[PATCH net-next 1/9] net: vlan: don't adjust MAC header in __vlan_insert_inner_tag() unless set

[Vladimir Oltean]

This is a preparatory change for the deletion of skb_reset_mac_header(skb)
from __dev_queue_xmit(). After that deletion, skb_mac_header(skb) will
no longer be set in TX paths, from which __vlan_insert_inner_tag() can
still be called (perhaps indirectly).

If we don't make this change, then an unset MAC header (equal to ~0U)
will become set after the adjustment with VLAN_HLEN.

Signed-off-by: Vladimir Oltean <vladimir.oltean@xxxxxxx>
---
 include/linux/if_vlan.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/include/linux/if_vlan.h b/include/linux/if_vlan.h
index 6864b89ef868..90b76d63c11c 100644
--- a/include/linux/if_vlan.h
+++ b/include/linux/if_vlan.h
@@ -351,7 +351,8 @@ static inline int __vlan_insert_inner_tag(struct sk_buff *skb,
 	/* Move the mac header sans proto to the beginning of the new header. */
 	if (likely(mac_len > ETH_TLEN))
 		memmove(skb->data, skb->data + VLAN_HLEN, mac_len - ETH_TLEN);
-	skb->mac_header -= VLAN_HLEN;
+	if (skb_mac_header_was_set(skb))
+		skb->mac_header -= VLAN_HLEN;
 
 	veth = (struct vlan_ethhdr *)(skb->data + mac_len - ETH_HLEN);
 
-- 
2.34.1