net/mlx5e: fix csum adjustments caused by RXFCS

As shown by Dmitris, we need to use csum_block_add() instead of csum_add()
when adding the FCS contribution to skb csum.

Before 4.18 (more exactly commit 88078d98 "net: pskb_trim_rcsum()
and CHECKSUM_COMPLETE are friends"), the whole skb csum was thrown away,
so RXFCS changes were ignored.

Then before commit d55bef50 ("net: fix pskb_trim_rcsum_slow() with
odd trim offset") both mlx5 and pskb_trim_rcsum_slow() bugs were canceling
each other.

Now we fixed pskb_trim_rcsum_slow() we need to fix mlx5.

Note that this patch also rewrites mlx5e_get_fcs() to :

- Use skb_header_pointer() instead of reinventing it.
- Use __get_unaligned_cpu32() to avoid possible non aligned accesses
  as Dmitris pointed out.

Fixes: 902a5459 ("net/mlx5e: When RXFCS is set, add FCS data into checksum calculation")
Reported-by: Paweł Staszewski <pstaszewski@itcare.pl>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Eran Ben Elisha <eranbe@mellanox.com>
Cc: Saeed Mahameed <saeedm@mellanox.com>
Cc: Dimitris Michailidis <dmichail@google.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Paweł Staszewski <pstaszewski@itcare.pl>
Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com>
Tested-By: Maria Pasechnik <mariap@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/mellanox/mlx5/core/en_rx.c

@@ -713,43 +713,15 @@ static inline void mlx5e_enable_ecn(struct mlx5e_rq *rq, struct sk_buff *skb)
 	rq->stats->ecn_mark += !!rc;
 }
 
-static __be32 mlx5e_get_fcs(struct sk_buff *skb)
+static u32 mlx5e_get_fcs(const struct sk_buff *skb)
 {
-	int last_frag_sz, bytes_in_prev, nr_frags;
-	u8 *fcs_p1, *fcs_p2;
-	skb_frag_t *last_frag;
-	__be32 fcs_bytes;
+	const void *fcs_bytes;
+	u32 _fcs_bytes;
 
-	if (!skb_is_nonlinear(skb))
-		return *(__be32 *)(skb->data + skb->len - ETH_FCS_LEN);
+	fcs_bytes = skb_header_pointer(skb, skb->len - ETH_FCS_LEN,
+				       ETH_FCS_LEN, &_fcs_bytes);
 
-	nr_frags = skb_shinfo(skb)->nr_frags;
-	last_frag = &skb_shinfo(skb)->frags[nr_frags - 1];
-	last_frag_sz = skb_frag_size(last_frag);
-
-	/* If all FCS data is in last frag */
-	if (last_frag_sz >= ETH_FCS_LEN)
-		return *(__be32 *)(skb_frag_address(last_frag) +
-				   last_frag_sz - ETH_FCS_LEN);
-
-	fcs_p2 = (u8 *)skb_frag_address(last_frag);
-	bytes_in_prev = ETH_FCS_LEN - last_frag_sz;
-
-	/* Find where the other part of the FCS is - Linear or another frag */
-	if (nr_frags == 1) {
-		fcs_p1 = skb_tail_pointer(skb);
-	} else {
-		skb_frag_t *prev_frag = &skb_shinfo(skb)->frags[nr_frags - 2];
-
-		fcs_p1 = skb_frag_address(prev_frag) +
-			    skb_frag_size(prev_frag);
-	}
-	fcs_p1 -= bytes_in_prev;
-
-	memcpy(&fcs_bytes, fcs_p1, bytes_in_prev);
-	memcpy(((u8 *)&fcs_bytes) + bytes_in_prev, fcs_p2, last_frag_sz);
-
-	return fcs_bytes;
+	return __get_unaligned_cpu32(fcs_bytes);
 }
 
 static u8 get_ip_proto(struct sk_buff *skb, __be16 proto)
@@ -797,8 +769,9 @@ static inline void mlx5e_handle_csum(struct net_device *netdev,
 						 network_depth - ETH_HLEN,
 						 skb->csum);
 		if (unlikely(netdev->features & NETIF_F_RXFCS))
-			skb->csum = csum_add(skb->csum,
-					     (__force __wsum)mlx5e_get_fcs(skb));
+			skb->csum = csum_block_add(skb->csum,
+						   (__force __wsum)mlx5e_get_fcs(skb),
+						   skb->len - ETH_FCS_LEN);
 		stats->csum_complete++;
 		return;
 	}