Fix L3 timestamps

[gpotter2]

Credits for the timestamp bug to Ron Henderson

fix #2423
fix #2425
fix #2277
fix #2429

[codecov]

Codecov Report

Merging #2426 into master will decrease coverage by 0.1%.
The diff coverage is 50%.

@@            Coverage Diff             @@
##           master    #2426      +/-   ##
==========================================
- Coverage   87.24%   87.14%   -0.11%     
==========================================
  Files         243      242       -1     
  Lines       50760    49861     -899     
==========================================
- Hits        44286    43451     -835     
+ Misses       6474     6410      -64

Impacted Files	Coverage Δ
scapy/arch/bpf/supersocket.py	27.87% <0%> (-47.58%)	⬇️
scapy/arch/common.py	73.43% <0%> (-15.96%)	⬇️
scapy/data.py	92.51% <100%> (ø)	⬆️
scapy/arch/linux.py	76.44% <100%> (-0.45%)	⬇️
scapy/supersocket.py	74.17% <50%> (-1.08%)	⬇️
scapy/arch/pcapdnet.py	65.78% <80%> (-3.78%)	⬇️
scapy/layers/dhcp6.py	64.18% <0%> (-21.58%)	⬇️
scapy/arch/bpf/core.py	81.37% <0%> (-5.89%)	⬇️
scapy/layers/tls/automaton_srv.py	73.18% <0%> (-4.98%)	⬇️
... and 78 more

[guedou]

Good catch!

[p-l-]

Great work @gpotter2. I have two minor (but I think useful) suggestions.

[p-l-]

What do you mean?

[gpotter2]

See https://github.com/torvalds/linux/blob/a2d79c7174aeb43b13020dd53d85a7aefdd9f3e5/include/uapi/asm-generic/socket.h#L109-L115

• SO_TIMESTAMPNS_OLD = 35 returns timespecs that use long to store the date = not 2038 safe
• SO_TIMESTAMPNS_NEW returns timespecs that are long long
• SO_TIMESTAMPNS defaults for the old one

We would use the "NEW" one but it's very recent and isn't supported well yet.
https://lwn.net/Articles/776219/

[gpotter2]

Should be ready to merge