
	atomic_sub(sizeof(struct frag_queue), &ip6_frag_mem);
	kfree(fq);
}

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

	if(!fq)
		return NULL;
	atomic_add(sizeof(struct frag_queue), &ip6_frag_mem);
	return fq;
}

/* Destruction primitives. */

/* Complete destruction of fq. */
static void ip6_frag_destroy(struct frag_queue *fq, 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 frag_queue *fq, int *work)
{
	if (atomic_dec_and_test(&fq->refcnt))
		ip6_frag_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 frag_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 ip6_evictor(void)
{
	struct frag_queue *fq;
	struct list_head *tmp;
	int work;

	work = atomic_read(&ip6_frag_mem) - sysctl_ip6frag_low_thresh;
	if (work <= 0)
		return;

	while(work > 0) {
		read_lock(&ip6_frag_lock);
		if (list_empty(&ip6_frag_lru_list)) {
			read_unlock(&ip6_frag_lock);
			return;
		}
		tmp = ip6_frag_lru_list.next;
		fq = list_entry(tmp, struct frag_queue, lru_list);
		atomic_inc(&fq->refcnt);
		read_unlock(&ip6_frag_lock);

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

		fq_put(fq, &work);
		IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
	}
}

static void ip6_frag_expire(unsigned long data)
{
	struct frag_queue *fq = (struct frag_queue *) data;

	spin_lock(&fq->lock);

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

	fq_kill(fq);

	IP6_INC_STATS_BH(IPSTATS_MIB_REASMTIMEOUT);
	IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);

	/* Send error only if the first segment arrived. */
	if (fq->last_in&FIRST_IN && fq->fragments) {
		struct net_device *dev = dev_get_by_index(fq->iif);

		/*
		   But use as source device on which LAST ARRIVED
		   segment was received. And do not use fq->dev
		   pointer directly, device might already disappeared.
		 */
		if (dev) {
			fq->fragments->dev = dev;
			icmpv6_send(fq->fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0,
				    dev);
			dev_put(dev);
		}
	}
out:
	spin_unlock(&fq->lock);
	fq_put(fq, NULL);
}

/* Creation primitives. */


static struct frag_queue *ip6_frag_intern(unsigned int hash,
					  struct frag_queue *fq_in)
{
	struct frag_queue *fq;
#ifdef CONFIG_SMP
	struct hlist_node *n;
#endif

	write_lock(&ip6_frag_lock);
#ifdef CONFIG_SMP
	hlist_for_each_entry(fq, n, &ip6_frag_hash[hash], list) {
		if (fq->id == fq_in->id && 
		    ipv6_addr_equal(&fq_in->saddr, &fq->saddr) &&
		    ipv6_addr_equal(&fq_in->daddr, &fq->daddr)) {
			atomic_inc(&fq->refcnt);
			write_unlock(&ip6_frag_lock);
			fq_in->last_in |= COMPLETE;
			fq_put(fq_in, NULL);
			return fq;
		}
	}
#endif
	fq = fq_in;

	if (!mod_timer(&fq->timer, jiffies + sysctl_ip6frag_time))
		atomic_inc(&fq->refcnt);

	atomic_inc(&fq->refcnt);
	hlist_add_head(&fq->list, &ip6_frag_hash[hash]);
	INIT_LIST_HEAD(&fq->lru_list);
	list_add_tail(&fq->lru_list, &ip6_frag_lru_list);
	ip6_frag_nqueues++;
	write_unlock(&ip6_frag_lock);
	return fq;
}


static struct frag_queue *
ip6_frag_create(unsigned int hash, u32 id, struct in6_addr *src, struct in6_addr *dst)
{
	struct frag_queue *fq;

	if ((fq = frag_alloc_queue()) == NULL)
		goto oom;

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

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

	init_timer(&fq->timer);
	fq->timer.function = ip6_frag_expire;
	fq->timer.data = (long) fq;
	spin_lock_init(&fq->lock);
	atomic_set(&fq->refcnt, 1);

	return ip6_frag_intern(hash, fq);

oom:
	IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
	return NULL;
}

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

	read_lock(&ip6_frag_lock);
	hlist_for_each_entry(fq, n, &ip6_frag_hash[hash], list) {
		if (fq->id == id && 
		    ipv6_addr_equal(src, &fq->saddr) &&
		    ipv6_addr_equal(dst, &fq->daddr)) {
			atomic_inc(&fq->refcnt);
			read_unlock(&ip6_frag_lock);
			return fq;
		}
	}
	read_unlock(&ip6_frag_lock);

	return ip6_frag_create(hash, id, src, dst);
}


static void ip6_frag_queue(struct frag_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)
		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) {
		IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
 		icmpv6_param_prob(skb,ICMPV6_HDR_FIELD, (u8*)&fhdr->frag_off - skb->nh.raw);
 		return;
	}

 	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))
			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
			 */
			IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
			icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, 
					  offsetof(struct ipv6hdr, payload_len));
			return;
		}
		if (end > fq->len) {
			/* Some bits beyond end -> corruption. */
			if (fq->last_in & LAST_IN)
				goto err;
			fq->len = end;
		}
	}

	if (end == offset)
		goto err;

	/* Point into the IP datagram 'data' part. */
	if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
		goto err;
	
	if (pskb_trim_rcsum(skb, end - offset))
		goto err;

