[IPV4]: Safer reassembly

Another spin of Herbert Xu's "safer ip reassembly" patch
for 2.6.16.

(The original patch is here:
http://marc.theaimsgroup.com/?l=linux-netdev&m=112281936522415&w=2
and my only contribution is to have tested it.)

This patch (optionally) does additional checks before accepting IP
fragments, which can greatly reduce the possibility of reassembling
fragments which originated from different IP datagrams.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Arthur Kepner <akepner@sgi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/networking/ip-sysctl.txt

@@ -46,6 +46,29 @@ ipfrag_secret_interval - INTEGER
 	for the hash secret) for IP fragments.
 	Default: 600
 
+ipfrag_max_dist - INTEGER
+	ipfrag_max_dist is a non-negative integer value which defines the 
+	maximum "disorder" which is allowed among fragments which share a 
+	common IP source address. Note that reordering of packets is 
+	not unusual, but if a large number of fragments arrive from a source 
+	IP address while a particular fragment queue remains incomplete, it 
+	probably indicates that one or more fragments belonging to that queue 
+	have been lost. When ipfrag_max_dist is positive, an additional check 
+	is done on fragments before they are added to a reassembly queue - if 
+	ipfrag_max_dist (or more) fragments have arrived from a particular IP 
+	address between additions to any IP fragment queue using that source 
+	address, it's presumed that one or more fragments in the queue are 
+	lost. The existing fragment queue will be dropped, and a new one 
+	started. An ipfrag_max_dist value of zero disables this check.
+
+	Using a very small value, e.g. 1 or 2, for ipfrag_max_dist can
+	result in unnecessarily dropping fragment queues when normal
+	reordering of packets occurs, which could lead to poor application 
+	performance. Using a very large value, e.g. 50000, increases the 
+	likelihood of incorrectly reassembling IP fragments that originate 
+	from different IP datagrams, which could result in data corruption.
+	Default: 64
+
 INET peer storage:
 
 inet_peer_threshold - INTEGER

include/linux/sysctl.h

@@ -390,6 +390,7 @@ enum
 	NET_IPV4_ICMP_ERRORS_USE_INBOUND_IFADDR=109,
 	NET_TCP_CONG_CONTROL=110,
 	NET_TCP_ABC=111,
+	NET_IPV4_IPFRAG_MAX_DIST=112,
 };
 
 enum {

include/net/inetpeer.h

@@ -25,6 +25,7 @@ struct inet_peer
 	__u32			v4daddr;	/* peer's address */
 	__u16			avl_height;
 	__u16			ip_id_count;	/* IP ID for the next packet */
+	atomic_t		rid;		/* Frag reception counter */
 	__u32			tcp_ts;
 	unsigned long		tcp_ts_stamp;
 };

include/net/ip.h

@@ -45,6 +45,7 @@ struct inet_skb_parm
 #define IPSKB_TRANSLATED	2
 #define IPSKB_FORWARDED		4
 #define IPSKB_XFRM_TUNNEL_SIZE	8
+#define IPSKB_FRAG_COMPLETE	16
 };
 
 struct ipcm_cookie
@@ -168,6 +169,7 @@ extern int sysctl_ipfrag_high_thresh;
 extern int sysctl_ipfrag_low_thresh;
 extern int sysctl_ipfrag_time;
 extern int sysctl_ipfrag_secret_interval;
+extern int sysctl_ipfrag_max_dist;
 
 /* From inetpeer.c */
 extern int inet_peer_threshold;

net/ipv4/inetpeer.c

@@ -401,6 +401,7 @@ struct inet_peer *inet_getpeer(__u32 daddr, int create)
 		return NULL;
 	n->v4daddr = daddr;
 	atomic_set(&n->refcnt, 1);
+	atomic_set(&n->rid, 0);
 	n->ip_id_count = secure_ip_id(daddr);
 	n->tcp_ts_stamp = 0;
 

net/ipv4/ip_fragment.c

@@ -22,6 +22,7 @@
  *		Patrick McHardy :	LRU queue of frag heads for evictor.
  */
 
+#include <linux/compiler.h>
 #include <linux/config.h>
 #include <linux/module.h>
 #include <linux/types.h>
@@ -38,6 +39,7 @@
 #include <net/ip.h>
 #include <net/icmp.h>
 #include <net/checksum.h>
+#include <net/inetpeer.h>
 #include <linux/tcp.h>
 #include <linux/udp.h>
 #include <linux/inet.h>
@@ -56,6 +58,8 @@
 int sysctl_ipfrag_high_thresh = 256*1024;
 int sysctl_ipfrag_low_thresh = 192*1024;
 
