Merge branch 'net-two-fixes-for-qdisc_pkt_len_init'

Eric Dumazet says:

====================
net: two fixes for qdisc_pkt_len_init()

Inspired by one syzbot report.

At least one qdisc (fq_codel) depends on qdisc_skb_cb(skb)->pkt_len
having a sane value (not zero)

With the help of af_packet, syzbot was able to fool qdisc_pkt_len_init()
to precisely set qdisc_skb_cb(skb)->pkt_len to zero.

First patch fixes this issue.

Second one (a separate one to help future bisections) adds
more sanity check to SKB_GSO_DODGY users.
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>

Author: Paolo Abeni

net/core/dev.c

@@ -3758,18 +3758,22 @@ static void qdisc_pkt_len_init(struct sk_buff *skb)
 						sizeof(_tcphdr), &_tcphdr);
 			if (likely(th))
 				hdr_len += __tcp_hdrlen(th);
-		} else {
+		} else if (shinfo->gso_type & SKB_GSO_UDP_L4) {
 			struct udphdr _udphdr;
 
 			if (skb_header_pointer(skb, hdr_len,
 					       sizeof(_udphdr), &_udphdr))
 				hdr_len += sizeof(struct udphdr);
 		}
 
-		if (shinfo->gso_type & SKB_GSO_DODGY)
-			gso_segs = DIV_ROUND_UP(skb->len - hdr_len,
-						shinfo->gso_size);
+		if (unlikely(shinfo->gso_type & SKB_GSO_DODGY)) {
+			int payload = skb->len - hdr_len;
 
+			/* Malicious packet. */
+			if (payload <= 0)
+				return;
+			gso_segs = DIV_ROUND_UP(payload, shinfo->gso_size);
+		}
 		qdisc_skb_cb(skb)->pkt_len += (gso_segs - 1) * hdr_len;
 	}
 }