HOWTO AIX:TROUBLESHOOT BLOCK MULTIPLEXER ADAPTER INITIAL CONFIGURATION PROBLEMS.
ITEM: RTA000026261
PLATFORM: RISCSYSTEM
OPERATING SYSTEM:
AIX FOR RISC SYSTEM/6000 Version 3 Release 2 Modification 1
*********************************************************************
** This HOWTO item was written and tested with the above operating **
** system(s). The provided explanations, techniques and procedures**
** have been reviewed for technical accuracy and applicability. **
** Though the techniques and information contained in this item **
** may work on other levels of the operating system(s), it has not **
** necessarily been tested. Normal precautions should be taken in **
** adopting these same techniques and procedures in your own **
** environment. **
*********************************************************************
SCENARIO/EXPLANATION:
The most common errors made during configuration of the Block
Multiplexer deal with getting the host and device (RISC System/6000
with Block Multiplexer (BMX) adapter, referenced here as the "system
unit") to shake hands successfully. Any problems are most commonly
manifested in the error message "Device cannot be varied online
because no channel path is available." This HOWTO explains what the
message really means and how to determine/resolve the actual problem.
SOLUTION:
A. REASONS FOR THE MESSAGE "no channel path is available"
=========================================================
This usually happens when you are trying to "vary on" and "attach"
devices on the host side. Basically, an "attach" makes the device
available to the host application (in this case the device is the
is the system unit and the application is TCP/IP). Prior to that,
a "vary on" must be performed. This operation establishes that the
devices are there. The details of the "vary on" operation are as
follows:
1. The channel will attempt to select the device with a command
like a "Sense ID".
2. When the channel issues the command, two possible scenarios can
occur. If the device is really there, it will respond to the
command and respond with the Sense ID information, and the vary on
will succeed. If the device is not there (or available), then the
channel will receive a special signal, "select in", to indicate
that the addressed device is not available to the channel.
3. If the channel detects a "select in" signal, then a condition
code 3 (CC3) is posted to the host, which translates to a "no
channel path is available" message.
B. THERE ARE SEVERAL REASONS FOR A CC3:
=======================================
1. The physical cabling is wrong.
Have your CE check all your connections from the host to the
system unit.
2. The IOCP GEN is wrong.
Check with your host System programmer to ensure the device is
gen'd to the system correctly.
3. The CHPID is offline.
Although this is not too common, if the system unit is being
brought up for the first time, and the CHPID was taken offline
for the cable installation, it is possible that the CHPID was
not brought online after the cables were installed. Have your
operator vary on the chipid.
4. The problem lies with the system unit.
If the system unit is the suspected problem, it could be any of
the following:
a. The S/B switch is in the wrong position.
This switch is located on the cable itself. For normal mode,
it should be set to "select", or "S". The "B" is "bypass" and
is essentially a loopback (select out to select in), and thus
does not route to the BMX adapter.
b. No application is running on the system unit side.
The connection requires that an application be initiated and
running on the system unit side before the channels will
be operational on the host side. This is because the system
unit must start a subchannel to bring the adapter "online".
When the adapter is "online", it will recognize its address
when the channel attempts to select it. If the adapter is
"offline", the adapter will loop the "select out" signal (used
to select the specified device) to "select in", which indicates
to the host channel that the device is not there, as described
above. This "looping" is done on the card (there are two places
for this loop, the cable switch and the adapter).
c. The wrong address has been coded on the system unit or host.
Check the SMIT DEVICES panel and the TCPIP PROFILE to ensure
the addresses are the same, and that it matches the address in
the IOCP gen.
C. HOW TO DETECT IF THE PROBLEM IS HOST OR SYSTEM UNIT
======================================================
Use the program "catcsx" in non-claw mode to determine if the
problem lies on the system unit or at the host. Generally this will
be used when the "channel path not available" message has been
encountered. Make sure you have a subchannel assigned to the Block
Multiplexer which is non-claw mode, as in the example.
NOTE: catcsx is NOT a supported diagnostic tool. It has been supplied
for IBM CEs only. Use at your own discretion.
NOTE: Do not use catcsx in claw mode. CLAW mode requires a certain
host protocol that is not used in a simple vary on command.
Here are the steps for making this determination.