	/* Find out which fragments are in front and at the back of us
	 * in the chain of fragments so far.  We must know where to put
	 * this fragment, right?
	 */
	prev = NULL;
	for(next = fq->fragments; next != NULL; next = next->next) {
		if (FRAG6_CB(next)->offset >= offset)
			break;	/* bingo! */
		prev = next;
	}

	/* We found where to put this one.  Check for overlap with
	 * preceding fragment, and, if needed, align things so that
	 * any overlaps are eliminated.
	 */
	if (prev) {
		int i = (FRAG6_CB(prev)->offset + prev->len) - offset;

		if (i > 0) {
			offset += i;
			if (end <= offset)
				goto err;
			if (!pskb_pull(skb, i))
				goto err;
			if (skb->ip_summed != CHECKSUM_UNNECESSARY)
				skb->ip_summed = CHECKSUM_NONE;
		}
	}

	/* Look for overlap with succeeding segments.
	 * If we can merge fragments, do it.
	 */
	while (next && FRAG6_CB(next)->offset < end) {
		int i = end - FRAG6_CB(next)->offset; /* overlap is 'i' bytes */

		if (i < next->len) {
			/* Eat head of the next overlapped fragment
			 * and leave the loop. The next ones cannot overlap.
			 */
			if (!pskb_pull(next, i))
				goto err;
			FRAG6_CB(next)->offset += i;	/* next fragment */
			fq->meat -= i;
			if (next->ip_summed != CHECKSUM_UNNECESSARY)
				next->ip_summed = CHECKSUM_NONE;
			break;
		} else {
			struct sk_buff *free_it = next;

			/* Old fragment is completely overridden with
			 * new one drop it.
			 */
			next = next->next;

			if (prev)
				prev->next = next;
			else
				fq->fragments = next;

			fq->meat -= free_it->len;
			frag_kfree_skb(free_it, NULL);
		}
	}

	FRAG6_CB(skb)->offset = offset;

	/* Insert this fragment in the chain of fragments. */
	skb->next = next;
	if (prev)
		prev->next = skb;
	else
		fq->fragments = skb;

	if (skb->dev)
		fq->iif = skb->dev->ifindex;
	skb->dev = NULL;
	skb_get_timestamp(skb, &fq->stamp);
	fq->meat += skb->len;
	atomic_add(skb->truesize, &ip6_frag_mem);

	/* The first fragment.
	 * nhoffset is obtained from the first fragment, of course.
	 */
	if (offset == 0) {
		fq->nhoffset = nhoff;
		fq->last_in |= FIRST_IN;
	}
	write_lock(&ip6_frag_lock);
	list_move_tail(&fq->lru_list, &ip6_frag_lru_list);
	write_unlock(&ip6_frag_lock);
	return;

err:
	IP6_INC_STATS(IPSTATS_MIB_REASMFAILS);
	kfree_skb(skb);
}

/*
 *	Check if this packet is complete.
 *	Returns NULL on failure by any reason, and pointer
 *	to current nexthdr field in reassembled frame.
 *
 *	It is called with locked fq, and caller must check that
 *	queue is eligible for reassembly i.e. it is not COMPLETE,
 *	the last and the first frames arrived and all the bits are here.
 */
static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff **skb_in,
			  struct net_device *dev)
{
	struct sk_buff *fp, *head = fq->fragments;
	int    payload_len;
	unsigned int nhoff;

	fq_kill(fq);

	BUG_TRAP(head != NULL);
	BUG_TRAP(FRAG6_CB(head)->offset == 0);

	/* Unfragmented part is taken from the first segment. */
	payload_len = (head->data - head->nh.raw) - sizeof(struct ipv6hdr) + fq->len - sizeof(struct frag_hdr);
	if (payload_len > IPV6_MAXPLEN)
		goto out_oversize;

	/* Head of list must not be cloned. */
	if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC))
		goto out_oom;

	/* If the first fragment is fragmented itself, we split
	 * it to two chunks: the first with data and paged part
	 * and the second, holding only fragments. */
	if (skb_shinfo(head)->frag_list) {
		struct sk_buff *clone;
		int i, plen = 0;

		if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
			goto out_oom;
		clone->next = head->next;
		head->next = clone;
		skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
		skb_shinfo(head)->frag_list = NULL;
		for (i=0; i<skb_shinfo(head)->nr_frags; i++)
			plen += skb_shinfo(head)->frags[i].size;
		clone->len = clone->data_len = head->data_len - plen;
		head->data_len -= clone->len;
		head->len -= clone->len;
		clone->csum = 0;
		clone->ip_summed = head->ip_summed;
		atomic_add(clone->truesize, &ip6_frag_mem);
	}

	/* We have to remove fragment header from datagram and to relocate
	 * header in order to calculate ICV correctly. */
	nhoff = fq->nhoffset;
	head->nh.raw[nhoff] = head->h.raw[0];
	memmove(head->head + sizeof(struct frag_hdr), head->head, 
		(head->data - head->head) - sizeof(struct frag_hdr));
	head->mac.raw += sizeof(struct frag_hdr);
	head->nh.raw += sizeof(struct frag_hdr);

	skb_shinfo(head)->frag_list = head->next;
	head->h.raw = head->data;
	skb_push(head, head->data - head->nh.raw);
	atomic_sub(head->truesize, &ip6_frag_mem);

	for (fp=head->next; fp; fp = fp->next) {
		head->data_len += fp->len;
		head->len += fp->len;
		if (head->ip_summed != fp->ip_summed)
			head->ip_summed = CHECKSUM_NONE;
		else if (head->ip_summed == CHECKSUM_HW)
			head->csum = csum_add(head->csum, fp->csum);
		head->truesize += fp->truesize;
		atomic_sub(fp->truesize, &ip6_frag_mem);
	}

	head->next = NULL;
	head->dev = dev;
	skb_set_timestamp(head, &fq->stamp);
	head->nh.ipv6h->payload_len = htons(payload_len);
	IP6CB(head)->nhoff = nhoff;

