net: Fix possible wrong checksum generation.

Patch cef401de7be8c4e (net: fix possible wrong checksum
generation) fixed wrong checksum calculation but it broke TSO by
defining new GSO type but not a netdev feature for that type.
net_gso_ok() would not allow hardware checksum/segmentation
offload of such packets without the feature.

Following patch fixes TSO and wrong checksum. This patch uses
same logic that Eric Dumazet used. Patch introduces new flag
SKBTX_SHARED_FRAG if at least one frag can be modified by
the user. but SKBTX_SHARED_FRAG flag is kept in skb shared
info tx_flags rather than gso_type.

tx_flags is better compared to gso_type since we can have skb with
shared frag without gso packet. It does not link SHARED_FRAG to
GSO, So there is no need to define netdev feature for this.

Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/macvtap.c

@@ -543,7 +543,6 @@ static int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *from,
 		skb->data_len += len;
 		skb->len += len;
 		skb->truesize += truesize;
-		skb_shinfo(skb)->gso_type |= SKB_GSO_SHARED_FRAG;
 		atomic_add(truesize, &skb->sk->sk_wmem_alloc);
 		while (len) {
 			int off = base & ~PAGE_MASK;
@@ -599,7 +598,7 @@ static int macvtap_skb_from_vnet_hdr(struct sk_buff *skb,
 
 	if (vnet_hdr->gso_type != VIRTIO_NET_HDR_GSO_NONE) {
 		skb_shinfo(skb)->gso_size = vnet_hdr->gso_size;
-		skb_shinfo(skb)->gso_type |= gso_type;
+		skb_shinfo(skb)->gso_type = gso_type;
 
 		/* Header must be checked, and gso_segs computed. */
 		skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
@@ -743,6 +742,7 @@ static ssize_t macvtap_get_user(struct macvtap_queue *q, struct msghdr *m,
 	if (zerocopy) {
 		skb_shinfo(skb)->destructor_arg = m->msg_control;
 		skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
+		skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
 	}
 	if (vlan)
 		macvlan_start_xmit(skb, vlan->dev);

drivers/net/tun.c

@@ -1019,7 +1019,6 @@ static int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *from,
 		skb->data_len += len;
 		skb->len += len;
 		skb->truesize += truesize;
-		skb_shinfo(skb)->gso_type |= SKB_GSO_SHARED_FRAG;
 		atomic_add(truesize, &skb->sk->sk_wmem_alloc);
 		while (len) {
 			int off = base & ~PAGE_MASK;
@@ -1165,18 +1164,16 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
 	}
 
 	if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
-		unsigned short gso_type = 0;
-
 		pr_debug("GSO!\n");
 		switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
 		case VIRTIO_NET_HDR_GSO_TCPV4:
-			gso_type = SKB_GSO_TCPV4;
+			skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
 			break;
 		case VIRTIO_NET_HDR_GSO_TCPV6:
-			gso_type = SKB_GSO_TCPV6;
+			skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
 			break;
 		case VIRTIO_NET_HDR_GSO_UDP:
-			gso_type = SKB_GSO_UDP;
+			skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
 			break;
 		default:
 			tun->dev->stats.rx_frame_errors++;
@@ -1185,10 +1182,9 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
 		}
 
 		if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
-			gso_type |= SKB_GSO_TCP_ECN;
+			skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
 
 		skb_shinfo(skb)->gso_size = gso.gso_size;
-		skb_shinfo(skb)->gso_type |= gso_type;
 		if (skb_shinfo(skb)->gso_size == 0) {
 			tun->dev->stats.rx_frame_errors++;
 			kfree_skb(skb);
@@ -1204,6 +1200,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
 	if (zerocopy) {
 		skb_shinfo(skb)->destructor_arg = msg_control;
 		skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
+		skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
 	}
 
 	skb_reset_network_header(skb);

drivers/net/virtio_net.c

@@ -227,7 +227,7 @@ static void set_skb_frag(struct sk_buff *skb, struct page *page,
 	skb->len += size;
 	skb->truesize += PAGE_SIZE;
 	skb_shinfo(skb)->nr_frags++;
-	skb_shinfo(skb)->gso_type |= SKB_GSO_SHARED_FRAG;
+	skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
 	*len -= size;
 }
 
