sk_buff: avoid potentially clearing 'slow_gro' field

If skb_dst_set_noref() is invoked with a NULL dst, the 'slow_gro'
field is cleared, too. That could lead to wrong behavior if
the skb later enters the GRO stage.

Fix the potential issue replacing preserving a non-zero value of
the 'slow_gro' field.

Additionally, fix a comment typo.

Reported-by: Sabrina Dubroca <sd@queasysnail.net>
Reported-by: Jakub Kicinski <kuba@kernel.org>
Fixes: 8a886b1 ("sk_buff: track dst status in slow_gro")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Link: https://lore.kernel.org/r/aa42529252dc8bb02bd42e8629427040d1058537.1627662501.git.pabeni@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

include/linux/skbuff.h

@@ -1009,7 +1009,7 @@ static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
 {
 	WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
-	skb->slow_gro = !!dst;
+	skb->slow_gro |= !!dst;
 	skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
 }
 

net/core/dev.c

@@ -6022,7 +6022,7 @@ static void gro_list_prepare(const struct list_head *head,
 				       skb_mac_header(skb),
 				       maclen);
 
-		/* in most common scenarions _state is 0
+		/* in most common scenarions 'slow_gro' is 0
 		 * otherwise we are already on some slower paths
 		 * either skip all the infrequent tests altogether or
 		 * avoid trying too hard to skip each of them individually