- ICMP Echo and Echo reply packets mangled the ICMP checksum when no
mangling was needed.
- Other ICMP message types may still be wrong since they potentially contain
IP header information that may or may not need IP addresses manipulated
- Increased the ARP table size from 8 to 64 since the ARP RFC (826) suggests
that ARP info from all received ARP packets be gathered as ARP broadcasts
are received. A normal LAN can easily blow over an 8 entry ARP table.
- Fix last_coni to properly reflect cycles
- Make DHCP mode default
- Fix display of SHOW IMP information to be complete
- Create a separate unit to precisely perform once per wall clock second
activities (DHCP timing and retries, and arp aging).
- Properly reference host and network data byte order items in DHCP packets
- Include Ethernet network address type in DHCP client id
- Properly identify DHCP supplied netmask bit length
- Clear DHCP acquired state when a DHCP NAK is received
- Track DHCP lease expiring only when a lease has been issued
- Fix ipv4_inet_ntoa to work on either endian host systems
- Properly call eth_filter with the mac address and broadcast address
avoiding all multicast mode.
- Set a default unique MAC address in an old BBN address OID address block
- Pick an initial DHCP transaction id (XID) that is unique and defined to be
derived from the MAC address which is presumably unique on the LAN
- Enable asynchronous Ethernet operating mode
- Enable renew and rebind times to be directly supplied via DHCP and
default to be derived from the explicitly returned lease time
- Implement exponential interval retry while waiting for a DHCP responses
- Properly send the a DHCP release packet at detach time.
- Broadcast an ARP reply to the LAN with the DHCP supplied IP address
- Move ARP data to be part of the imp_device structure
- Allow static ARP entries
- Add a SET IMP ARP=ddd.ddd.ddd.ddd=XX:XX:XX:XX:XX:XX command to create
static ARP entries
- Define a static ARP entry when a IMP GW address is defined at attach time
This allows wall clock delays (with sim_activate_after) to be as close
as possible right from when instruction execution starts without having
to execute for the seconds it will take for a proper execution rate
to be calibrated.