	*skb_in = head;

	/* Yes, and fold redundant checksum back. 8) */
	if (head->ip_summed == CHECKSUM_HW)
		head->csum = csum_partial(head->nh.raw, head->h.raw-head->nh.raw, head->csum);

	IP6_INC_STATS_BH(IPSTATS_MIB_REASMOKS);
	fq->fragments = NULL;
	return 1;

out_oversize:
	if (net_ratelimit())
		printk(KERN_DEBUG "ip6_frag_reasm: payload len = %d\n", payload_len);
	goto out_fail;
out_oom:
	if (net_ratelimit())
		printk(KERN_DEBUG "ip6_frag_reasm: no memory for reassembly\n");
out_fail:
	IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
	return -1;
}

static int ipv6_frag_rcv(struct sk_buff **skbp)
{
	struct sk_buff *skb = *skbp; 
	struct net_device *dev = skb->dev;
	struct frag_hdr *fhdr;
	struct frag_queue *fq;
	struct ipv6hdr *hdr;

	hdr = skb->nh.ipv6h;

	IP6_INC_STATS_BH(IPSTATS_MIB_REASMREQDS);

	/* Jumbo payload inhibits frag. header */
	if (hdr->payload_len==0) {
		IP6_INC_STATS(IPSTATS_MIB_INHDRERRORS);
		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb->h.raw-skb->nh.raw);
		return -1;
	}
	if (!pskb_may_pull(skb, (skb->h.raw-skb->data)+sizeof(struct frag_hdr))) {
		IP6_INC_STATS(IPSTATS_MIB_INHDRERRORS);
		icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb->h.raw-skb->nh.raw);
		return -1;
	}

	hdr = skb->nh.ipv6h;
	fhdr = (struct frag_hdr *)skb->h.raw;

	if (!(fhdr->frag_off & htons(0xFFF9))) {
		/* It is not a fragmented frame */
		skb->h.raw += sizeof(struct frag_hdr);
		IP6_INC_STATS_BH(IPSTATS_MIB_REASMOKS);

		IP6CB(skb)->nhoff = (u8*)fhdr - skb->nh.raw;
		return 1;
	}

	if (atomic_read(&ip6_frag_mem) > sysctl_ip6frag_high_thresh)
		ip6_evictor();

	if ((fq = fq_find(fhdr->identification, &hdr->saddr, &hdr->daddr)) != NULL) {
		int ret = -1;

		spin_lock(&fq->lock);

		ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);

		if (fq->last_in == (FIRST_IN|LAST_IN) &&
		    fq->meat == fq->len)
			ret = ip6_frag_reasm(fq, skbp, dev);

		spin_unlock(&fq->lock);
		fq_put(fq, NULL);
		return ret;
	}

	IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
	kfree_skb(skb);
	return -1;
}

static struct inet6_protocol frag_protocol =
{
	.handler	=	ipv6_frag_rcv,
	.flags		=	INET6_PROTO_NOPOLICY,
};

void __init ipv6_frag_init(void)
{
	if (inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT) < 0)
		printk(KERN_ERR "ipv6_frag_init: Could not register protocol\n");

	ip6_frag_hash_rnd = (u32) ((num_physpages ^ (num_physpages>>7)) ^
				   (jiffies ^ (jiffies >> 6)));

