net: mctp: unshare packets when reassembling

Ensure that the frag_list used for reassembly isn't shared with other
packets. This avoids incorrect reassembly when packets are cloned, and
prevents a memory leak due to circular references between fragments and
their skb_shared_info.

The upcoming MCTP-over-USB driver uses skb_clone which can trigger the
problem - other MCTP drivers don't share SKBs.

A kunit test is added to reproduce the issue.

Signed-off-by: Matt Johnston <[email protected]>
Fixes: 4a992bb ("mctp: Implement message fragmentation & reassembly")
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>

Author: mkj

net/mctp/route.c

@@ -332,8 +332,14 @@ static int mctp_frag_queue(struct mctp_sk_key *key, struct sk_buff *skb)
 		& MCTP_HDR_SEQ_MASK;
 
 	if (!key->reasm_head) {
-		key->reasm_head = skb;
-		key->reasm_tailp = &(skb_shinfo(skb)->frag_list);
+		/* Since we're manipulating the shared frag_list, ensure it isn't
+		 * shared with any other SKBs.
+		 */
+		key->reasm_head = skb_unshare(skb, GFP_ATOMIC);
+		if (!key->reasm_head)
+			return -ENOMEM;
+
+		key->reasm_tailp = &(skb_shinfo(key->reasm_head)->frag_list);
 		key->last_seq = this_seq;
 		return 0;
 	}

net/mctp/test/route-test.c

@@ -921,6 +921,114 @@ static void mctp_test_route_input_sk_fail_frag(struct kunit *test)
 	__mctp_route_test_fini(test, dev, rt, sock);
 }
 
+/* Input route to socket, using a fragmented message created from clones.
+ */
+static void mctp_test_route_input_cloned_frag(struct kunit *test)
+{
+	/* 5 packet fragments, forming 2 complete messages */
+	const struct mctp_hdr hdrs[5] = {
+		RX_FRAG(FL_S, 0),
+		RX_FRAG(0, 1),
+		RX_FRAG(FL_E, 2),
+		RX_FRAG(FL_S, 0),
+		RX_FRAG(FL_E, 1),
+	};
+	struct mctp_test_route *rt;
+	struct mctp_test_dev *dev;
+	struct sk_buff *skb[5];
+	struct sk_buff *rx_skb;
+	struct socket *sock;
+	size_t data_len;
+	u8 compare[100];
+	u8 flat[100];
+	size_t total;
+	void *p;
+	int rc;
+
+	/* Arbitrary length */
+	data_len = 3;
+	total = data_len + sizeof(struct mctp_hdr);
+
+	__mctp_route_test_init(test, &dev, &rt, &sock, MCTP_NET_ANY);
+
+	/* Create a single skb initially with concatenated packets */
+	skb[0] = mctp_test_create_skb(&hdrs[0], 5 * total);
+	mctp_test_skb_set_dev(skb[0], dev);
+	memset(skb[0]->data, 0 * 0x11, skb[0]->len);
+	memcpy(skb[0]->data, &hdrs[0], sizeof(struct mctp_hdr));
+
+	/* Extract and populate packets */
+	for (int i = 1; i < 5; i++) {
+		skb[i] = skb_clone(skb[i - 1], GFP_ATOMIC);
+		KUNIT_ASSERT_TRUE(test, skb[i]);
+		p = skb_pull(skb[i], total);
+		KUNIT_ASSERT_TRUE(test, p);
+		skb_reset_network_header(skb[i]);
+		memcpy(skb[i]->data, &hdrs[i], sizeof(struct mctp_hdr));
+		memset(&skb[i]->data[sizeof(struct mctp_hdr)], i * 0x11, data_len);
+	}
+	for (int i = 0; i < 5; i++)
+		skb_trim(skb[i], total);
+
+	/* SOM packets have a type byte to match the socket */
+	skb[0]->data[4] = 0;
+	skb[3]->data[4] = 0;
+
+	skb_dump("pkt1 ", skb[0], false);
+	skb_dump("pkt2 ", skb[1], false);
+	skb_dump("pkt3 ", skb[2], false);
+	skb_dump("pkt4 ", skb[3], false);
+	skb_dump("pkt5 ", skb[4], false);
+
+	for (int i = 0; i < 5; i++) {
+		KUNIT_EXPECT_EQ(test, refcount_read(&skb[i]->users), 1);
+		/* Take a reference so we can check refcounts at the end */
+		skb_get(skb[i]);
+	}
+
+	/* Feed the fragments into MCTP core */
+	for (int i = 0; i < 5; i++) {
+		rc = mctp_route_input(&rt->rt, skb[i]);
+		KUNIT_EXPECT_EQ(test, rc, 0);
+	}
+
+	/* Receive first reassembled message */
+	rx_skb = skb_recv_datagram(sock->sk, MSG_DONTWAIT, &rc);
+	KUNIT_EXPECT_EQ(test, rc, 0);
+	KUNIT_EXPECT_EQ(test, rx_skb->len, 3 * data_len);
+	rc = skb_copy_bits(rx_skb, 0, flat, rx_skb->len);
+	for (int i = 0; i < rx_skb->len; i++)
+		compare[i] = (i / data_len) * 0x11;
+	/* Set type byte */
+	compare[0] = 0;
+
+	KUNIT_EXPECT_MEMEQ(test, flat, compare, rx_skb->len);
+	KUNIT_EXPECT_EQ(test, refcount_read(&rx_skb->users), 1);
+	kfree_skb(rx_skb);
+
+	/* Receive second reassembled message */
+	rx_skb = skb_recv_datagram(sock->sk, MSG_DONTWAIT, &rc);
+	KUNIT_EXPECT_EQ(test, rc, 0);
+	KUNIT_EXPECT_EQ(test, rx_skb->len, 2 * data_len);
+	rc = skb_copy_bits(rx_skb, 0, flat, rx_skb->len);
+	for (int i = 0; i < rx_skb->len; i++)
+		compare[i] = (i / data_len + 3) * 0x11;
+	/* Set type byte */
+	compare[0] = 0;
+
+	KUNIT_EXPECT_MEMEQ(test, flat, compare, rx_skb->len);
+	KUNIT_EXPECT_EQ(test, refcount_read(&rx_skb->users), 1);
+	kfree_skb(rx_skb);
+
+	/* Check input skb refcounts */
+	for (int i = 0; i < 5; i++) {
+		KUNIT_EXPECT_EQ(test, refcount_read(&skb[i]->users), 1);
+		kfree_skb(skb[i]);
+	}
+
+	__mctp_route_test_fini(test, dev, rt, sock);
+}
+
 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
 
 static void mctp_test_flow_init(struct kunit *test,
@@ -1144,6 +1252,7 @@ static struct kunit_case mctp_test_cases[] = {
 	KUNIT_CASE(mctp_test_packet_flow),
 	KUNIT_CASE(mctp_test_fragment_flow),
 	KUNIT_CASE(mctp_test_route_output_key_create),
+	KUNIT_CASE(mctp_test_route_input_cloned_frag),
 	{}
 };