This chapter provides the following information:
Hardware Diagnostics Checking Port LEDs Troubleshooting Checklist Checking if Proper Drivers are Loaded Running Cable Diagnostics Testing Network Connectivity Software Problems and Solutions
Loopback diagnostic tests are available for testing the adapter hardware under Windows. These tests provide access to the adapter's internal/external diagnostics, where packet information is transmitted across the physical link (refer to "Diagnostics", in the Broadcom Advanced Server Control Suite chapter).
The adapter has four LEDs, one for each
port speed option (10 Mbps, 100 Mbps, and 1 Gbps), and one for Activity. The
three port speed LEDs indicate active links, and the Activity LED indicates
data transfer status. Once the driver is loaded and the cables are connected
properly, the appropriate speed LED is lit and the data LED is on if data traffic
is present. Before the port LEDs can provide troubleshooting information, the adapter must
be connected to the network (see "Installing the Hardware"),
and the network drivers for your particular operating system must be installed.
LED | State | Description |
1000 | On | Good Gigabit Ethernet link. |
Off | No 1000 Mbps link; possible link at different speed, possible bad cable, bad connector, or configuration mismatch. | |
100 | On | Good 100 Mbps Fast Ethernet link. |
Off | No 100 Mbps link; possible link at different speed, possible bad cable, bad connector, or configuration mismatch. | |
10 | On | Good 10 Mbps Fast Ethernet link. |
Off | No 10 Mbps link; possible link at different speed, possible bad cable, bad connector, or configuration mismatch. | |
ACT | Blinking | Brief bursts of data detected on the port. |
On | Streams of data detected on the port. | |
Off | No data detected on the port. |
![]() |
WARNING — Before opening the cabinet of your system for removing or inserting the adapter, please review all precautions outlined under "Safety Precautions." |
The following checklist provides recommended actions to take to resolve problems installing the Gigabit Ethernet Adapter or running it in your system.
The following section describes how to check if the proper drivers are loaded for Windows, NetWare, and Linux.
The Broadcom Advanced Server Control Suite, Vital Sign screen allows you to view vital adapter information, network status, and network connectivity. Active adapters are listed.
![]() |
NOTE – Information for non-Broadcom adapters is less comprehensive than information listed for Broadcom adapters. |
To verify that the driver is loaded properly, type
LOAD B57.LAN FRAME_ETHERNET_II NAME=B57_1_EII
This command automatically verifies if the link is active, and if so displays "Link is up".
From the command line, type config then press Enter. The following status information is displayed:
Broadcom NetXtreme Gigabit Ethernet Adapter
Version:
Hardware Setting:
Mode Address:
Frame Type:
Board Name:
Lan Protocol: ARP (see note)
LAN Protocol: IP Addr: (see note)
![]() |
*NOTE – The LAN protocol status appears after assigning an IP address to the adapter (i.e., bind). |
To verify that the bcm5700.o driver is loaded properly, run
lsmod
If
the driver is loaded, a line similar to the one below appears, where <size>
Module
|
Size
|
Used by
|
BCM5700
|
<size>
|
<n>
|
To verify that the bcme driver is loaded properly, run
/etc/conf/bin/idinstall -g -s bcme
A line appears showing that the driver is loaded.
To verify that the bcme driver is loaded properly, run
/sbin/modadmin -s bcme
A line appears showing that the driver is loaded.
To verify that the bcme driver is loaded properly, run
modinfo | grep bcme
A line appears showing that the driver is loaded.
The following section describes how to run the cable diagnostics from the Broadcom Advanced Server Control Suite. Refer to "Broadcom Advanced Server Control Suite," for setups and initialization.
![]() |
NOTE – This test is designed to work for a 1 gigabit copper link (1000BASE-T) only. |
From the Cable Analysis screen the user can monitor conditions of an Ethernet CAT5 cable connection within a cable plant in an Ethernet network.
![]() |
NOTE – The network connection will be lost when running these tests. |
The Length sub tab allows you to verify cable length and determine whether your configuration has the appropriate cable, which are calculated by a Return Loss algorithm. This utility allows you to determine whether the problem is with the adapter or in the cable plant.
Interface components for the BASCS Cable Analysis/Channel Pairs window are described below:
Distance: This field presents the estimated cable length in meters by averaging all four channels using Return Loss algorithms.
Margin: Margin yields the minimum distance between the measured cable pair and the maximum IEEE 802.3ab limits. The unit is in dB.
