Hi Cody,

Hi,

quoted

Just a generic thought - might be worse to move functions that parse ipv6 header extentions
and related strcutures into rte_net.
I am sure they might be reused by some other code.

Sorry, I am misunderstanding. Do you mean it might be better to move
struct ipv6_opt_hdr and ipv6_ext_hdr() into rte_net since they are not
fragmentation specific? That seems fine to me.

Yes, that's was my thought.

quoted

pktmbuf_read() is quite heavy-weight one.
Do we really need it here?
From my perspective - add an assumption that all whole IPv6 header will be inside
one segment seems reasonable enough.

It is my understanding that rte_pktmbuf_read() will almost always just
invoke a light weight rte_pktmbuf_mtod_offset(). It only runs the
heavy weight __rte_pktmbuf_read() in the case that the assumption you
mentioned is broken.

Ah, yes you right.
Konstantin