+int sysctl_ipfrag_max_dist = 64;
+
 /* Important NOTE! Fragment queue must be destroyed before MSL expires.
  * RFC791 is wrong proposing to prolongate timer each fragment arrival by TTL.
  */
@@ -89,8 +93,10 @@ struct ipq {
 	spinlock_t	lock;
 	atomic_t	refcnt;
 	struct timer_list timer;	/* when will this queue expire?		*/
-	int		iif;
 	struct timeval	stamp;
+	int             iif;
+	unsigned int    rid;
+	struct inet_peer *peer;
 };
 
 /* Hash table. */
@@ -195,6 +201,9 @@ static void ip_frag_destroy(struct ipq *qp, int *work)
 	BUG_TRAP(qp->last_in&COMPLETE);
 	BUG_TRAP(del_timer(&qp->timer) == 0);
 
+	if (qp->peer)
+		inet_putpeer(qp->peer);
+
 	/* Release all fragment data. */
 	fp = qp->fragments;
 	while (fp) {
@@ -353,6 +362,7 @@ static struct ipq *ip_frag_create(unsigned hash, struct iphdr *iph, u32 user)
 	qp->meat = 0;
 	qp->fragments = NULL;
 	qp->iif = 0;
+	qp->peer = sysctl_ipfrag_max_dist ? inet_getpeer(iph->saddr, 1) : NULL;
 
 	/* Initialize a timer for this entry. */
 	init_timer(&qp->timer);
@@ -398,6 +408,56 @@ static inline struct ipq *ip_find(struct iphdr *iph, u32 user)
 	return ip_frag_create(hash, iph, user);
 }
 
+/* Is the fragment too far ahead to be part of ipq? */
+static inline int ip_frag_too_far(struct ipq *qp)
+{
+	struct inet_peer *peer = qp->peer;
+	unsigned int max = sysctl_ipfrag_max_dist;
+	unsigned int start, end;
+
+	int rc;
+
+	if (!peer || !max)
+		return 0;
+
+	start = qp->rid;
+	end = atomic_inc_return(&peer->rid);
+	qp->rid = end;
+
+	rc = qp->fragments && (end - start) > max;
+
+	if (rc) {
+		IP_INC_STATS_BH(IPSTATS_MIB_REASMFAILS);
+	}
+
+	return rc;
+}
+
+static int ip_frag_reinit(struct ipq *qp)
+{
+	struct sk_buff *fp;
+
+	if (!mod_timer(&qp->timer, jiffies + sysctl_ipfrag_time)) {
+		atomic_inc(&qp->refcnt);
+		return -ETIMEDOUT;
+	}
+
+	fp = qp->fragments;
+	do {
+		struct sk_buff *xp = fp->next;
+		frag_kfree_skb(fp, NULL);
+		fp = xp;
+	} while (fp);
+
+	qp->last_in = 0;
+	qp->len = 0;
+	qp->meat = 0;
+	qp->fragments = NULL;
+	qp->iif = 0;
+
+	return 0;
+}
+
 /* Add new segment to existing queue. */
 static void ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
 {
@@ -408,6 +468,12 @@ static void ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
 	if (qp->last_in & COMPLETE)
 		goto err;
 
+	if (!(IPCB(skb)->flags & IPSKB_FRAG_COMPLETE) &&
+	    unlikely(ip_frag_too_far(qp)) && unlikely(ip_frag_reinit(qp))) {
+		ipq_kill(qp);
+		goto err;
+	}
+
  	offset = ntohs(skb->nh.iph->frag_off);
 	flags = offset & ~IP_OFFSET;
 	offset &= IP_OFFSET;

net/ipv4/ip_output.c

@@ -445,6 +445,7 @@ int ip_fragment(struct sk_buff *skb, int (*output)(struct sk_buff*))
 
 	hlen = iph->ihl * 4;
 	mtu = dst_mtu(&rt->u.dst) - hlen;	/* Size of data space */
+	IPCB(skb)->flags |= IPSKB_FRAG_COMPLETE;
 
 	/* When frag_list is given, use it. First, check its validity:
 	 * some transformers could create wrong frag_list or break existing

net/ipv4/sysctl_net_ipv4.c

@@ -22,6 +22,7 @@
 extern int sysctl_ip_nonlocal_bind;
 
 #ifdef CONFIG_SYSCTL
+static int zero;
 static int tcp_retr1_max = 255; 
 static int ip_local_port_range_min[] = { 1, 1 };
 static int ip_local_port_range_max[] = { 65535, 65535 };
@@ -613,6 +614,15 @@ ctl_table ipv4_table[] = {
 		.proc_handler	= &proc_dointvec_jiffies,
 		.strategy	= &sysctl_jiffies
 	},
+	{
+		.ctl_name	= NET_IPV4_IPFRAG_MAX_DIST,
+		.procname	= "ipfrag_max_dist",
+		.data		= &sysctl_ipfrag_max_dist,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec_minmax,
+		.extra1		= &zero
+	},
 	{
 		.ctl_name	= NET_TCP_NO_METRICS_SAVE,
 		.procname	= "tcp_no_metrics_save",