Each channel’s frequency response is displayed based the computation by the cable algorithms. The two graphs represent the values calculated by the Cable Loss and Return Loss algorithms. The vertical axis represents the gain in dB and the horizontal axis represents the operating frequency. The blue graph is the IEEE 802.3ab limit and the red graph is the actual computed values for a particular twisted pair. The minimum Margin and Frequency Margin fields in the Cable Length display are derived from these graphs.
![]() |
NOTE – Network connection will be lost when running this test. |
The combination of the cable length and the cable status reveal the state of the cable connection between the local adapter and the far end node (hub/switch/routers). The cable status will display possible problems associated with the cable if the cable is not in working order. In this case, displayed cable length is meaningless since the correct cable length cannot be calculated properly with a broken cable. The status window displays various error messages that diagnose possible problems associated with the cable.
The following section describes how to test network connectivity for Windows, NetWare, and Linux.
Use the ping command to determine if network connectivity is working.
Ping an IP host on the network to verify connection has been established:
From the command line, type load ping <IP address>, then press Enter.
This will display the packet send/receive status.
To verify that the Ethernet interface is up and running, run 'ifconfig' to check the status of the Ethernet interface. 'netstat -i' can also be used to check the statistics on the Ethernet interface. Consult manual pages for more information on 'ifconfig' and 'netstat'.
This section provides a list of known software problems and solutions for the operating systems below.
The following table lists know problems and solutions using Windows 2000:
Problem: Able to create Broadcom LAC connections icons having same name under network properties menu. |
Solution: In Windows 2000, each network adapter installed properly will have an icon in Network Properties window. The name of the icon is usually in the form of "Local Area Connection ##" and where ## is a number starting from 1. The names of icon can be changed by right clicking the icon and selecting rename. The name of the icon is only meaningful to the Network Properties window. Sometimes the system administrators will rename these icons to easily differentiate the network connections. When BASP is configured, BASP will create additional adapter icons and rename the icons. The naming convention is to allow end users to quickly identify the adapter with corresponding team and VLAN. The names of the icons created by BASP therefore are not recommended to be changed. |
Problem: Uninstalling the BASP software and user is prompted to reboot early |
Solution: In Windows 2000, after configuring BASP team, the system may sometimes prompt user to reboot. This is because Plug and Play on W2k may fail to commit any change in the network protocol binding. Users can choose not to reboot and continue configuring intermediate driver without having any side effect. When user has finished all the configuration, it is required to reboot. |
Problem: Cannot enable VLAN after it is disabled |
Solution: In Windows 2000, BASP creates additional network connections in "Network Connection and Dail-up" Window. Similar to physical network connections, these virtual connections can be disabled via the context menu. However, if these virtual connections are re-enabled, the system will report error as "connection failed!". This is known problem with Windows 2000. When this happens, reboot the system will enable the virtual connection again. |
Problem: Able to configure Internet Connection Sharing (ICS) when there aren't any unassigned adapters. | Solution: Windows 2000 networking is shipped with the Internet Connection Sharing (ICS), which is designed to allow multiple computers accessing the Internet via a Windows 2000 system. To configure ICS, a user would need to select a network connection as "outside" connection, and another network connection as "inside" connection. Broadcom has observed that ICS allow any available network connection to be "outside" and "inside" connection, regardless of the BASP team configuration. Broadcom recommends that the user not select any network connection that is part of the BASP team to be "outside" and "inside." |
Problem: Team configuration is not retained when user goes back to edit. |
Solution: In configuring BASP team configuration, the actual changes of the configuration is not committed until the user click "OK" in "Network Properties" window. If a user chooses not click OK and instead goes back to the BASP team configuration window, all the previous uncommitted changes will be lost and user will need to reenter the configurations. As a workaround, the user should always click "OK" in "Network Properties" window after making changes. |
Problem: IP address is configurable on a member of the team. |
Solution: When a team is created, the TCP/IP properties for the adapters are unselected. The user can manually select and configure TCP/IP properties of the adapters and configure an IP address. This is a limitation of the Windows 2000 network installation paradigm, where this invalid configuration is still allowed. |
