Merge branch 'ipv6-defrag-rbtree'

Peter Oskolkov says:

====================
net: IP defrag: use rbtrees in IPv6 defragmentation

Currently, IPv6 defragmentation code drops non-last fragments that
are smaller than 1280 bytes: see
commit 0ed4229b ("ipv6: defrag: drop non-last frags smaller than min mtu")

This behavior is not specified in IPv6 RFCs and appears to break compatibility
with some IPv6 implementations, as reported here:
https://www.spinics.net/lists/netdev/msg543846.html

This patchset contains four patches:
- patch 1 moves rbtree-related code from IPv4 to files shared b/w
IPv4/IPv6
- patch 2 changes IPv6 defragmenation code to use rbtrees for defrag
queue
- patch 3 changes nf_conntrack IPv6 defragmentation code to use rbtrees
- patch 4 changes ip_defrag selftest to test changes made in the
previous three patches.

Along the way, the 1280-byte restrictions are removed.

I plan to introduce similar changes to 6lowpan defragmentation code
once I figure out how to test it.
====================
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/net/inet_frag.h

@@ -77,8 +77,8 @@ struct inet_frag_queue {
 	struct timer_list	timer;
 	spinlock_t		lock;
 	refcount_t		refcnt;
-	struct sk_buff		*fragments;  /* Used in IPv6. */
-	struct rb_root		rb_fragments; /* Used in IPv4. */
+	struct sk_buff		*fragments;  /* used in 6lopwpan IPv6. */
+	struct rb_root		rb_fragments; /* Used in IPv4/IPv6. */
 	struct sk_buff		*fragments_tail;
 	struct sk_buff		*last_run_head;
 	ktime_t			stamp;
@@ -153,4 +153,16 @@ static inline void add_frag_mem_limit(struct netns_frags *nf, long val)
 
 extern const u8 ip_frag_ecn_table[16];
 
+/* Return values of inet_frag_queue_insert() */
+#define IPFRAG_OK	0
+#define IPFRAG_DUP	1
+#define IPFRAG_OVERLAP	2
+int inet_frag_queue_insert(struct inet_frag_queue *q, struct sk_buff *skb,
+			   int offset, int end);
+void *inet_frag_reasm_prepare(struct inet_frag_queue *q, struct sk_buff *skb,
+			      struct sk_buff *parent);
+void inet_frag_reasm_finish(struct inet_frag_queue *q, struct sk_buff *head,
+			    void *reasm_data);
+struct sk_buff *inet_frag_pull_head(struct inet_frag_queue *q);
+
 #endif

include/net/ipv6_frag.h

@@ -82,8 +82,15 @@ ip6frag_expire_frag_queue(struct net *net, struct frag_queue *fq)
 	__IP6_INC_STATS(net, __in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT);
 
 	/* Don't send error if the first segment did not arrive. */
-	head = fq->q.fragments;
-	if (!(fq->q.flags & INET_FRAG_FIRST_IN) || !head)
+	if (!(fq->q.flags & INET_FRAG_FIRST_IN))
+		goto out;
+
+	/* sk_buff::dev and sk_buff::rbnode are unionized. So we
+	 * pull the head out of the tree in order to be able to
+	 * deal with head->dev.
+	 */
+	head = inet_frag_pull_head(&fq->q);
+	if (!head)
 		goto out;
 
 	head->dev = dev;