	init_timer(&ip6_frag_secret_timer);
	ip6_frag_secret_timer.function = ip6_frag_secret_rebuild;
	ip6_frag_secret_timer.expires = jiffies + sysctl_ip6frag_secret_interval;
	add_timer(&ip6_frag_secret_timer);
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* IPv6 fragment reassembly
	3	* Linux INET6 implementation
	4	*
	5	* Authors:
	6	* Pedro Roque <roque@di.fc.ul.pt>
	7	*
	8	* $Id: reassembly.c,v 1.26 2001/03/07 22:00:57 davem Exp $
	9	*
	10	* Based on: net/ipv4/ip_fragment.c
	11	*
	12	* This program is free software; you can redistribute it and/or
	13	* modify it under the terms of the GNU General Public License
	14	* as published by the Free Software Foundation; either version
	15	* 2 of the License, or (at your option) any later version.
	16	*/
	17
	18	/*
	19	* Fixes:
	20	* Andi Kleen Make it work with multiple hosts.
	21	* More RFC compliance.
	22	*
	23	* Horst von Brand Add missing #include <linux/string.h>
	24	* Alexey Kuznetsov SMP races, threading, cleanup.
	25	* Patrick McHardy LRU queue of frag heads for evictor.
	26	* Mitsuru KANDA @USAGI Register inet6_protocol{}.
	27	* David Stevens and
	28	* YOSHIFUJI,H. @USAGI Always remove fragment header to
	29	* calculate ICV correctly.
	30	*/
	31	#include <linux/config.h>
	32	#include <linux/errno.h>
	33	#include <linux/types.h>
	34	#include <linux/string.h>
	35	#include <linux/socket.h>
	36	#include <linux/sockios.h>
	37	#include <linux/jiffies.h>
	38	#include <linux/net.h>
	39	#include <linux/list.h>
	40	#include <linux/netdevice.h>
	41	#include <linux/in6.h>
	42	#include <linux/ipv6.h>
	43	#include <linux/icmpv6.h>
	44	#include <linux/random.h>
	45	#include <linux/jhash.h>
	46
	47	#include <net/sock.h>
	48	#include <net/snmp.h>
	49
	50	#include <net/ipv6.h>
	51	#include <net/protocol.h>
	52	#include <net/transp_v6.h>
	53	#include <net/rawv6.h>
	54	#include <net/ndisc.h>
	55	#include <net/addrconf.h>
	56
	57	int sysctl_ip6frag_high_thresh = 256*1024;
	58	int sysctl_ip6frag_low_thresh = 192*1024;
	59
	60	int sysctl_ip6frag_time = IPV6_FRAG_TIMEOUT;
	61
	62	struct ip6frag_skb_cb
	63	{
	64	struct inet6_skb_parm h;
65	int offset;
66	};
67
68	#define FRAG6_CB(skb) ((struct ip6frag_skb_cb*)((skb)->cb))
69
70
71	/*
72	* Equivalent of ipv4 struct ipq
73	*/
74
75	struct frag_queue
76	{
e7c8a41e	77	struct hlist_node list;
1da177e4 LT	78	struct list_head lru_list; /* lru list member */
	79
	80	__u32 id; /* fragment id */
	81	struct in6_addr saddr;
	82	struct in6_addr daddr;
	83
	84	spinlock_t lock;
	85	atomic_t refcnt;
	86	struct timer_list timer; /* expire timer */
	87	struct sk_buff *fragments;
	88	int len;
	89	int meat;
	90	int iif;
	91	struct timeval stamp;
	92	unsigned int csum;
	93	__u8 last_in; /* has first/last segment arrived? */
	94	#define COMPLETE 4
	95	#define FIRST_IN 2
	96	#define LAST_IN 1
	97	__u16 nhoffset;
1da177e4 LT	98	};
	99
	100	/* Hash table. */
	101
	102	#define IP6Q_HASHSZ 64
	103
e7c8a41e	104	static struct hlist_head ip6_frag_hash[IP6Q_HASHSZ];
1da177e4 LT	105	static DEFINE_RWLOCK(ip6_frag_lock);
	106	static u32 ip6_frag_hash_rnd;
	107	static LIST_HEAD(ip6_frag_lru_list);
	108	int ip6_frag_nqueues = 0;
	109
	110	static __inline__ void __fq_unlink(struct frag_queue *fq)
	111	{
e7c8a41e	112	hlist_del(&fq->list);
1da177e4 LT	113	list_del(&fq->lru_list);
	114	ip6_frag_nqueues--;
	115	}
	116
	117	static __inline__ void fq_unlink(struct frag_queue *fq)
	118	{
	119	write_lock(&ip6_frag_lock);
	120	__fq_unlink(fq);
	121	write_unlock(&ip6_frag_lock);
	122	}
	123
	124	static unsigned int ip6qhashfn(u32 id, struct in6_addr *saddr,
	125	struct in6_addr *daddr)
	126	{
	127	u32 a, b, c;
	128
	129	a = saddr->s6_addr32[0];
	130	b = saddr->s6_addr32[1];
	131	c = saddr->s6_addr32[2];
	132
	133	a += JHASH_GOLDEN_RATIO;
	134	b += JHASH_GOLDEN_RATIO;
	135	c += ip6_frag_hash_rnd;
	136	__jhash_mix(a, b, c);
	137
	138	a += saddr->s6_addr32[3];
	139	b += daddr->s6_addr32[0];
	140	c += daddr->s6_addr32[1];
	141	__jhash_mix(a, b, c);
	142
	143	a += daddr->s6_addr32[2];
	144	b += daddr->s6_addr32[3];
	145	c += id;
	146	__jhash_mix(a, b, c);
	147
	148	return c & (IP6Q_HASHSZ - 1);
	149	}
	150
	151	static struct timer_list ip6_frag_secret_timer;
	152	int sysctl_ip6frag_secret_interval = 10 * 60 * HZ;
	153
	154	static void ip6_frag_secret_rebuild(unsigned long dummy)
	155	{
	156	unsigned long now = jiffies;
