Hi Simon,
another thought ... we're concentrating on the PPP routines now, but it's
confirmed that pppifOutput is called only 2 times (to send the first 2 1024
bytes segments).
But if pppifOutput itself is called only 2 times from the tcp_out ... isn't
that more a thing to concentrate on the TCP side? We're out of the
responsibility of the PPP routines at this point.
Is there any internal "time-to-live" for a fragment which might kick in if the
output won't happen within a specific time?
Inserting this debug output in tcp_out.c after line 920:
wnd = LWIP_MIN(pcb->snd_wnd, pcb->cwnd);
for (seg = pcb->unsent; seg != NULL; seg = seg->next, i++)
{
LWIP_PLATFORM_DIAG (("unsentq Seg:%hu Len:%hu\n", i, seg->len));
}
i = 0;
gives me this fir the 2x1400 bytes packets to be sent:
unsentq Seg:0 Len:1024<LF>
unsentq Seg:1 Len:376<LF>
unsentq Seg:0 Len:1024<LF>
At least 376 bytes are still missing at this point. The final wireshark output
gives me 2 packets with 1024 bytes each, because it's "naggled" later ...
_______________________________________________
lwip-users mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/lwip-users
