Summary: 100M switching hub problems

From: Aditya Talwar (
Date: Thu Jan 30 1997 - 16:09:35 CST

My question was... [some folks could not read my mail. i had sent in as ascii
but came through as attachment form.]
Last week i upgraded my network from a simple 10m ethernet to 100/10M switched
ethernet via cabletron's MMAC PLUS 100M and 10M cards. Our 2 ultra 170E's
are directly connected to the 100M port, and concentrators hub are hooked
to the sitched 10M ports. I also changed domainnames for the office, so had
to build new maps, and i threw in a new defaultrouter, a cisco box.
With all this change one would think of a fast/zippy network... however
my network seems slow/sluggish/out of sync..

on the ultras with solaris 2.5 I have applied latest/greatest patches..
101655-04 102837-01 102980-04 103093-03 103226-06 103285-01 103492-01
101821-01 102839-01 102980-10 103093-07 103238-01 103300-02 103505-01
102055-01 102841-01 102982-01 103162-01 103241-01 103318-01 103506-01
102113-03 102846-01 102984-01 103169-06 103242-02 103325-01 103586-01
102114-03 102850-01 103017-04 103172-01 103242-04 103447-06 103667-03
102141-01 102934-04 103026-01 103186-21 103244-01 103468-01 103708-02
102211-03 102935-03 103048-02 103187-02 103244-03 103468-02 103746-01
102537-02 102964-05 103048-03 103187-04 103247-07 103476-01 103815-01
102802-01 102971-01 103060-01 103187-11 103266-01 103477-01
102832-01 102979-01 103066-01 103210-01 103276-01 103477-06
102832-02 102979-02 103075-02 103210-04 103279-01 103480-02
102835-01 102980-03 103093-02 103226-03 103279-02 103480-03

this seems to have helped a bit. as i no longer get NFS server not responding

1. on the ultra how do i know i am running at 100M/full duplex/half duplex?

   do i need to add ndd commands to increase the perfomance or highwatermarks?
   If so, What are they? Do i have to force the 100M full duplex networking?
   has anyone tinkered with cabletron parameters for sun 100M operation?
   cabletron/sun support have no specific instructions to my knowledge.
2. on sparc 5's/2.5 dtmail hangs when it tries to read mail from a solaris 2.4
   mail server. has anyone encountered this problem. i did not have this problem
   before the switchover. i have applied the latest CDE patch thou.
3. any public domain tools to monitor/isolate problems..

First thanks to the gurus who answered my queries. Nobody nailed the issue!
although their replies contained tons of valuable knowlegde to work with
100m on suns and 100m switching hubs in general ...

As to my problem, i solved them using Sun Tech's help and installed latest
network patches from Sun. Then i got a patch for the Cabletron Hub!
it seems that Suns work well in auto-negotiation mode at 100m Half duplex mode.
when i forced full duplex on the sun, it was out of sync with the 100M on the
hub. i still have to tune/work with this mode (so don't take my
word on this). (I am still looking for network management/monitoring tools
from PD any pointers will be appreciated). my network is humming smooth now.

Mighty thanks to... (Tom Southerland)
Satish Somanath <> (Joerg P. Kottusch)
Tony Jago <>

the valuable insights i got from their emails was...
If autonegotiation fails and the driver is not at the desired speed and duplex
mode, you can turn off negotiation and force the speed and mode of the link.

Note that only half duplex capability is fully supported at the present.
IEEE Industry standards for Full Duplex interoperability between vendors will
be completed in the near future. ie: no guarantee that full-duplex will
perform as expected.

see the following manuals for more info:
      SunFastEthernet Adapter 2.0 Installation and User's Guide - PN
       Platform Notes: The hme Fast Ethernet Device Driver - PN 802-3970-10

Solaris 2.x allows three ways to set hme driver parameters.
       example: turn off auto negotiation and force 100 MB full duplex:
/etc/system addtions.
       set hme:hme_adv_autoneg_cap=0
       set hme:hme_adv_100hdx_cap=0
       set hme:hme_adv_100fdx_cap=1
Manual ndd commands:
       # ndd -set /dev/hme adv_autoneg_cap 0
       # ndd -set /dev/hme adv_100hdx_cap 0
       # ndd -set /dev/hme adv_100fdx_cap 1