net/ipv4/inet_fragment.c

@@ -25,6 +25,62 @@
 #include <net/sock.h>
 #include <net/inet_frag.h>
 #include <net/inet_ecn.h>
+#include <net/ip.h>
+#include <net/ipv6.h>
+
+/* Use skb->cb to track consecutive/adjacent fragments coming at
+ * the end of the queue. Nodes in the rb-tree queue will
+ * contain "runs" of one or more adjacent fragments.
+ *
+ * Invariants:
+ * - next_frag is NULL at the tail of a "run";
+ * - the head of a "run" has the sum of all fragment lengths in frag_run_len.
+ */
+struct ipfrag_skb_cb {
+	union {
+		struct inet_skb_parm	h4;
+		struct inet6_skb_parm	h6;
+	};
+	struct sk_buff		*next_frag;
+	int			frag_run_len;
+};
+
+#define FRAG_CB(skb)		((struct ipfrag_skb_cb *)((skb)->cb))
+
+static void fragcb_clear(struct sk_buff *skb)
+{
+	RB_CLEAR_NODE(&skb->rbnode);
+	FRAG_CB(skb)->next_frag = NULL;
+	FRAG_CB(skb)->frag_run_len = skb->len;
+}
+
+/* Append skb to the last "run". */
+static void fragrun_append_to_last(struct inet_frag_queue *q,
+				   struct sk_buff *skb)
+{
+	fragcb_clear(skb);
+
+	FRAG_CB(q->last_run_head)->frag_run_len += skb->len;
+	FRAG_CB(q->fragments_tail)->next_frag = skb;
+	q->fragments_tail = skb;
+}
+
+/* Create a new "run" with the skb. */
+static void fragrun_create(struct inet_frag_queue *q, struct sk_buff *skb)
+{
+	BUILD_BUG_ON(sizeof(struct ipfrag_skb_cb) > sizeof(skb->cb));
+	fragcb_clear(skb);
+
+	if (q->last_run_head)
+		rb_link_node(&skb->rbnode, &q->last_run_head->rbnode,
+			     &q->last_run_head->rbnode.rb_right);
+	else
+		rb_link_node(&skb->rbnode, NULL, &q->rb_fragments.rb_node);
+	rb_insert_color(&skb->rbnode, &q->rb_fragments);
+
+	q->fragments_tail = skb;
+	q->last_run_head = skb;
+}
 
 /* Given the OR values of all fragments, apply RFC 3168 5.3 requirements
  * Value : 0xff if frame should be dropped.
@@ -123,6 +179,28 @@ static void inet_frag_destroy_rcu(struct rcu_head *head)
 	kmem_cache_free(f->frags_cachep, q);
 }
 