1. Run the following command or use SMIT DEVICES to determine that
the BMX adapter is configured correctly:
lsattr -E -l cat0
Output:
=======
dma_bus_mem 0x00F40000 Bus memory address False
bus_intr_lvl 4 Bus interrupt level False
dma_lvl 0x6 DMA arbitration level False
xmitsz 4096 Transmit buffer SIZE
xmitno 26 NUMBER of transmit buffers
recvsz 4096 Receive buffer SIZE
recvno 26 NUMBER of receive buffers
rdto 0 Receive data transfer OFFSET
speed 2 Channel SPEED
first_sc 0x00 STARTING subchannel address
num_sc 36 NUMBER of subchannel addresses
clawset 0x24 CLAW mode subchannel set
switch online Online/Offline SWITCH
indicator online Online/Offline INDICATOR
ucode_type 1 N/A
host_name HOST N/A
adapter_name PSCA N/A
Check the xmitsz, recvsz, first_sc, num_sc, host_name, and
adapter_name fields. They should match what was entered in the
TCPIP PROFILE file.
2. By default the subchannels are non-claw mode. So in this example,
there are 36 subchannels available: subchannels 00-23 and 26-27 are
non-claw, and claw mode subchannels are 24-25. Recall this is
in hexadecimal. Claw mode uses a subchannel pair, whereas non-claw
mode uses single subchannels. The non-claw subchannel 0x00 is used
in the following example. Now run /usr/lpp/blkmux/test/catcsx
by entering the following sequences to the prompts:
0) Open a device.
0) Open /dev/cat0.
/dev/cat0 is open.
1) Start a subchannel on an open device.
MAIN Read select timeout:
Enter 1 for claw mode, 0 for non-claw mode.
0
Enter the hex value of the subchannel you want started
Or the even subchannel of the claw mode subchannel pair.
0x00
Enter 1 to start with unsolisited DE.
Issue start.
Wait for start done.
Link_id returned was 00000000
option(1) was 00000000
option(3) was 00000000
Device /dev/cat0 has subchannel 00 online.
3. At this point, go to the host and vary on 900. This proves the
block multiplexer was able to successfully start a subchannel to
the host. Therefore the problem is not the bmx adapter.
D. HOW TO CLEAN UP A BADLY CONFIGURED TCP/IP ON THE SYSTEM UNIT
===============================================================
If you have TCP/IP running but suspect it is incorrect, it is best
to remove it from the ODM and start fresh. Here are the steps to
do so:
1. Close the device driver and shut down the subchannel:
ifconfig ca0 down
ifconfig ca0 detach
2. Remove the device driver and all configuration info from ODM:
rmdev -l ca0 -d
NOTE: If this command doesnt work, you may have to remove the adapter,
reboot the machine, run diag -a and remove all references to
ca0.
3. Use SMIT to reconfigure the interface:
smit tcpip
Further Configuration
Network Interfaces
Network Interface Selection
Add a Network Interface
NOTE: DO NOT use the Minimum Configuration menu. There are bugsĒ
At this point fill in the correct address, the subchannel IN
DECIMAL, and choose the "activate now" option. This will recreate
the interface and put it in the ODM.
NOTE: All references to subchannels must be in decimal for TCP/IP,
and in hex for device references.
NOTE: You will not be able to ping your adapter address. In point to
point connections there is no loopback mechanism.
*****************************************************************
** If you have found this information to be informative and **
** useful, please let us know via a HONE FEEDBACK or via ASKQ. **
** Make sure in your evaluation to reference the appropriate **
** INFOSYS or FLASH number. **
*****************************************************************
---------- ---------- ---------- --------- ---------- ----------
This item was created from library item Q604637 2VRDW
S e a r c h - k e y w o r d s:
$$A ADAPTER AIX ALTERNATE ATTACH ATTACHED ATTACHMENT BLKD BLKMUX
BLOCK BMX CHAN CHANNEL COMMUNICATIO CONFIGURATIO HOWTO INDEX INIT
INITIAL INITIALIZE INITIATE INITIATOR IX MULTIPLEXER PROBLEMS RISC
RISCOC RISCSYST RISCSYSTEM SEP92 SOFTWARE SYSTEM S370 TROUBLESHOOT
V3 2VRDW
WWQA: ITEM: RTA000026261 ITEM: RTA000026261
Dated: 07/1998 Category: RISCSNA
This HTML file was generated 99/06/24~12:43:09
Comments or suggestions?
Contact us