RE: "Happy Meal Ethernet": Extremely slow and "MAX Packet size" errors

Fred,

You might find that you will get a quicker response my sending your message
directly to the mailing list ([email protected]) so that others have
chance to respond.

Use ethtool for the hardware address. The latest version I could find
sources for was 1.5, available at http://sourceforge.net/projects/gkernel/
or http://rpmfind.net. Earlier versions my not have this feature. Assuming
you already have both interfaces configured the command might look like
this:

ethtool -s eth0 physad 00000C248FDC
ethtool -s eth1 physad 00000C244FA8

Just in-case, here is how the interfaces might be configured:

First, make sure there are entries in /etc/modules.conf

-----------------------------------------------------
Excerpt from /etc/modules.conf
-----------------------------------------------------
alias eth0 sunhme
alias eth1 sunhme
-----------------------------------------------------

Then run "depmod"

You should get no error messages.

You can use the "-v" argument to depmod to watch the list of modules scroll
while depmod checks dependencies.

Now, make sure that there are startup configuration files for each
interface.

cd /etc/sysconfig/network-scripts/

-----------------------------------------------------
  Sample /etc/sysconfig/network-scripts/ifcfg-eth0
-----------------------------------------------------
DEVICE=eth0
BOOTPROTO=static
BROADCAST=192.168.2.255
IPADDR=192.168.2.5
NETMASK=255.255.255.0
NETWORK=192.168.2.0
ONBOOT=yes
-----------------------------------------------------
-----------------------------------------------------
  Sample /etc/sysconfig/network-scripts/ifcfg-eth1
-----------------------------------------------------
DEVICE=eth1
BOOTPROTO=static
BROADCAST=192.168.2.255
IPADDR=192.168.2.6
NETMASK=255.255.255.0
NETWORK=192.168.2.0
ONBOOT=yes
-----------------------------------------------------

Then, shutdown and restart the interfaces:
(you must be at the console or connected to a console port)

ifdown eth0
ifdown eth1
ifup eth0
ifup eth1

Now, check to see that they're working:

netstat -i

Kernel Interface table
Iface   MTU Met    RX-OK RX-ERR RX-DRP RX-OVR    TX-OK TX-ERR TX-DRP TX-OVR
Flg
eth0   1500   0      377      0      0      0      128      0      0      0
BRU
eth1   1500   0        0      0      0      0        3      0      0      0
BRU
lo     8020   0        7      0      0      0        7      0      0      0
LRU

lsmod

Module                  Size  Used by
sunhme                128368   2  (autoclean)

> -----Original Message-----
> From: Fred Kutell [mailto:[email protected]]
> Sent: Thursday, April 04, 2002 5:11 PM
> To: [email protected]
> Subject: Re: "Happy Meal Ethernet": Extremely slow and "MAX Packet size"
> errors
>
>
> I saw that you seem to have hardware similar to mine.  I have a
> Netra T1 105
> Ultrasparc.
>
>
> Do you know how to make both of the Happy Meal interfaces have different
> hardware addresses ?
>
>
>
>
> "Cal Webster" <[email protected]> wrote in message
> news:<[email protected]>...
> > I have been unable to get respectable transfer rates through the
> integrated
> > Ethernet interface on our Sun Ultra IIi running RedHat Linux 6.2
> > (2.2.19-6.2.12).
> >
> > The machine was connected to a 3Com SSII 3300 switch and allowed to auto
> > negotiate.  It sets itself to 100Mb/s at Full Duplex. Initially, I was
> > getting somewhat slow, but workable rates on the inbound  transfers, but
> > sometimes unbearable 0.18 Kbps on outbound. Examination of
> > /var/log/messages showed that the interface was resetting frequently due
> to
> > "MAX Packet size" errors.
> >
> > After many hours of searching lists and news groups I found several
> messages
> > indicating that Sparc machines sometimes had problems with auto
> negotiation.
> > The  recommended fix was to manually set the card to run at half duplex.
> The
> > "sunhme"  driver does not return any arguments from "modinfo".
> After many
> > more hours I found  a reference to "ethtool", which I used to reset the
> > card. This caused other errors  when connected to the switch. I tried
> > various manual settings on the switch port with little success
> so I moved
> it
> > to a 3Com SSII Dual Speed Hub.
> >
> > The card still resets itself several times each day under heavy
> use. I now
> > run a  script every minute from the root crontab that checks to
> see if the
> > interface is at full duplex, in which case it is reset to half again.
> >
> > According to my users, this machine still has performance problems when
> > transferring large files or a large number of files. Most of the users
> > access Samba  shares on this machine. Whenever they complain about
> > performance I can find "MAX Packet size" error messages similar to those
> > below in the logs.
> >
> > There is obviously a larger problem here. resetting the
> interface for half
> > duplex  is only a Band-Aid. Does anyone have a good explanation for the
> "MAX
> > Packet size"  errors and poor network performance? More
> importantly, does
> > anyone have a better  solution then resetting the interface with
> "ethtool"?
> >
> > I would very much appreciate help in any form.
> >
> > Cal Webster
> > [email protected]
> >
> > ---------------------------------------------------------------------
> > Some benchmark tests transferring a file (WinNT Service Pack 6)
> with ftp.
> > These two  machines are on the same switch:
> >
> > At Full Duplex (best case):
> > nozzle -> winggear:
> > 	35725704 bytes sent in 7.14 secs (4.9e+03 Kbytes/sec)
> > nozzle <- winggear:
> > 	35725704 bytes received in 87 secs (4e+02 Kbytes/sec)
> >
> > At Half Duplex (best case):
> > winggear -> nozzle
> > 	4439568 bytes sent in 0.71 seconds (6.1e+03 Kbytes/s)
> > winggear <- nozzle
> > 	4439568 bytes received in 0.67 seconds (6.5e+03 Kbytes/s)
> >
> >
> > Typical entries from /var/log/messages:
> > (last line is from "ethtool" running from root crontab)
> >
> > kernel: eth0: Error interrupt for happy meal, status = 00000400
> > kernel: eth0: Happy Meal MAX Packet size error.
> > kernel: eth0: Resetting...
> > kernel: eth0: Link is up using internal transceiver at 100Mb/s, Half
> Duplex.
> > kernel: eth0: Link has been forced up using internal transceiver at
> 100Mb/s,
> > Half  Duplex.
> > ---------------------------------------------------------------------
> >
> >
> >
> > _______________________________________________
> > Sparc-list mailing list
> > [email protected]
> > https://listman.redhat.com/mailman/listinfo/sparc-list
>