+unsigned int inet_frag_rbtree_purge(struct rb_root *root)
+{
+	struct rb_node *p = rb_first(root);
+	unsigned int sum = 0;
+
+	while (p) {
+		struct sk_buff *skb = rb_entry(p, struct sk_buff, rbnode);
+
+		p = rb_next(p);
+		rb_erase(&skb->rbnode, root);
+		while (skb) {
+			struct sk_buff *next = FRAG_CB(skb)->next_frag;
+
+			sum += skb->truesize;
+			kfree_skb(skb);
+			skb = next;
+		}
+	}
+	return sum;
+}
+EXPORT_SYMBOL(inet_frag_rbtree_purge);
+
 void inet_frag_destroy(struct inet_frag_queue *q)
 {
 	struct sk_buff *fp;
@@ -224,3 +302,218 @@ struct inet_frag_queue *inet_frag_find(struct netns_frags *nf, void *key)
 	return fq;
 }
 EXPORT_SYMBOL(inet_frag_find);
+
+int inet_frag_queue_insert(struct inet_frag_queue *q, struct sk_buff *skb,
+			   int offset, int end)
+{
+	struct sk_buff *last = q->fragments_tail;
+
+	/* RFC5722, Section 4, amended by Errata ID : 3089
+	 *                          When reassembling an IPv6 datagram, if
+	 *   one or more its constituent fragments is determined to be an
+	 *   overlapping fragment, the entire datagram (and any constituent
+	 *   fragments) MUST be silently discarded.
+	 *
+	 * Duplicates, however, should be ignored (i.e. skb dropped, but the
+	 * queue/fragments kept for later reassembly).
+	 */
+	if (!last)
+		fragrun_create(q, skb);  /* First fragment. */
+	else if (last->ip_defrag_offset + last->len < end) {
+		/* This is the common case: skb goes to the end. */
+		/* Detect and discard overlaps. */
+		if (offset < last->ip_defrag_offset + last->len)
+			return IPFRAG_OVERLAP;
+		if (offset == last->ip_defrag_offset + last->len)
+			fragrun_append_to_last(q, skb);
+		else
+			fragrun_create(q, skb);
+	} else {
+		/* Binary search. Note that skb can become the first fragment,
+		 * but not the last (covered above).
+		 */
+		struct rb_node **rbn, *parent;
+
+		rbn = &q->rb_fragments.rb_node;
+		do {
+			struct sk_buff *curr;
+			int curr_run_end;
+
+			parent = *rbn;
+			curr = rb_to_skb(parent);
+			curr_run_end = curr->ip_defrag_offset +
+					FRAG_CB(curr)->frag_run_len;
+			if (end <= curr->ip_defrag_offset)
+				rbn = &parent->rb_left;
+			else if (offset >= curr_run_end)
+				rbn = &parent->rb_right;
+			else if (offset >= curr->ip_defrag_offset &&
+				 end <= curr_run_end)
+				return IPFRAG_DUP;
+			else
+				return IPFRAG_OVERLAP;
+		} while (*rbn);
+		/* Here we have parent properly set, and rbn pointing to
+		 * one of its NULL left/right children. Insert skb.
+		 */
+		fragcb_clear(skb);
+		rb_link_node(&skb->rbnode, parent, rbn);
+		rb_insert_color(&skb->rbnode, &q->rb_fragments);
+	}
+
+	skb->ip_defrag_offset = offset;
+
+	return IPFRAG_OK;
+}
+EXPORT_SYMBOL(inet_frag_queue_insert);
+
+void *inet_frag_reasm_prepare(struct inet_frag_queue *q, struct sk_buff *skb,
+			      struct sk_buff *parent)
+{
+	struct sk_buff *fp, *head = skb_rb_first(&q->rb_fragments);
+	struct sk_buff **nextp;
+	int delta;
+
+	if (head != skb) {
+		fp = skb_clone(skb, GFP_ATOMIC);
+		if (!fp)
+			return NULL;
+		FRAG_CB(fp)->next_frag = FRAG_CB(skb)->next_frag;
+		if (RB_EMPTY_NODE(&skb->rbnode))
+			FRAG_CB(parent)->next_frag = fp;
+		else
+			rb_replace_node(&skb->rbnode, &fp->rbnode,
+					&q->rb_fragments);
+		if (q->fragments_tail == skb)