	157	int i;
	158
	159	write_lock(&ip6_frag_lock);
	160	get_random_bytes(&ip6_frag_hash_rnd, sizeof(u32));
	161	for (i = 0; i < IP6Q_HASHSZ; i++) {
	162	struct frag_queue *q;
e7c8a41e	163	struct hlist_node p, n;
1da177e4	164
e7c8a41e	165	hlist_for_each_entry_safe(q, p, n, &ip6_frag_hash[i], list) {
1da177e4 LT	166	unsigned int hval = ip6qhashfn(q->id,
	167	&q->saddr,
	168	&q->daddr);
	169
	170	if (hval != i) {
e7c8a41e	171	hlist_del(&q->list);
1da177e4 LT	172
1da177e4 LT	173	/* Relink to new hash chain. */
e7c8a41e YK	174	hlist_add_head(&q->list,
e7c8a41e YK	175	&ip6_frag_hash[hval]);
1da177e4	176
e7c8a41e	177	}
1da177e4 LT	178	}
	179	}
	180	write_unlock(&ip6_frag_lock);
	181
	182	mod_timer(&ip6_frag_secret_timer, now + sysctl_ip6frag_secret_interval);
	183	}
	184
	185	atomic_t ip6_frag_mem = ATOMIC_INIT(0);
	186
	187	/* Memory Tracking Functions. */
	188	static inline void frag_kfree_skb(struct sk_buff skb, int work)
	189	{
	190	if (work)
	191	*work -= skb->truesize;
	192	atomic_sub(skb->truesize, &ip6_frag_mem);
	193	kfree_skb(skb);
	194	}
	195
	196	static inline void frag_free_queue(struct frag_queue fq, int work)
	197	{
	198	if (work)
	199	*work -= sizeof(struct frag_queue);
	200	atomic_sub(sizeof(struct frag_queue), &ip6_frag_mem);
	201	kfree(fq);
	202	}
	203
	204	static inline struct frag_queue *frag_alloc_queue(void)
	205	{
	206	struct frag_queue *fq = kmalloc(sizeof(struct frag_queue), GFP_ATOMIC);
	207
	208	if(!fq)
	209	return NULL;
	210	atomic_add(sizeof(struct frag_queue), &ip6_frag_mem);
	211	return fq;
	212	}
	213
	214	/* Destruction primitives. */
	215
	216	/* Complete destruction of fq. */
	217	static void ip6_frag_destroy(struct frag_queue fq, int work)
	218	{
	219	struct sk_buff *fp;
	220
	221	BUG_TRAP(fq->last_in&COMPLETE);
	222	BUG_TRAP(del_timer(&fq->timer) == 0);
	223
	224	/* Release all fragment data. */
	225	fp = fq->fragments;
	226	while (fp) {
	227	struct sk_buff *xp = fp->next;
	228
	229	frag_kfree_skb(fp, work);
	230	fp = xp;
	231	}
	232
	233	frag_free_queue(fq, work);
	234	}
	235
	236	static __inline__ void fq_put(struct frag_queue fq, int work)
	237	{
	238	if (atomic_dec_and_test(&fq->refcnt))
	239	ip6_frag_destroy(fq, work);
	240	}
	241
242	/* Kill fq entry. It is not destroyed immediately,
243	* because caller (and someone more) holds reference count.
244	*/
245	static __inline__ void fq_kill(struct frag_queue *fq)
246	{
247	if (del_timer(&fq->timer))
248	atomic_dec(&fq->refcnt);
249
250	if (!(fq->last_in & COMPLETE)) {
251	fq_unlink(fq);
252	atomic_dec(&fq->refcnt);
253	fq->last_in \|= COMPLETE;
254	}
255	}
256
257	static void ip6_evictor(void)
258	{
259	struct frag_queue *fq;
260	struct list_head *tmp;
261	int work;
262
263	work = atomic_read(&ip6_frag_mem) - sysctl_ip6frag_low_thresh;
264	if (work <= 0)
265	return;
266
267	while(work > 0) {
268	read_lock(&ip6_frag_lock);
269	if (list_empty(&ip6_frag_lru_list)) {
270	read_unlock(&ip6_frag_lock);
271	return;
272	}
273	tmp = ip6_frag_lru_list.next;
274	fq = list_entry(tmp, struct frag_queue, lru_list);
275	atomic_inc(&fq->refcnt);
276	read_unlock(&ip6_frag_lock);
277
278	spin_lock(&fq->lock);
279	if (!(fq->last_in&COMPLETE))
280	fq_kill(fq);
281	spin_unlock(&fq->lock);
282
283	fq_put(fq, &work);
284	IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
285	}
286	}
287
288	static void ip6_frag_expire(unsigned long data)
289	{
290	struct frag_queue fq = (struct frag_queue ) data;
291
292	spin_lock(&fq->lock);
293
294	if (fq->last_in & COMPLETE)
295	goto out;
296
297	fq_kill(fq);
298
299	IP6_INC_STATS_BH(IPSTATS_MIB_REASMTIMEOUT);
300	IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
301
302	/* Send error only if the first segment arrived. */
303	if (fq->last_in&FIRST_IN && fq->fragments) {
304	struct net_device *dev = dev_get_by_index(fq->iif);
305
306	/*
307	But use as source device on which LAST ARRIVED
308	segment was received. And do not use fq->dev
309	pointer directly, device might already disappeared.
310	*/
311	if (dev) {
312	fq->fragments->dev = dev;
313	icmpv6_send(fq->fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0,
314	dev);
315	dev_put(dev);
316	}
317	}
318	out:
319	spin_unlock(&fq->lock);
320	fq_put(fq, NULL);
321	}
322
323	/* Creation primitives. */
324
325
326	static struct frag_queue *ip6_frag_intern(unsigned int hash,
327	struct frag_queue *fq_in)
328	{
