You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
fix(link): a FRAD pad octet is lost or invented when a whole frame is rebuilt #2716
Found by the Sonnet review of the #2710 fix. Both reproduce identically on main at 57a6c366d, and nothing is logged.
A pad before a direct NLPID is lost.0c21 03 00 cc + IPv4 rebuilds as 0c21 03 cc …, without the pad.
A pad is invented before 0x80.0c21 03 80 000000 0806 + ARP rebuilds as 0c21 03 00 80 ….
RFC 2427 §3 allows the single zero pad octet before any NLPID, "when required". The data model does not record whether a pad was captured, so make() cannot rebuild the frame exactly.
The fix needs a make() parameter for the pad, and #2692 rewrites the makers, so this is blocked until #2692 merges.
#2692 has merged, so make() can take the pad parameter. The #2710 worker is doing it as a second commit stacked on #2710, which goes out as its own PR. wip.
Activity
JarryShaw commentedon Nov 3, 2026
#2692 has merged, so
make()can take the pad parameter. The #2710 worker is doing it as a second commit stacked on #2710, which goes out as its own PR.wip.JarryShaw commentedon Nov 3, 2026
The PR is #2721.