@@ -387,18 +387,16 @@ static void receive_buf(struct receive_queue *rq, void *buf, unsigned int len)
 		 ntohs(skb->protocol), skb->len, skb->pkt_type);
 
 	if (hdr->hdr.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
-		unsigned short gso_type = 0;
-
 		pr_debug("GSO!\n");
 		switch (hdr->hdr.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
 		case VIRTIO_NET_HDR_GSO_TCPV4:
-			gso_type = SKB_GSO_TCPV4;
+			skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
 			break;
 		case VIRTIO_NET_HDR_GSO_UDP:
-			gso_type = SKB_GSO_UDP;
+			skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
 			break;
 		case VIRTIO_NET_HDR_GSO_TCPV6:
-			gso_type = SKB_GSO_TCPV6;
+			skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
 			break;
 		default:
 			net_warn_ratelimited("%s: bad gso type %u.\n",
@@ -407,7 +405,7 @@ static void receive_buf(struct receive_queue *rq, void *buf, unsigned int len)
 		}
 
 		if (hdr->hdr.gso_type & VIRTIO_NET_HDR_GSO_ECN)
-			gso_type |= SKB_GSO_TCP_ECN;
+			skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
 
 		skb_shinfo(skb)->gso_size = hdr->hdr.gso_size;
 		if (skb_shinfo(skb)->gso_size == 0) {
@@ -415,7 +413,6 @@ static void receive_buf(struct receive_queue *rq, void *buf, unsigned int len)
 			goto frame_err;
 		}
 
-		skb_shinfo(skb)->gso_type |= gso_type;
 		/* Header must be checked, and gso_segs computed. */
 		skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
 		skb_shinfo(skb)->gso_segs = 0;

include/linux/skbuff.h

@@ -230,6 +230,13 @@ enum {
 
 	/* generate wifi status information (where possible) */
 	SKBTX_WIFI_STATUS = 1 << 4,
+
+	/* This indicates at least one fragment might be overwritten
+	 * (as in vmsplice(), sendfile() ...)
+	 * If we need to compute a TX checksum, we'll need to copy
+	 * all frags to avoid possible bad checksum
+	 */
+	SKBTX_SHARED_FRAG = 1 << 5,
 };
 
 /*
@@ -307,13 +314,6 @@ enum {
 	SKB_GSO_TCPV6 = 1 << 4,
 
 	SKB_GSO_FCOE = 1 << 5,
-
-	/* This indicates at least one fragment might be overwritten
-	 * (as in vmsplice(), sendfile() ...)
-	 * If we need to compute a TX checksum, we'll need to copy
-	 * all frags to avoid possible bad checksum
-	 */
-	SKB_GSO_SHARED_FRAG = 1 << 6,
 };
 
 #if BITS_PER_LONG > 32
@@ -2220,7 +2220,8 @@ static inline int skb_linearize(struct sk_buff *skb)
  */
 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
 {
-	return skb_shinfo(skb)->gso_type & SKB_GSO_SHARED_FRAG;
+	return skb_is_nonlinear(skb) &&
+	       skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
 }
 
 /**

net/core/skbuff.c

@@ -2326,8 +2326,7 @@ void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len)
 {
 	int pos = skb_headlen(skb);
 
-	skb_shinfo(skb1)->gso_type = skb_shinfo(skb)->gso_type;
-
+	skb_shinfo(skb)->tx_flags = skb_shinfo(skb1)->tx_flags & SKBTX_SHARED_FRAG;
 	if (len < pos)	/* Split line is inside header. */
 		skb_split_inside_header(skb, skb1, len, pos);
 	else		/* Second chunk has no header, nothing to copy. */
@@ -2833,7 +2832,7 @@ struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features)
 		skb_copy_from_linear_data_offset(skb, offset,
 						 skb_put(nskb, hsize), hsize);
 
