netdev: bfin_mac: invalid data cache only once for each new rx skb buffer

The skb buffer isn't actually used until we finish transferring and pass
it up to higher layers, so only invalidate the range once before we start
receiving actual data.  This also avoids the problem with data invalidating
on Blackfin systems -- there is no invalidate-only, just invalidate+flush.
So when running in writeback mode, there is the small (but not uncommon)
possibility of the flush overwriting valid DMA-ed data from the cache.

Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/drivers/net/bfin_mac.c b/drivers/net/bfin_mac.c
index 16f2a37..2b364ba 100644
--- a/drivers/net/bfin_mac.c
+++ b/drivers/net/bfin_mac.c
@@ -203,6 +203,11 @@
 			goto init_error;
 		}
 		skb_reserve(new_skb, NET_IP_ALIGN);
+		/* Invidate the data cache of skb->data range when it is write back
+		 * cache. It will prevent overwritting the new data from DMA
+		 */
+		blackfin_dcache_invalidate_range((unsigned long)new_skb->head,
+					 (unsigned long)new_skb->end);
 		r->skb = new_skb;
 
 		/*
@@ -1011,19 +1016,17 @@
 	}
 	/* reserve 2 bytes for RXDWA padding */
 	skb_reserve(new_skb, NET_IP_ALIGN);
-	current_rx_ptr->skb = new_skb;
-	current_rx_ptr->desc_a.start_addr = (unsigned long)new_skb->data - 2;
-
 	/* Invidate the data cache of skb->data range when it is write back
 	 * cache. It will prevent overwritting the new data from DMA
 	 */
 	blackfin_dcache_invalidate_range((unsigned long)new_skb->head,
 					 (unsigned long)new_skb->end);
 
+	current_rx_ptr->skb = new_skb;
+	current_rx_ptr->desc_a.start_addr = (unsigned long)new_skb->data - 2;
+
 	len = (unsigned short)((current_rx_ptr->status.status_word) & RX_FRLEN);
 	skb_put(skb, len);
-	blackfin_dcache_invalidate_range((unsigned long)skb->head,
-					 (unsigned long)skb->tail);
 
 	skb->protocol = eth_type_trans(skb, dev);