329	struct frag_queue *fq;
e7c8a41e YK	330	#ifdef CONFIG_SMP
	331	struct hlist_node *n;
	332	#endif
1da177e4 LT	333
	334	write_lock(&ip6_frag_lock);
	335	#ifdef CONFIG_SMP
e7c8a41e	336	hlist_for_each_entry(fq, n, &ip6_frag_hash[hash], list) {
1da177e4 LT	337	if (fq->id == fq_in->id &&
	338	ipv6_addr_equal(&fq_in->saddr, &fq->saddr) &&
	339	ipv6_addr_equal(&fq_in->daddr, &fq->daddr)) {
	340	atomic_inc(&fq->refcnt);
	341	write_unlock(&ip6_frag_lock);
	342	fq_in->last_in \|= COMPLETE;
	343	fq_put(fq_in, NULL);
	344	return fq;
	345	}
	346	}
	347	#endif
	348	fq = fq_in;
	349
	350	if (!mod_timer(&fq->timer, jiffies + sysctl_ip6frag_time))
	351	atomic_inc(&fq->refcnt);
	352
	353	atomic_inc(&fq->refcnt);
e7c8a41e	354	hlist_add_head(&fq->list, &ip6_frag_hash[hash]);
1da177e4 LT	355	INIT_LIST_HEAD(&fq->lru_list);
	356	list_add_tail(&fq->lru_list, &ip6_frag_lru_list);
	357	ip6_frag_nqueues++;
	358	write_unlock(&ip6_frag_lock);
	359	return fq;
	360	}
	361
	362
	363	static struct frag_queue *
	364	ip6_frag_create(unsigned int hash, u32 id, struct in6_addr src, struct in6_addr dst)
	365	{
	366	struct frag_queue *fq;
	367
	368	if ((fq = frag_alloc_queue()) == NULL)
	369	goto oom;
	370
	371	memset(fq, 0, sizeof(struct frag_queue));
	372
	373	fq->id = id;
	374	ipv6_addr_copy(&fq->saddr, src);
	375	ipv6_addr_copy(&fq->daddr, dst);
	376
	377	init_timer(&fq->timer);
	378	fq->timer.function = ip6_frag_expire;
	379	fq->timer.data = (long) fq;
	380	spin_lock_init(&fq->lock);
	381	atomic_set(&fq->refcnt, 1);
	382
	383	return ip6_frag_intern(hash, fq);
	384
	385	oom:
	386	IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
	387	return NULL;
	388	}
	389
	390	static __inline__ struct frag_queue *
	391	fq_find(u32 id, struct in6_addr src, struct in6_addr dst)
	392	{
	393	struct frag_queue *fq;
e7c8a41e	394	struct hlist_node *n;
1da177e4 LT	395	unsigned int hash = ip6qhashfn(id, src, dst);
	396
	397	read_lock(&ip6_frag_lock);
e7c8a41e	398	hlist_for_each_entry(fq, n, &ip6_frag_hash[hash], list) {
1da177e4 LT	399	if (fq->id == id &&
	400	ipv6_addr_equal(src, &fq->saddr) &&
	401	ipv6_addr_equal(dst, &fq->daddr)) {
	402	atomic_inc(&fq->refcnt);
	403	read_unlock(&ip6_frag_lock);
	404	return fq;
	405	}
	406	}
	407	read_unlock(&ip6_frag_lock);
	408
	409	return ip6_frag_create(hash, id, src, dst);
	410	}
	411
	412
	413	static void ip6_frag_queue(struct frag_queue fq, struct sk_buff skb,
	414	struct frag_hdr *fhdr, int nhoff)
	415	{
	416	struct sk_buff prev, next;
	417	int offset, end;
	418
	419	if (fq->last_in & COMPLETE)
	420	goto err;
	421
	422	offset = ntohs(fhdr->frag_off) & ~0x7;
	423	end = offset + (ntohs(skb->nh.ipv6h->payload_len) -
	424	((u8 ) (fhdr + 1) - (u8 ) (skb->nh.ipv6h + 1)));
	425
	426	if ((unsigned int)end > IPV6_MAXPLEN) {
	427	IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
	428	icmpv6_param_prob(skb,ICMPV6_HDR_FIELD, (u8*)&fhdr->frag_off - skb->nh.raw);
	429	return;
	430	}
	431
	432	if (skb->ip_summed == CHECKSUM_HW)
	433	skb->csum = csum_sub(skb->csum,
	434	csum_partial(skb->nh.raw, (u8*)(fhdr+1)-skb->nh.raw, 0));
	435
	436	/* Is this the final fragment? */
	437	if (!(fhdr->frag_off & htons(IP6_MF))) {
	438	/* If we already have some bits beyond end
	439	* or have different end, the segment is corrupted.
	440	*/
	441	if (end < fq->len \|\|
	442	((fq->last_in & LAST_IN) && end != fq->len))
	443	goto err;
	444	fq->last_in \|= LAST_IN;
	445	fq->len = end;
	446	} else {
	447	/* Check if the fragment is rounded to 8 bytes.
	448	* Required by the RFC.
	449	*/
	450	if (end & 0x7) {
	451	/* RFC2460 says always send parameter problem in
	452	* this case. -DaveM
	453	*/
	454	IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
	455	icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
	456	offsetof(struct ipv6hdr, payload_len));
	457	return;
	458	}
	459	if (end > fq->len) {
	460	/* Some bits beyond end -> corruption. */
	461	if (fq->last_in & LAST_IN)
	462	goto err;
463	fq->len = end;
464	}
465	}
466
467	if (end == offset)
468	goto err;
469
470	/* Point into the IP datagram 'data' part. */
471	if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
472	goto err;
42ca89c1 SH	473
	474	if (pskb_trim_rcsum(skb, end - offset))
	475	goto err;
1da177e4 LT	476
	477	/* Find out which fragments are in front and at the back of us
	478	* in the chain of fragments so far. We must know where to put