-		skb_shinfo(nskb)->gso_type = skb_shinfo(skb)->gso_type;
+		skb_shinfo(nskb)->tx_flags = skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
 
 		while (pos < offset + len && i < nfrags) {
 			*frag = skb_shinfo(skb)->frags[i];

net/ipv4/af_inet.c

@@ -1287,7 +1287,6 @@ static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
 		       SKB_GSO_UDP |
 		       SKB_GSO_DODGY |
 		       SKB_GSO_TCP_ECN |
-		       SKB_GSO_SHARED_FRAG |
 		       0)))
 		goto out;
 

net/ipv4/ip_output.c

@@ -598,6 +598,7 @@ slow_path:
 	/* for offloaded checksums cleanup checksum before fragmentation */
 	if ((skb->ip_summed == CHECKSUM_PARTIAL) && skb_checksum_help(skb))
 		goto fail;
+	iph = ip_hdr(skb);
 
 	left = skb->len - hlen;		/* Space per frame */
 	ptr = hlen;		/* Where to start from */

net/ipv4/tcp.c

@@ -897,8 +897,7 @@ new_segment:
 			get_page(page);
 			skb_fill_page_desc(skb, i, page, offset, copy);
 		}
-
-		skb_shinfo(skb)->gso_type |= SKB_GSO_SHARED_FRAG;
+		skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
 
 		skb->len += copy;
 		skb->data_len += copy;
@@ -3044,7 +3043,6 @@ struct sk_buff *tcp_tso_segment(struct sk_buff *skb,
 			       SKB_GSO_DODGY |
 			       SKB_GSO_TCP_ECN |
 			       SKB_GSO_TCPV6 |
-			       SKB_GSO_SHARED_FRAG |
 			       0) ||
 			     !(type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6))))
 			goto out;

net/ipv4/tcp_input.c

@@ -1239,13 +1239,13 @@ static bool tcp_shifted_skb(struct sock *sk, struct sk_buff *skb,
 	 */
 	if (!skb_shinfo(prev)->gso_size) {
 		skb_shinfo(prev)->gso_size = mss;
-		skb_shinfo(prev)->gso_type |= sk->sk_gso_type;
+		skb_shinfo(prev)->gso_type = sk->sk_gso_type;
 	}
 
 	/* CHECKME: To clear or not to clear? Mimics normal skb currently */
 	if (skb_shinfo(skb)->gso_segs <= 1) {
 		skb_shinfo(skb)->gso_size = 0;
-		skb_shinfo(skb)->gso_type &= SKB_GSO_SHARED_FRAG;
+		skb_shinfo(skb)->gso_type = 0;
 	}
 
 	/* Difference in this won't matter, both ACKed by the same cumul. ACK */

net/ipv4/tcp_output.c

@@ -1133,7 +1133,6 @@ static void tcp_queue_skb(struct sock *sk, struct sk_buff *skb)
 static void tcp_set_skb_tso_segs(const struct sock *sk, struct sk_buff *skb,
 				 unsigned int mss_now)
 {
-	skb_shinfo(skb)->gso_type &= SKB_GSO_SHARED_FRAG;
 	if (skb->len <= mss_now || !sk_can_gso(sk) ||
 	    skb->ip_summed == CHECKSUM_NONE) {
 		/* Avoid the costly divide in the normal
@@ -1141,10 +1140,11 @@ static void tcp_set_skb_tso_segs(const struct sock *sk, struct sk_buff *skb,
 		 */
 		skb_shinfo(skb)->gso_segs = 1;
 		skb_shinfo(skb)->gso_size = 0;
+		skb_shinfo(skb)->gso_type = 0;
 	} else {
 		skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss_now);
 		skb_shinfo(skb)->gso_size = mss_now;
-		skb_shinfo(skb)->gso_type |= sk->sk_gso_type;
+		skb_shinfo(skb)->gso_type = sk->sk_gso_type;
 	}
 }
 

net/ipv6/ip6_offload.c

@@ -100,7 +100,6 @@ static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
 		       SKB_GSO_DODGY |
 		       SKB_GSO_TCP_ECN |
 		       SKB_GSO_TCPV6 |
-		       SKB_GSO_SHARED_FRAG |
 		       0)))
 		goto out;