That did it. Thanks Alan. You know, I'd never needed to add the
primary_router parm before. I read about it in the Device Drivers manual,
but the OSA I was using configured as OSD was routing just fine until a few
weeks back. There was a microcode update about a month ago -- I wonder if
a microcode upgrade for/in the OSA effected the need for this -- ? It
sounds reasonable. Because the card did work fine until I noticed I wasn't
getting 100Mb throughput, made sure the card was explicitly set to "100
Mb/Full Duplex" at the HMC (instead of "Auto negotiate" which doesn't seem
to work here), and disabled/enabled the port.
Anyway, for anyone else looking through the archives that may have the same
question at a later date, here's how I added it:
tuxmstr1:/usr/local/src/iftop-0.11 # cat /proc/qeth
devnos (hex) CHPID device cardtype port chksum prio-q'ing rtr fsz C
cnt
-------------- --- ---------- -------------- -- -- ---------- --- --- -
---
0734/0735/0733 x0D hsi11 HiperSockets 0 no always q 2 no 64k
128
0730/0731/0732 x0C hsi10 HiperSockets 0 no always q 2 no 64k
128
0728/0729/0727 x0B hsi9 HiperSockets 0 no always q 2 no 16k
128
0724/0725/0726 x0A hsi8 HiperSockets 0 no always q 2 no 16k
128
0722/0723/0721 x09 hsi7 HiperSockets 0 no always q 2 no 16k
128
0718/0719/071A x07 hsi6 HiperSockets 0 no always q 2 no 16k
128
0716/0717/0715 x06 hsi5 HiperSockets 0 no always q 2 no 16k
128
0712/0713/0714 x05 hsi4 HiperSockets 0 no always q 2 no 16k
128
070A/070B/0709 x04 hsi3 HiperSockets 0 no always q 2 no 16k
128
0706/0707/0708 x03 hsi2 HiperSockets 0 no always q 2 no 16k
128
0704/0705/0703 x02 hsi1 HiperSockets 0 no always q 2 no 16k
128
0700/0701/0702 x01 hsi0 HiperSockets 0 no always q 2 no 16k
128
F800/F801/F802 xF8 eth0 OSD_100 0 no always q 2 no 64k
128
tuxmstr1:/usr/local/src/iftop-0.11 # echo primary_router eth0 > /proc/qeth
tuxmstr1:/usr/local/src/iftop-0.11 # cat /proc/qeth
devnos (hex) CHPID device cardtype port chksum prio-q'ing rtr fsz C
cnt
-------------- --- ---------- -------------- -- -- ---------- --- --- -
---
0734/0735/0733 x0D hsi11 HiperSockets 0 no always q 2 no 64k
128
0730/0731/0732 x0C hsi10 HiperSockets 0 no always q 2 no 64k
128
0728/0729/0727 x0B hsi9 HiperSockets 0 no always q 2 no 16k
128
0724/0725/0726 x0A hsi8 HiperSockets 0 no always q 2 no 16k
128
0722/0723/0721 x09 hsi7 HiperSockets 0 no always q 2 no 16k
128
0718/0719/071A x07 hsi6 HiperSockets 0 no always q 2 no 16k
128
0716/0717/0715 x06 hsi5 HiperSockets 0 no always q 2 no 16k
128
0712/0713/0714 x05 hsi4 HiperSockets 0 no always q 2 no 16k
128
070A/070B/0709 x04 hsi3 HiperSockets 0 no always q 2 no 16k
128
0706/0707/0708 x03 hsi2 HiperSockets 0 no always q 2 no 16k
128
0704/0705/0703 x02 hsi1 HiperSockets 0 no always q 2 no 16k
128
0700/0701/0702 x01 hsi0 HiperSockets 0 no always q 2 no 16k
128
F800/F801/F802 xF8 eth0 OSD_100 0 no always q 2 pri 64k
128
Of course this is dynamic. To make it perm you can add it as a QETH parm
in /etc/chandev.conf:
add_parms,0x10,0xF800,0xF802,portname:F800PORT,primary_router
qeth0,0xF800,0xF801,0xF802
You should be able to "rcchandev reload" without causing the interface to
come down (but I didn't want to chance it in the middle of the day.)
Matt Lashley
Idaho State Controller's Office
Alan Altmark
<[EMAIL PROTECTED] To: [EMAIL PROTECTED]
ibm.com> cc:
Sent by: Linux on Subject: Re: OSA Fast Ethernet (OSD) -
QDIO mode
390 Port
<[EMAIL PROTECTED]
IST.EDU>
07/27/2003 10:06
PM
Please respond to
Linux on 390 Port
On Friday, 07/25/2003 at 03:39 CST, Matt Lashley/SCO
<[EMAIL PROTECTED]> wrote:
> Is it normal for ifconfig to display all 0's for the MAC address of an
OSA
> interface?
Yes, for QDIO mode. No, for LCS.
> Will this cause problems for the ARP tables/cache of the Cisco router to
> which the card is plugged?
No. All you are seeing is that the QETH driver is not extracting the MAC
address from the card. QDIO is a layer 3 transport protocol that
insulates the host from having to deal with layer 2 issues, including MAC.
The card handles ARP request & response without help from the host. In
LCS mode, the device operates as a layer 2 transport, so MAC is directly
visible to the host.
> Strangely, on the same machine after taking down the QDIO interface, I
can
> bring up a second OSA using LCS, configure it with the IP of the QDIO
card
> and everything routes to the 192.168.x.x machines just fine.
It sounds like you forgot to specify "primary" as a QETH configuration
option in chandev.conf. Rule: If you are going to route through an OSA
attached to a guest, the OSA connection should be defined as primary.
> Oh, and by the way, the LCS card, although configured 100Mb/Full Duplex
> won't get over 10Mb. -- ? The LAN to which it is connected is
100Mbit/Full
> as well.
Remember that on VM you only get a slice of the CPU. If the guest doesn't
get CPU, it can't drive I/O.
Alan Altmark
Sr. Software Engineer
IBM z/VM Development