	479	* this fragment, right?
	480	*/
	481	prev = NULL;
	482	for(next = fq->fragments; next != NULL; next = next->next) {
	483	if (FRAG6_CB(next)->offset >= offset)
	484	break; /* bingo! */
	485	prev = next;
	486	}
	487
	488	/* We found where to put this one. Check for overlap with
	489	* preceding fragment, and, if needed, align things so that
	490	* any overlaps are eliminated.
	491	*/
	492	if (prev) {
	493	int i = (FRAG6_CB(prev)->offset + prev->len) - offset;
	494
	495	if (i > 0) {
	496	offset += i;
	497	if (end <= offset)
	498	goto err;
	499	if (!pskb_pull(skb, i))
	500	goto err;
	501	if (skb->ip_summed != CHECKSUM_UNNECESSARY)
	502	skb->ip_summed = CHECKSUM_NONE;
	503	}
	504	}
	505
	506	/* Look for overlap with succeeding segments.
	507	* If we can merge fragments, do it.
	508	*/
	509	while (next && FRAG6_CB(next)->offset < end) {
	510	int i = end - FRAG6_CB(next)->offset; /* overlap is 'i' bytes */
	511
	512	if (i < next->len) {
	513	/* Eat head of the next overlapped fragment
	514	* and leave the loop. The next ones cannot overlap.
	515	*/
	516	if (!pskb_pull(next, i))
	517	goto err;
	518	FRAG6_CB(next)->offset += i; /* next fragment */
	519	fq->meat -= i;
	520	if (next->ip_summed != CHECKSUM_UNNECESSARY)
	521	next->ip_summed = CHECKSUM_NONE;
	522	break;
	523	} else {
	524	struct sk_buff *free_it = next;
	525
	526	/* Old fragment is completely overridden with
	527	* new one drop it.
	528	*/
	529	next = next->next;
	530
	531	if (prev)
	532	prev->next = next;
	533	else
	534	fq->fragments = next;
	535
	536	fq->meat -= free_it->len;
	537	frag_kfree_skb(free_it, NULL);
	538	}
	539	}
540
541	FRAG6_CB(skb)->offset = offset;
542
543	/* Insert this fragment in the chain of fragments. */
544	skb->next = next;
545	if (prev)
546	prev->next = skb;
547	else
548	fq->fragments = skb;
549
550	if (skb->dev)
551	fq->iif = skb->dev->ifindex;
552	skb->dev = NULL;
a61bbcf2	553	skb_get_timestamp(skb, &fq->stamp);
1da177e4 LT	554	fq->meat += skb->len;
	555	atomic_add(skb->truesize, &ip6_frag_mem);
	556
	557	/* The first fragment.
	558	* nhoffset is obtained from the first fragment, of course.
	559	*/
	560	if (offset == 0) {
	561	fq->nhoffset = nhoff;
	562	fq->last_in \|= FIRST_IN;
	563	}
	564	write_lock(&ip6_frag_lock);
	565	list_move_tail(&fq->lru_list, &ip6_frag_lru_list);
	566	write_unlock(&ip6_frag_lock);
	567	return;
	568
	569	err:
	570	IP6_INC_STATS(IPSTATS_MIB_REASMFAILS);
	571	kfree_skb(skb);
	572	}
	573
	574	/*
	575	* Check if this packet is complete.
	576	* Returns NULL on failure by any reason, and pointer
	577	* to current nexthdr field in reassembled frame.
	578	*
	579	* It is called with locked fq, and caller must check that
	580	* queue is eligible for reassembly i.e. it is not COMPLETE,
	581	* the last and the first frames arrived and all the bits are here.
	582	*/
	583	static int ip6_frag_reasm(struct frag_queue fq, struct sk_buff *skb_in,
1da177e4 LT	584	struct net_device *dev)
	585	{
	586	struct sk_buff fp, head = fq->fragments;
	587	int payload_len;
	588	unsigned int nhoff;
	589
	590	fq_kill(fq);
	591
	592	BUG_TRAP(head != NULL);
	593	BUG_TRAP(FRAG6_CB(head)->offset == 0);
	594
	595	/* Unfragmented part is taken from the first segment. */
	596	payload_len = (head->data - head->nh.raw) - sizeof(struct ipv6hdr) + fq->len - sizeof(struct frag_hdr);
	597	if (payload_len > IPV6_MAXPLEN)
	598	goto out_oversize;
	599
	600	/* Head of list must not be cloned. */
	601	if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC))
	602	goto out_oom;
	603
	604	/* If the first fragment is fragmented itself, we split
	605	* it to two chunks: the first with data and paged part
	606	* and the second, holding only fragments. */
	607	if (skb_shinfo(head)->frag_list) {
	608	struct sk_buff *clone;
	609	int i, plen = 0;
	610
	611	if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
	612	goto out_oom;
	613	clone->next = head->next;
	614	head->next = clone;
	615	skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
	616	skb_shinfo(head)->frag_list = NULL;
	617	for (i=0; i<skb_shinfo(head)->nr_frags; i++)
	618	plen += skb_shinfo(head)->frags[i].size;
	619	clone->len = clone->data_len = head->data_len - plen;
	620	head->data_len -= clone->len;
	621	head->len -= clone->len;
	622	clone->csum = 0;
	623	clone->ip_summed = head->ip_summed;