+			q->fragments_tail = fp;
+		skb_morph(skb, head);
+		FRAG_CB(skb)->next_frag = FRAG_CB(head)->next_frag;
+		rb_replace_node(&head->rbnode, &skb->rbnode,
+				&q->rb_fragments);
+		consume_skb(head);
+		head = skb;
+	}
+	WARN_ON(head->ip_defrag_offset != 0);
+
+	delta = -head->truesize;
+
+	/* Head of list must not be cloned. */
+	if (skb_unclone(head, GFP_ATOMIC))
+		return NULL;
+
+	delta += head->truesize;
+	if (delta)
+		add_frag_mem_limit(q->net, delta);
+
+	/* 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_has_frag_list(head)) {
+		struct sk_buff *clone;
+		int i, plen = 0;
+
+		clone = alloc_skb(0, GFP_ATOMIC);
+		if (!clone)
+			return NULL;
+		skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
+		skb_frag_list_init(head);
+		for (i = 0; i < skb_shinfo(head)->nr_frags; i++)
+			plen += skb_frag_size(&skb_shinfo(head)->frags[i]);
+		clone->data_len = head->data_len - plen;
+		clone->len = clone->data_len;
+		head->truesize += clone->truesize;
+		clone->csum = 0;
+		clone->ip_summed = head->ip_summed;
+		add_frag_mem_limit(q->net, clone->truesize);
+		skb_shinfo(head)->frag_list = clone;
+		nextp = &clone->next;
+	} else {
+		nextp = &skb_shinfo(head)->frag_list;
+	}
+
+	return nextp;
+}
+EXPORT_SYMBOL(inet_frag_reasm_prepare);
+
+void inet_frag_reasm_finish(struct inet_frag_queue *q, struct sk_buff *head,
+			    void *reasm_data)
+{
+	struct sk_buff **nextp = (struct sk_buff **)reasm_data;
+	struct rb_node *rbn;
+	struct sk_buff *fp;
+
+	skb_push(head, head->data - skb_network_header(head));
+
+	/* Traverse the tree in order, to build frag_list. */
+	fp = FRAG_CB(head)->next_frag;
+	rbn = rb_next(&head->rbnode);
+	rb_erase(&head->rbnode, &q->rb_fragments);
+	while (rbn || fp) {
+		/* fp points to the next sk_buff in the current run;
+		 * rbn points to the next run.
+		 */
+		/* Go through the current run. */
+		while (fp) {
+			*nextp = fp;
+			nextp = &fp->next;
+			fp->prev = NULL;
+			memset(&fp->rbnode, 0, sizeof(fp->rbnode));
+			fp->sk = NULL;
+			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_COMPLETE)
+				head->csum = csum_add(head->csum, fp->csum);
+			head->truesize += fp->truesize;
+			fp = FRAG_CB(fp)->next_frag;
+		}
+		/* Move to the next run. */
+		if (rbn) {
+			struct rb_node *rbnext = rb_next(rbn);
+
+			fp = rb_to_skb(rbn);
+			rb_erase(rbn, &q->rb_fragments);
+			rbn = rbnext;
+		}
+	}
+	sub_frag_mem_limit(q->net, head->truesize);
+
+	*nextp = NULL;
+	skb_mark_not_on_list(head);
+	head->prev = NULL;
+	head->tstamp = q->stamp;
+}
+EXPORT_SYMBOL(inet_frag_reasm_finish);
+
+struct sk_buff *inet_frag_pull_head(struct inet_frag_queue *q)
+{
+	struct sk_buff *head;
+
+	if (q->fragments) {
+		head = q->fragments;
+		q->fragments = head->next;
+	} else {
+		struct sk_buff *skb;
+
+		head = skb_rb_first(&q->rb_fragments);
+		if (!head)
+			return NULL;
+		skb = FRAG_CB(head)->next_frag;
+		if (skb)
+			rb_replace_node(&head->rbnode, &skb->rbnode,
+					&q->rb_fragments);
+		else
+			rb_erase(&head->rbnode, &q->rb_fragments);
+		memset(&head->rbnode, 0, sizeof(head->rbnode));
+		barrier();
+	}
+	if (head == q->fragments_tail)
+		q->fragments_tail = NULL;
+
+	sub_frag_mem_limit(q->net, head->truesize);
+
+	return head;
+}
+EXPORT_SYMBOL(inet_frag_pull_head);

