Netra T1s hate SUNGEM and QFE
Hello,
You might remember my problem with a Netra T1 105 and a qfe card. Any
packet that hit the qfe card locked the box up with no error message.
A power cycle was needed.
The solution was to try the qfe in a Netra T1 AC200. Problem #1,
the Debian kernel in the install CD doesn't understand the IDE chip.
OK, so we rip the drives out of the T1 105 and put them into the T1
AC200.
Ah, Sun in their wisdom changed the interface from an hme to an eri,
bastards. This means you need to use the SUNGEM driver. OK, recompile
with new driver, I got eth0 to eth5 (0-3 qfe using happy meal driver,
4-5 using SUNGEM driver).
Kernel is 2.4.17
First problem, what does this mean?
# hwclock --show
Timed out waiting for time change.
Sounds like the public service or something.
kernel: eth0-3: Quattro HME (PCI/CheerIO) 10/100baseT Ethernet DEC 21153 PCI Bridge
kernel: eth0: Quattro HME slot 0 (PCI/CheerIO) 10/100baseT Ethernet 08:00:20:bf:01:40
kernel: eth1: Quattro HME slot 1 (PCI/CheerIO) 10/100baseT Ethernet 08:00:20:bf:01:41
kernel: eth2: Quattro HME slot 2 (PCI/CheerIO) 10/100baseT Ethernet 08:00:20:bf:01:42
kernel: eth3: Quattro HME slot 3 (PCI/CheerIO) 10/100baseT Ethernet 08:00:20:bf:01:43
kernel: sungem.c:v0.96 11/17/01 David S. Miller (davem@redhat.com)
kernel: eth4: Sun GEM (PCI) 10/100/1000BaseT Ethernet 00:03:ba:0b:be:3e
kernel: eth5: Sun GEM (PCI) 10/100/1000BaseT Ethernet 00:03:ba:0b:be:3d
All is well until we start putting some load on the eth4 and eth5
interfaces, first we get this stuff:
kernel: eth5: TX MAC max packet size error.
kernel: eth5: TX MAC xmit underrun.
kernel: NETDEV WATCHDOG: eth5: transmit timed out
kernel: eth5: transmit timed out, resetting
kernel: eth5: TX_STATE[003ffc05:00000000:00000019]
kernel: eth5: RX_STATE[0300c805:00000001:00000001]
kernel: eth5: Link is up at 100 Mbps, half-duplex.
and
kernel: eth5: TX MAC xmit underrun.
kernel: eth5: RX MAC fifo overflow.
kernel: eth5: TX MAC xmit underrun.
and then the netra decides to give the whole game away and just stop
sending or receiving packets on that interface. It only happens with
eth4 and eth5 (the eri using GEM) and when there is reasonable load,
like downloading a big webpage, which really is not that big at all.
Any suggestions? I've moved eth5 onto eth3 so I only got one dodgy
interface left. Please CC me.
--
Craig Small VK2XLZ GnuPG:1C1B D893 1418 2AF4 45EE 95CB C76C E5AC 12CA DFA5
Eye-Net Consulting http://www.eye-net.com.au/ <csmall@eye-net.com.au>
MIEEE <csmall@ieee.org> Debian developer <csmall@debian.org>
Reply to: