This is caused by commit ade446403bfb ("net: ipv4: do not handle duplicate fragments as overlapping"). I have a patch that fixes the test - waiting for net-next to reopen. Or maybe I should send it to net?

On Fri, Jan 4, 2019 at 9:10 AM Dan Rue dan.rue@linaro.org wrote:

Hi! We've noticed selftests/net/ip_defrag.sh started failing on mainline between v4.20-rc7-23-gab63e725b49c (good) and v4.20-rc7-202-g1d51b4b1d3f2 (bad). Output looks like:

selftests: net: ip_defrag.sh
ipv4 defrag
PASS
seed = 1545325420
ipv4 defrag with overlaps
seed = 1545325421
./ip_defrag: recv: expected timeout; got 2272
not ok 1..16 selftests: net: ip_defrag.sh [FAIL]
selftests: net_ip_defrag.sh [FAIL]

I see a lot of net/ changes in the diff - we can bisect it if necessary but wanted to see if anyone else had already investigated.

Thanks, Dan

-- Linaro - Kernel Validation