tools/testing/selftests/net/ip_defrag.c

@@ -20,6 +20,7 @@ static bool		cfg_do_ipv4;
 static bool		cfg_do_ipv6;
 static bool		cfg_verbose;
 static bool		cfg_overlap;
+static bool		cfg_permissive;
 static unsigned short	cfg_port = 9000;
 
 const struct in_addr addr4 = { .s_addr = __constant_htonl(INADDR_LOOPBACK + 2) };
@@ -35,7 +36,7 @@ const struct in6_addr addr6 = IN6ADDR_LOOPBACK_INIT;
 static int payload_len;
 static int max_frag_len;
 
-#define MSG_LEN_MAX	60000	/* Max UDP payload length. */
+#define MSG_LEN_MAX	10000	/* Max UDP payload length. */
 
 #define IP4_MF		(1u << 13)  /* IPv4 MF flag. */
 #define IP6_MF		(1)  /* IPv6 MF flag. */
@@ -59,13 +60,14 @@ static void recv_validate_udp(int fd_udp)
 	msg_counter++;
 
 	if (cfg_overlap) {
+		if (ret == -1 && (errno == ETIMEDOUT || errno == EAGAIN))
+			return;  /* OK */
+		if (!cfg_permissive) {
 			if (ret != -1)
 				error(1, 0, "recv: expected timeout; got %d",
 					(int)ret);
-		if (errno != ETIMEDOUT && errno != EAGAIN)
-			error(1, errno, "recv: expected timeout: %d",
-				 errno);
-		return;  /* OK */
+			error(1, errno, "recv: expected timeout: %d", errno);
+		}
 	}
 
 	if (ret == -1)
@@ -203,7 +205,6 @@ static void send_udp_frags(int fd_raw, struct sockaddr *addr,
 {
 	struct ip *iphdr = (struct ip *)ip_frame;
 	struct ip6_hdr *ip6hdr = (struct ip6_hdr *)ip_frame;
-	const bool ipv4 = !ipv6;
 	int res;
 	int offset;
 	int frag_len;
@@ -251,7 +252,7 @@ static void send_udp_frags(int fd_raw, struct sockaddr *addr,
 	}
 
 	/* Occasionally test IPv4 "runs" (see net/ipv4/ip_fragment.c) */
-	if (ipv4 && !cfg_overlap && (rand() % 100 < 20) &&
+	if (!cfg_overlap && (rand() % 100 < 20) &&
 			(payload_len > 9 * max_frag_len)) {
 		offset = 6 * max_frag_len;
 		while (offset < (UDP_HLEN + payload_len)) {
@@ -276,41 +277,38 @@ static void send_udp_frags(int fd_raw, struct sockaddr *addr,
 	while (offset < (UDP_HLEN + payload_len)) {
 		send_fragment(fd_raw, addr, alen, offset, ipv6);
 		/* IPv4 ignores duplicates, so randomly send a duplicate. */
-		if (ipv4 && (1 == rand() % 100))
+		if (rand() % 100 == 1)
 			send_fragment(fd_raw, addr, alen, offset, ipv6);
 		offset += 2 * max_frag_len;
 	}
 
 	if (cfg_overlap) {
-		/* Send an extra random fragment. */
-		if (ipv6) {
-			struct ip6_frag *fraghdr = (struct ip6_frag *)(ip_frame + IP6_HLEN);
-			/* sendto() returns EINVAL if offset + frag_len is too small. */
-			offset = rand() % (UDP_HLEN + payload_len - 1);
-			frag_len = max_frag_len + rand() % 256;
-			/* In IPv6 if !!(frag_len % 8), the fragment is dropped. */
-			frag_len &= ~0x7;
-			fraghdr->ip6f_offlg = htons(offset / 8 | IP6_MF);
-			ip6hdr->ip6_plen = htons(frag_len);
-			frag_len += IP6_HLEN;
-		} else {
-			/* In IPv4, duplicates and some fragments completely inside
+		/* Send an extra random fragment.
+		 *
+		 * Duplicates and some fragments completely inside
 		 * previously sent fragments are dropped/ignored. So
 		 * random offset and frag_len can result in a dropped
-			 * fragment instead of a dropped queue/packet. So we
+		 * fragment instead of a dropped queue/packet. Thus we
 		 * hard-code offset and frag_len.
-			 *
-			 * See ade446403bfb ("net: ipv4: do not handle duplicate
-			 * fragments as overlapping").
 		 */
 		if (max_frag_len * 4 < payload_len || max_frag_len < 16) {
-				/* not enough payload to play with random offset and frag_len. */
+			/* not enough payload for random offset and frag_len. */
 			offset = 8;
-				frag_len = IP4_HLEN + UDP_HLEN + max_frag_len;
+			frag_len = UDP_HLEN + max_frag_len;
 		} else {
 			offset = rand() % (payload_len / 2);
 			frag_len = 2 * max_frag_len + 1 + rand() % 256;
 		}
+		if (ipv6) {
+			struct ip6_frag *fraghdr = (struct ip6_frag *)(ip_frame + IP6_HLEN);
+			/* sendto() returns EINVAL if offset + frag_len is too small. */
+			/* In IPv6 if !!(frag_len % 8), the fragment is dropped. */
+			frag_len &= ~0x7;
+			fraghdr->ip6f_offlg = htons(offset / 8 | IP6_MF);
+			ip6hdr->ip6_plen = htons(frag_len);
+			frag_len += IP6_HLEN;
+		} else {
+			frag_len += IP4_HLEN;
 			iphdr->ip_off = htons(offset / 8 | IP4_MF);
 			iphdr->ip_len = htons(frag_len);
 		}
@@ -327,7 +325,7 @@ static void send_udp_frags(int fd_raw, struct sockaddr *addr,
 	while (offset < (UDP_HLEN + payload_len)) {
 		send_fragment(fd_raw, addr, alen, offset, ipv6);
 		/* IPv4 ignores duplicates, so randomly send a duplicate. */
-		if (ipv4 && (1 == rand() % 100))
+		if (rand() % 100 == 1)
 			send_fragment(fd_raw, addr, alen, offset, ipv6);
 		offset += 2 * max_frag_len;
 	}
@@ -342,7 +340,7 @@ static void run_test(struct sockaddr *addr, socklen_t alen, bool ipv6)
 	 */
 	struct timeval tv = { .tv_sec = 1, .tv_usec = 10 };
 	int idx;
-	int min_frag_len = ipv6 ? 1280 : 8;
+	int min_frag_len = 8;
 
