Software System Detection Mechanisms - Extreme Networks BlackDiamond 6804 Troubleshooting Manual

Extreme Networks' Complementary Detection of Packet Errors Between Wires
transmitted, but an invalid CRC value is included with the packet. Therefore, the receiving device will
detect an invalid CRC value and will drop the packet.
In Summit stackable switches, the packet checksum is calculated by the MAC ASIC on the receiving
port and is compared against the verification checksum calculated by the MAC ASIC on the
transmitting port, as described above.
In Alpine 3800 series switches, the packet checksum is calculated by the MAC ASIC on the receiving
port on the I/O module on which the packet is received. The packet checksum and packet are passed to
the switch fabric, which is on the Alpine switch backplane, and then from the switch fabric to the
transmitting MAC ASIC on the I/O module on which the packet is to be transmitted. There, the
verification checksum is computed and compared against the packet checksum.
In BlackDiamond 6800 series switches, the packet checksum is computed by the MAC ASIC on the
receiving port on the I/O module on which the packet is received. The packet checksum and the packet
traverse the switch fabric on the I/O module and are handed off to either an external MAC ASIC,
connected to a network port, or to an internal MAC ASIC, connected to a BlackDiamond backplane link.
In either case, the behavior of the MAC ASIC is the same: it computes the verification checksum and
compares it against the packet checksum to detect any changes in packet data. Similarly, whether the
packet is transmitted out the external port to the network, or out the internal port to the BlackDiamond
backplane, the packet is accompanied by an Ethernet CRC.
The behavior of the BlackDiamond MSM module is identical to that of the BlackDiamond I/O module,
except that all MAC ASICs on the MSM are internal (not to network ports). Regardless, the behavior of
the receiving and transmitting MAC ASICs is the same for packets traversing an MSM module as for
packets traversing an I/O module.
Thus far, all of the systems described have been involved in fast-path forwarding. Therefore, any
checksum errors detected using the mechanisms described above are referred to as fast-path checksum
errors.
Packets that must be processed by the switch CPU are also validated by checksum values. When a
packet is received, it might be destined specifically for the CPU (as in the case of protocol packets) or it
might be passed to the CPU for assistance in making a forwarding decision (if the switch fabric lacks
the information required to forward the packet correctly). In either case, the receiving MAC ASIC still
computes and prepends a packet checksum just as it does for fast-path packets, but because the packet
is not passed to a transmitting MAC ASIC before it is forwarded, the switch fabric itself is responsible
for computing the verification checksum and comparing it against the packet checksum. If a mismatch
is found, the switch fabric reports the checksum error condition to the CPU as it passes the packet up to
the CPU. These types of checksum errors are one instance of a class of checksum errors known as
slow-path checksum errors.

Software System Detection Mechanisms

As described above, each MAC ASIC maintains a port-by-port count of every checksum error detected.
ExtremeWare contains mechanisms that can retrieve the checksum error counts from the MAC ASICs in
the switch and act on it. Current versions of ExtremeWare retrieve the checksum error counts from all
MAC ASICs in the switch at twenty-second intervals. The counts at the end of the twenty-second
interval are compared with the counts at the beginning of the twenty-second interval on a port-by-port
basis. If, for any given port, the count is found to be different, then ExtremeWare is said to have
detected a checksum error. Depending on the ExtremeWare version, the configuration settings, the
frequency and count of checksum errors, and a variety of other factors, ExtremeWare will initiate one of
several actions, described in the section "System (CPU and Backplane) Health Check" on page 69. For
Advanced System Diagnostics and Troubleshooting Guide 29