	624	atomic_add(clone->truesize, &ip6_frag_mem);
	625	}
	626
	627	/* We have to remove fragment header from datagram and to relocate
	628	* header in order to calculate ICV correctly. */
	629	nhoff = fq->nhoffset;
	630	head->nh.raw[nhoff] = head->h.raw[0];
	631	memmove(head->head + sizeof(struct frag_hdr), head->head,
	632	(head->data - head->head) - sizeof(struct frag_hdr));
	633	head->mac.raw += sizeof(struct frag_hdr);
	634	head->nh.raw += sizeof(struct frag_hdr);
	635
	636	skb_shinfo(head)->frag_list = head->next;
	637	head->h.raw = head->data;
	638	skb_push(head, head->data - head->nh.raw);
	639	atomic_sub(head->truesize, &ip6_frag_mem);
	640
	641	for (fp=head->next; fp; fp = fp->next) {
	642	head->data_len += fp->len;
	643	head->len += fp->len;
	644	if (head->ip_summed != fp->ip_summed)
	645	head->ip_summed = CHECKSUM_NONE;
	646	else if (head->ip_summed == CHECKSUM_HW)
	647	head->csum = csum_add(head->csum, fp->csum);
648	head->truesize += fp->truesize;
649	atomic_sub(fp->truesize, &ip6_frag_mem);
650	}
651
652	head->next = NULL;
653	head->dev = dev;
a61bbcf2	654	skb_set_timestamp(head, &fq->stamp);
1da177e4	655	head->nh.ipv6h->payload_len = htons(payload_len);
951dbc8a	656	IP6CB(head)->nhoff = nhoff;
1da177e4 LT	657
	658	*skb_in = head;
	659
	660	/* Yes, and fold redundant checksum back. 8) */
	661	if (head->ip_summed == CHECKSUM_HW)
	662	head->csum = csum_partial(head->nh.raw, head->h.raw-head->nh.raw, head->csum);
	663
	664	IP6_INC_STATS_BH(IPSTATS_MIB_REASMOKS);
	665	fq->fragments = NULL;
1da177e4 LT	666	return 1;
	667
	668	out_oversize:
	669	if (net_ratelimit())
	670	printk(KERN_DEBUG "ip6_frag_reasm: payload len = %d\n", payload_len);
	671	goto out_fail;
	672	out_oom:
	673	if (net_ratelimit())
	674	printk(KERN_DEBUG "ip6_frag_reasm: no memory for reassembly\n");
	675	out_fail:
	676	IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
	677	return -1;
	678	}
	679
951dbc8a	680	static int ipv6_frag_rcv(struct sk_buff **skbp)
1da177e4 LT	681	{
	682	struct sk_buff skb = skbp;
	683	struct net_device *dev = skb->dev;
	684	struct frag_hdr *fhdr;
	685	struct frag_queue *fq;
	686	struct ipv6hdr *hdr;
	687
	688	hdr = skb->nh.ipv6h;
	689
	690	IP6_INC_STATS_BH(IPSTATS_MIB_REASMREQDS);
	691
	692	/* Jumbo payload inhibits frag. header */
	693	if (hdr->payload_len==0) {
	694	IP6_INC_STATS(IPSTATS_MIB_INHDRERRORS);
	695	icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb->h.raw-skb->nh.raw);
	696	return -1;
	697	}
	698	if (!pskb_may_pull(skb, (skb->h.raw-skb->data)+sizeof(struct frag_hdr))) {
	699	IP6_INC_STATS(IPSTATS_MIB_INHDRERRORS);
	700	icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb->h.raw-skb->nh.raw);
	701	return -1;
	702	}
	703
	704	hdr = skb->nh.ipv6h;
	705	fhdr = (struct frag_hdr *)skb->h.raw;
	706
	707	if (!(fhdr->frag_off & htons(0xFFF9))) {
	708	/* It is not a fragmented frame */
	709	skb->h.raw += sizeof(struct frag_hdr);
	710	IP6_INC_STATS_BH(IPSTATS_MIB_REASMOKS);
	711
951dbc8a	712	IP6CB(skb)->nhoff = (u8*)fhdr - skb->nh.raw;
1da177e4 LT	713	return 1;
	714	}
	715
	716	if (atomic_read(&ip6_frag_mem) > sysctl_ip6frag_high_thresh)
	717	ip6_evictor();
	718
	719	if ((fq = fq_find(fhdr->identification, &hdr->saddr, &hdr->daddr)) != NULL) {
	720	int ret = -1;
	721
	722	spin_lock(&fq->lock);
	723
951dbc8a	724	ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
1da177e4 LT	725
	726	if (fq->last_in == (FIRST_IN\|LAST_IN) &&
	727	fq->meat == fq->len)
951dbc8a	728	ret = ip6_frag_reasm(fq, skbp, dev);
1da177e4 LT	729
	730	spin_unlock(&fq->lock);
	731	fq_put(fq, NULL);
	732	return ret;
	733	}
	734
	735	IP6_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
	736	kfree_skb(skb);
	737	return -1;
	738	}
	739
	740	static struct inet6_protocol frag_protocol =
	741	{
	742	.handler = ipv6_frag_rcv,
	743	.flags = INET6_PROTO_NOPOLICY,
	744	};
	745
	746	void __init ipv6_frag_init(void)
	747	{
	748	if (inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT) < 0)
	749	printk(KERN_ERR "ipv6_frag_init: Could not register protocol\n");
	750
	751	ip6_frag_hash_rnd = (u32) ((num_physpages ^ (num_physpages>>7)) ^
	752	(jiffies ^ (jiffies >> 6)));
	753
	754	init_timer(&ip6_frag_secret_timer);
	755	ip6_frag_secret_timer.function = ip6_frag_secret_rebuild;
	756	ip6_frag_secret_timer.expires = jiffies + sysctl_ip6frag_secret_interval;
	757	add_timer(&ip6_frag_secret_timer);
	758	}