 	/* Initialize the payload. */
 	for (idx = 0; idx < MSG_LEN_MAX; ++idx)
@@ -434,7 +432,7 @@ static void parse_opts(int argc, char **argv)
 {
 	int c;
 
-	while ((c = getopt(argc, argv, "46ov")) != -1) {
+	while ((c = getopt(argc, argv, "46opv")) != -1) {
 		switch (c) {
 		case '4':
 			cfg_do_ipv4 = true;
@@ -445,6 +443,9 @@ static void parse_opts(int argc, char **argv)
 		case 'o':
 			cfg_overlap = true;
 			break;
+		case 'p':
+			cfg_permissive = true;
+			break;
 		case 'v':
 			cfg_verbose = true;
 			break;

tools/testing/selftests/net/ip_defrag.sh

@@ -20,6 +20,10 @@ setup() {
 	ip netns exec "${NETNS}" sysctl -w net.ipv6.ip6frag_low_thresh=7000000 >/dev/null 2>&1
 	ip netns exec "${NETNS}" sysctl -w net.ipv6.ip6frag_time=1 >/dev/null 2>&1
 
+	ip netns exec "${NETNS}" sysctl -w net.netfilter.nf_conntrack_frag6_high_thresh=9000000 >/dev/null 2>&1
+	ip netns exec "${NETNS}" sysctl -w net.netfilter.nf_conntrack_frag6_low_thresh=7000000  >/dev/null 2>&1
+	ip netns exec "${NETNS}" sysctl -w net.netfilter.nf_conntrack_frag6_timeout=1 >/dev/null 2>&1
+
 	# DST cache can get full with a lot of frags, with GC not keeping up with the test.
 	ip netns exec "${NETNS}" sysctl -w net.ipv6.route.max_size=65536 >/dev/null 2>&1
 }
@@ -43,4 +47,16 @@ ip netns exec "${NETNS}" ./ip_defrag -6
 echo "ipv6 defrag with overlaps"
 ip netns exec "${NETNS}" ./ip_defrag -6o
 
+# insert an nf_conntrack rule so that the codepath in nf_conntrack_reasm.c taken
+ip netns exec "${NETNS}" ip6tables -A INPUT  -m conntrack --ctstate INVALID -j ACCEPT
+
+echo "ipv6 nf_conntrack defrag"
+ip netns exec "${NETNS}" ./ip_defrag -6
+
+echo "ipv6 nf_conntrack defrag with overlaps"
+# netfilter will drop some invalid packets, so we run the test in
+# permissive mode: i.e. pass the test if the packet is correctly assembled
+# even if we sent an overlap
+ip netns exec "${NETNS}" ./ip_defrag -6op
+
 echo "all tests done"