wenxu <we...@ucloud.cn> wrote: > > 在 2020/7/6 22:38, Florian Westphal 写道: > > we...@ucloud.cn <we...@ucloud.cn> wrote: > >> From: wenxu <we...@ucloud.cn> > >> > >> Add nf_ct_frag_gather for conntrack defrag and it will > >> elide the CB clear when packets are defragmented by > >> connection tracking > > Why is this patch required? > > Can't you rework ip_defrag to avoid the cb clear if you need that? > > The ip_defrag used by ip stack and can work with the cb setting.
Yes, but does it have to? If yes, why does nf_ct_frag not need it whereas ip_defrag has to? > Defragment case only for conntrack maybe need to avoid the cb > > clear. So it is more clear to nf_ct_frag_gather for conntrack like the > > function nf_ct_frag6_gather for ipv6. nf_ct_frag6_gather() is only re-using less code from ipv6 for historical reasons. If anything, ipv6 conntrack defrag should re-use more code from ipv6 defrag, rather than making ipv4 conntrack defrag look more like ipv6 conntrack defrag.