       Example of setting 2nd hme interface (hme1) to force 100mb half-duplex
       # ndd -set /dev/hme instance 1
       # ndd -set /dev/hme adv_autoneg_cap 0
       # ndd -set /dev/hme adv_100hdx_cap 1
       # ndd -set /dev/hme adv_100fdx_cap 0
create /kernel/drv/hme.conf

       name="hme" class="sbus"


       0x8c04000,0x00002000,0xe,0x8c06000,0x00002000,0xe,0x8c07000, 0x00000020

       adv_autoneg_cap=0 adv_100fdx_cap=1 use_int_xcvr=1 ipg1=10 ipg2=8;


Solaris 2.5 implements NFS version 3.0. Solaris 2.4 uses NFS 2.0.

NFS 3.0 uses TCP as its transport and a block size of 32kb
NFS 2.0 uses UDP as its transport and a block size of 8kb

Try using these options for the mount:


Cisco Ether Switch CPW2200
        -2 dedicated 100 Mb ports
        -12 10/100 Mb ports
Does your switch have full duplex ports?

My U2 170e sits on the dedicated 100Mb port. After putting the U2 on this
port the NFS requests started increasing. U2 should automatically switch
to 100Mb according to Sun Tech support.

Mail: /var/mail mounted from mail server...
Any errors in /var/adm/messages related to NFS?

I recommend running some network monitoring tool on one workstation from
each port off the switch to see where most traffic is going. You might
want to shuffle the workstations on the hubs to keep the talkitive ones

/etc/rpc.etherd le0
ethertop -ts 30

Put the mail server on the switch.


>1. on the ultra how do i know i am running at 100M/full duplex/half duplex?

To find out the settings on the driver

# ndd /dev/hme/ \?
? (read only)
transceiver_inuse (read only)
link_status (read only)
link_speed (read only)
link_mode (read only)
ipg1 (read and write)
ipg2 (read and write)
use_int_xcvr (read and write)
pace_size (read and write)
adv_autoneg_cap (read and write)
adv_100T4_cap (read and write)
adv_100fdx_cap (read and write)
adv_100hdx_cap (read and write)
adv_10fdx_cap (read and write)
adv_10hdx_cap (read and write)
autoneg_cap (read only)
100T4_cap (read only)
100fdx_cap (read only)
100hdx_cap (read only)
10fdx_cap (read only)
10hdx_cap (read only)
lp_autoneg_cap (read only)
lp_100T4_cap (read only)
lp_100fdx_cap (read only)
lp_100hdx_cap (read only)
lp_10fdx_cap (read only)
lp_10hdx_cap (read only)
instance (read and write)
lance_mode (read and write)
ipg0 (read and write)

To get a particular value

# ndd /dev/hme adv_100fdx_cap

1 is full duplex

Also read the man page on the hme driver.

> do i need to add ndd commands to increase the perfomance or highwatermarks?
> If so, What are they? Do i have to force the 100M full duplex networking?

100M full duplex networking has not been standardised upon so stick to half
duplex, though you may be lucky with a cabletron/sun solution (though it
doesn't sound like it)

This is what I use. I have 2 hme ports one running at 100 FDX and one at
10 HDX. Hope this helps.

ndd -set /dev/tcp tcp_old_urp_interpretation 1

ndd -set /dev/hme instance 0
ndd -set /dev/hme adv_autoneg_cap 0
ndd -set /dev/hme adv_100fdx_cap 0
ndd -set /dev/hme adv_100hdx_cap 0
ndd -set /dev/hme adv_10fdx_cap 0
ndd -set /dev/hme adv_10hdx_cap 1

ndd -set /dev/hme instance 1
ndd -set /dev/hme adv_autoneg_cap 0
ndd -set /dev/hme adv_100fdx_cap 1
ndd -set /dev/hme adv_100hdx_cap 0
ndd -set /dev/hme adv_10fdx_cap 0
ndd -set /dev/hme adv_10hdx_cap 0

ndd -set /dev/ip ip_path_mtu_discovery 0
ndd -set /dev/ip ip_send_source_quench 1

ndd -set /dev/tcp tcp_xmit_hiwat 65535
ndd -set /dev/tcp tcp_recv_hiwat 65535
ndd -set /dev/tcp tcp_cwnd_max 65534
ndd -set /dev/tcp tcp_conn_req_max 64


Standard Disclaimers apply. the above are my own opinions/experiences only
and not my employers.

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:11:44 CDT