Hi James,

The ACL you have created allows all UDP traffic alright, but because there are no other permit lines in it, all other traffic is blocked - TCP, ICMP, raw IP, etc. Cisco ACLs always use the logic of "what is not permitted will be denied". I suppose that it is the other traffic that is denied that causes the resulting UDP communication to fail as well (if, say, there should be a control channel established through another TCP connection that couldn't be set up).

However, I am also surprised at the notion that the switch should be blocking the UDP communication on a specific port. Your 2950 has absolutely no reason to do that. There might be something else in play here. As you are dealing with an embedded software that's doing networking, there is a possibility that some of the lower layer operations are not correct, and the 2950 being a solid switch drops frames that in some way exhibit peculiar properties.

Can you please provide the following details?

• What exact source and destination IP address is used for this communication?
• What exact source and destination MAC address is used for this communication?
• Is it possible that the frames sent by the PC during the programming are shorter than 64 bytes in total (60 bytes shown in Wireshark as the frame checksum is not displayed)?
• Is it possible that during the programming, the port on the drive changes link speed? That would cause the switch port to become temporarily blocked by Spanning Tree Protocol (STP) for up to 30 seconds.

A Wireshark record of the programming session would be very helpful in providing these details. In fact, two independent recordings should be done: One that SPANs the port where the PC is connected, and another that SPANs the port where the drive is connected.

Before that, though, I suggest you do the following:

1. Remove the ACL you have configured
2. Set up all ports as static access ports - this will save some time during their link-up during which these ports try to find out if they're connected to another Cisco switch
3. Set up all ports as PortFast ports - this will tell STP that these ports are guaranteed to be connected to end hosts that will not cause a switching loop, and so can be unblocked immediately, shaving off 30 seconds of being blocked. This will also prevent blockings the port for 30 seconds if it flaps momentarily or changes speed, which is what could be happening during reflashing of the drive.

This config will do the trick (please keep the whitespaces around hyphens and commas in the interface range command exactly as shown, ideally do a copy&paste):

interface range fa0/1 - 24 , gi0/1 - 2
 no ip access-group 100 in
 switchport mode access
 spanning-tree portfast
 exit
no access-list 100

If these modifications do not help to make the programming work, would it be possible for you to record the SPANned traffic in Wireshark as suggested and post it here for further analysis?

Best regards,
Peter