Re: nfe0 watchdog timeout

On Mon, May 12, 2008 at 06:12:59PM +0200, Pascal Hofstee wrote:
 > Hi,
 > 

Hi,

 > I recently decided to switch my CURRNET from i386 over to amd64 and
 > have since noticed an occasional occurance of the following phenomenon:
 > 
 > Every couple of days the nfe0 interface (nVidia nForce MCP55 Ethernet
 > Controller ... Vitesse VSC6801 PHY) enters a mode from which it simply
 > will not recover.
 > 
 > When it happens the console keeps spewing out
 > 
 > nfe0: watchdog timeout (missed Tx interrupts) -- recovering
 > 
 > When this happens i have to reboot to get the interface working again.
 > (I'll probably build another custom kernel soon and disable the nfe
 > driver so i can try if a kldunload/kldload cycle will do the trick).
 > 

Can you see link state change reports of nfe(4) in your dmesg file?
If you find any entries related with nfe(4) please show them to me.
It may have a couple of link UP/DOWN messages and watchdog timeouts.

 > Considering i have never observed this phenomenon when running the
 > exact same system with CURRENT/i386 i wonder if this problem may be
 > some sort of 64-bit issue.
 > 

Probably not.

 > Any ideas and/or solutions would be appreciated.
 > 

When you know nfe(4) is not working anymore, would you check Rx MAC
is still able to see incoming packets by invoking tcpdump on
nfe(4) interface?

It would be even better I can see verbosed boot messages related
with nfe(4).

 > With kind regards,
 >   Pascal Hofstee
 > 
-- 
Regards,
Pyun YongHyeon