[eepro100] Yet another receive hang-up.

[Steve Egbert]

I've been using Ether Pro 100s on some sixty odd PC platforms of various
motherboards in my company.  I've noticed that despite NIC swapouts, the
exact same Hangup occurs on certain motherboards.

I've staggered my eepro100.o drivers' version on those affected machines to
no avail.

The only known workaround is a power reset as opposed to reboot.  I seem to
recall awhile back during Linux 2.3.11 that "reboot" command has been
changed to only do soft reboot (anyone confirm this?).  Perhaps, I should
modify reboot to do hard and see how this goes.

The problem seems to have gone away for motherboards made after 1997
(loosely tabulating my memory).  Seems to be a hard (or soft) reset line
issue (inadequate debounce)?  I'm beginning to think that the Intel
controller requires a minimum hard reset for NIC reusability.

Am I on the right track?

Steve.

> -----Original Message-----
> From: Mike Zeller [mailto:mike.zeller@crosskeys.com]
> Sent: Tuesday, March 20, 2001 3:42 PM
> To: eepro100@scyld.com
> Subject: [eepro100] Yet another receive hang-up.
> 
> 
> Hi,
> 
> I am using four etherpro 100s with the eepro100 v1.09j-t 
> driver and Redhat
> 6.2.  The version of the receive hang-up I see occurs on reboot.
> 
> The driver output from the initial load show that the driver 
> detected the
> "Receive Hangup Bug".  However, judging (guessing) from the 
> code, this is an
> occasional hang-up during normal operation.  As I say, the 
> version I see
> occurs on reboot.
> 
> The problem occurs intermittently.  About one reboot in four. 
>  I currently
> have the machine rebooting over night and looking at its 
> netstats to see if
> it has any data. I have a ping coming in to the card so that there is
> traffic.  No traffic after about a minute means I have a 
> problem, reboot and
> try again.  When I detect the problem on reboot, I do mii-diag eth1.
> Looking at the output of this command I can tell that it has 
> negotiated with
> the link partner so it is certainly receiving at a hardware/electronic
> level.
> 
> I also then try to ping out from the affected machine.  This 
> seems to cure
> the problem because the card will reset with a "Transmit 
> Timed Out" message.
> 
> Looking at the code for the eepro100 driver I notice that the transmit
> timeout code does a lot more than what mii-diag -R does.
> 
> Has anyone else seen a receive hang-up that only occurs on reboot?
> 
> Does anyone have a way to detect this problem without having 
> guaranteed
> incoming traffic on the card?
> 
> Thanks in advance.
> 
> Mike Zeller
> Crosskeys Systems, Inc.
> mike.zeller@crosskeys.com
> 610-530-2000 x 3125
> 
> 
> 
> 
> _______________________________________________
> eepro100 mailing list
> eepro100@scyld.com
> http://www.scyld.com/mailman/listinfo/eepro100
>