[tulip-bug] driver failure under high NFS load

Donald Becker becker@scyld.com
Wed May 15 12:50:01 2002


On Tue, 14 May 2002, Greg Wooledge wrote:
> Donald Becker (becker@scyld.com) wrote:
> > What does 'tulip-diag' report when the interface is in this state?
> It was "dead" when I got up this morning, so I ran
>
> { tulip-diag -ee; tulip-diag -aa; tulip-diag -mm; } >/tmp/tulip-diag.out

Note that you can stack the options: tulip-diag -aa -eee -mm

> tulip-diag.c:v2.10 3/08/2002 Donald Becker (becker@scyld.com)
>  http://www.scyld.com/diag/index.html
> Index #1: Found a ADMtek AL985 Centaur-P adapter at 0x6c00.
>  0x00: fff98000 ffffffff ffffffff 13800800 13800a00 fc664010 ff9f2117 ffffebff
...
>  Transmit started, Receive started, half-duplex.
>   The Rx process state is 'Waiting for packets'.
>   The Tx process state is 'Idle'.

This is the normal status for a running adapter.
It appears that the interface is operating normally.
Are you seeing the packet count increase in /proc/net/dev?

At this point I'm guessing that the problem is with NFS, not the driver.

>  MII PHY #1 transceiver registers:
>    2000 784d 0022 5410 0081 40a1 0007 2801
>    ffff ffff ffff ffff ffff ffff ffff ffff
>    01c0 0028 0790 8020 4f02 0304 0026 0000
>    0000 cfbf 0000 ffff ffff ffff ffff ffff.
>  Basic mode control register 0x2000: Auto-negotiation disabled!
>    Speed fixed at 100 mbps, half-duplex.

This matches your previously reported forced speed.

>   ADMtek vendor specific registers information:
>    Cable length is approximately 0 meters.
>    The receiver has lost lock 0 times since last check and had 0 error events.

This link is normal, and you are not seeing cable noise.
It's not always possible to estimate the cable length, especially if
there is good impedance matching at the remote end.

>       Rx link in pass state, PLL slipped since last read.
>       Events since last read:  Link detection fault  Link status change

These link failure events are likely from power-up.  The status is
"sticky" -- it reports any failure since the last read.

-- 
Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993