On Tue, Oct 31, 2023 at 10:22:05PM +0100, Linus Walleij wrote:

Hi Vladimir,

I got around to testing this too:

# ping -s 1472 192.168.1.137

The result:

SKB before padding:
37 (192.168.1.13skb len=1514 headroom=18 headlen=1514 tailroom=260
mac=(18,14) net=(32,20) trans=52
shinfo(txflags=0 nr_frags=0 gso(size=0 type=0 segs=0))
csum(0xd4ef2b1 ip_summed=0 complete_sw=0 valid=0 level=0)
hash(0x0 sw=0 l4=0) proto=0x0800 pkttype=0 iif=0
7): 1472 data bydev name=lan0 feat=0x0002000000005020
tes
sk family=2 type=3 proto=1

SKB after padding:
skb len=1518 headroom=18 headlen=1518 tailroom=256
mac=(18,14) net=(32,20) trans=52
shinfo(txflags=0 nr_frags=0 gso(size=0 type=0 segs=0))
csum(0xd4ef2b1 ip_summed=0 complete_sw=0 valid=0 level=0)
hash(0x0 sw=0 l4=0) proto=0x0800 pkttype=0 iif=0
dev name=lan0 feat=0x0002000000005020
sk family=2 type=3 proto=1

As expected the linear SKB is 4 bytes longer in this case.

Ok, I'm not seeing anything unusual about the skb geometry in the
skb_dump() output.