Problem: When creating 64 VLANs, all virtual adapters show disconnected or one of the 64 VLANs show disabled. |
Solution: The maximum VLAN configurations are 63 tagged and one untagged VLAN ID 0. If 64 tagged VLANs are created, they are disconnected. A reboot is required and only 63 tagged VLANs show links, while a 64th is disabled. |
The following table lists known problems and solutions using Linux:
Linux-Basp |
|||
Problem: Linux-Basp RPM fails to install on Mandrake 7.1.1. | Solution: Use the tar ball to install BASP, refer to the tar-ball installation instruction. | ||
Problem: Configuring Basp and/or changing my IP address on Caldera 2.4 causes my system to hang on shutdown or when running 'df'. | Solution: This is a third party issue that is not related to the bcm5700 core or basp modules. This is seen on Caldera 2.4 (kernel 2.2.14). The problem occurs when bringing up Caldera 2.4 with an interface that is configured to start at boot time with a particular IP address (i.e., 192.168.0.10), then later changing the address manually via ifconfig (ifconfig eth* 192.168.0.50). This apparently causes RPC to hang indefinitely. This is noticed when attempting to run 'df' or halting the OS. The problem has also been observed with other third party adapters. The work around would be to configure the interfaces that will be used for basp with blank IP addresses or an IP address that is all 0's, in the ../network-scripts/ifcfg-eth* file. For more information regarding changing your IP address using 'ifconfig', please consult your Caldera documentation. | ||
Problem: When obtaining the IP for a SLB's virtual interface via DHCP, the IP-gets lost under heavy traffic. | Solution: Always set a static IP for all the virtual interfaces in a SLB team. | ||
Problem: The BASP and baspcfg man pages are not found in Redhat 6.2. They are not in the default path. | Solution:
Manually copy the man pages to the default man page directory:
cp /usr/share/man/man4/basp.4.gz /usr/man/man4 |
||
Linux-Core |
|||
Problem: Installing the driver on RedHat 7.0 Enterprise, generates unresolve symbol. | Solution: Refer to README.rh7 for instructions. This file can be found in the Tar-Ball. | ||
Problem: Compiling the driver fails under SuSE 7.x. | Solution:
If compiling the driver under one of the SuSE 7.x kernels and errors are
reported, please follow the general guidelines below to rebuild the kernel
source tree:
cd /usr/src/linux-<kernel_version>.SuSE where <kernel_version>
is the actual kernel version used in the SuSE distribution. In SuSE 7.1, which may optionally come installed with two kernel versions, the /boot/vmlinuz.* files described above may be found as /boot/vmlinuz_24.* for the 2.4.0 kernel and /boot/vmlinuz.* for the 2.2.18 kernel. Note that the source files for these kernels, may need to be installed manually from the respective SuSE Distribution CD-ROM. |
||
Problem: The SusE 7.0 SMP hangs. | Solution: Recompile the kernel's source tree. | ||
Problem: Users of Linux-Mandrake 8.0 may experience system hangs during the boot time if they configure newly installed BCM570x NIC's using the Kudzu utility.
|
Solution: Mandrake has adopted RedHat's 7.1's pcitable configuration file which contains older BCM570x PCI information. This adoption contains a bug which causes Kudzu to attempt to 'insmod' the tulip driver module when detecting BCM570x NICs. This occurs even though the correct BCM570x module has been installed sucessfully. Note, that Mandrake's source does not appear to be patched with support for the BCM570x module. To work around this issue, select the "Do Nothing" or "Ignore" options when presented with the kudzu configuration for BCM570x NICs. "Ignore" is prefered to prevent the wizard from reappearing everytime the user reboots. It is safe to allow Mandrake's kudzu to remove the NIC configuration, but not to allow it to configure new instances of any BCM570x NICs. |
The following table lists known Solaris problems and solutions:
Problem:
The Broadcom NetXtreme Gigabit Ethernet adapter can not be initialized (in Solaris 7 or 8) eventhough the adapter's instance number is shown. |
Solution: Operating system might not have enough physically contiguous memory. In order to allocate more memory, set an O/S system variable "lomempages"in /etc/system. Please see "Installing the Driver" in Solaris Driver Software section for more details. |
The following table lists known UnixWare problems and solutions:
Problem: In OpenUnix 8, in some machines with hot swappable slots the driver may not bind to the NetXtreme Gigabit Adapter. | Solution: During the install of the OS do not choose to install the Compaq hot swap pci feature. If the Hot Plug feature has already been installed in system use "pkgrm cpqhpcd" to uninstall the package and then proceed with installing the NetXtreme Driver. |
The following table lists known miscellaneous problems and solutions:
Problem: The Broadcom NetXtreme Gigabit Ethernet adapter can not be seen on the PCI BUS. | Solution: This is because some of the older servers are advertising themselves as PCI-X capable systems. This causes the NetXtreme network adapter to operate in PCI-X mode; therefore; not to be seen by the PCI bus. This problem can be overcome by configuring the firmware to operate in forced pci mode. Refer to b57diag for configuration instructions. |