>Number:         6690
>Category:       kern
>Synopsis:       interrupt handling in ThunderLan driver seems to be broken
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people (Kernel Bug People)
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Dec 30 09:20:00 1998
>Last-Modified:
>Originator:     Matthias Scheler
>Organization:
Matthias Scheler                                http://home.owl.de/~tron/
>Release:        981227
>Environment:
System: NetBSD lyssa.owl.de 1.3I NetBSD 1.3I (LYSSA) #3: Wed Dec 30 10:53:14 GMT 1998 root@lyssa.owl.de:/src/sys/compile/LYSSA i386


>Description:
When trying to download data at a high speed rate to a system using a
ThunderLan chipset based ethernet card the system will freeze sooner
or later and drop into the kernel debugger after a while at location
"_Xintr0+0x6f". Because "trace" will list this location about a dozen
times I suppose the interrupt handler gets called recursively.

It is a Compaq dual port card sitting in a Pentium II based system which
is recognized fine at system startup:

tl0 at pci2 dev 0 function 0
tl0: Compaq Dual Port Netelligent 10/100 TX
tl0: Ethernet address 00:08:c7:a4:71:2f
tl0: interrupting at irq 11
nsphy0 at tl0 phy 1: DP83840 10/100 media interface, rev. 1
nsphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
tlphy0 at tl0 phy 31: ThunderLAN 10baseT media interface, rev. 5
tlphy0: 10base5/AUI

The server is a Pentium based system with a DEC 21140a card running 1.3.3.
The problem can be repeated with FTP and NFS at 100MBit/Sec (very quick)
and 10MBit/Sec (takes a while). It doesn't happen with light network
activity like SMTP or interactive sessions.

>How-To-Repeat:
Connect a ThunderLan chipset based ethernet card to fast ethernet hub and
download large amounts of data from a fast server via FTP or NFS.

>Fix:
Sorry, no idea.
>Audit-Trail:
>Unformatted: