On 10/4/26 12:14 PM, Eli Britstein via dev wrote: > RFC 791 permits zero-payload last fragments, but common stacks > including the Linux kernel discard them. Align the reassembly test > with that policy: use valid one-byte fragments, and update the expiry > coverage expectation for duplicate handling. > > Assisted-by: composer-2.5-fast, Cursor > Acked-by: Mike Pattrick <[email protected]> > Signed-off-by: Eli Britstein <[email protected]> > --- Continuation of the LLM review:
This patch replaces the zero-payload test fragments with one-byte fragments and adjusts the coverage expectation. > +dnl Make sure expiry and purge counters reflect the four packets sent. > +dnl The duplicate second fragment is counted as overlap, not a second expiry. > +AT_CHECK([ovs-appctl coverage/show | grep -c "^ipf.*total: 2"], [0], [1 > ]) Could this check the expiry and purge counters individually by name? At this commit, the second identical fragment causes ipf_drop_frag_chain() to delete the first fragment too, so nothing reaches expiry. The check still passes because the purge counter reaches two. After the later duplicate-handling patch, the first fragment survives and expires, so the expected expiry count is one. This assertion does not check that count either. Could the comment also be updated for the final series? The second fragment is then counted as a duplicate rather than an overlap, and that counter is reported by ipf-get-status rather than coverage/show. _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
