Subject: Re: de(4) and transmit underflow
To: None <port-i386@NetBSD.ORG>
From: Jason Thorpe <thorpej@nas.nasa.gov>
List: port-i386
Date: 04/21/1998 14:46:38
On Tue, 21 Apr 1998 14:42:50 -0700 
 Chuck <nospam@best.com> wrote:

 > Sometimes you have to share interrupts.  This system has an IDE disk
 > and four quad-port ethernet cards.
 > 
 > Below is `dmesg` from the system which is having problems and 
 > producing this error message:
 >   Apr 21 14:21:34 smdls4 /netbsd: de0: abnormal interrupt: transmit underflow
 >   (raising TX threshold to 96|256)

Don't worry about this too much...  It basically means that the kernel
wasn't feeding the Tulip buffers fast enough for 100Mbit transfers.  By
raising the TX threshold, the driver gives the Tulip more buffers for
every TX complete interrupt...

 > 
 > It has the following things on each IRQ:
 >    1: pc0
 >    3: com1
 >    4: com0
 >    5: com2 isapnp0
 >    6: fdc0
 >    7: sb0
 >    9: de1 de4 de10 de15 de18
 >   10: de0 de3 de6 de7 de8 de9 de13 de14 de16 de17 wss0
 >   11: de2 de5 de11 de12 de19
 >   12: pms0
 >   14: wdc0
 >   15: wdc1
 > 
 > The "transmit underflow" error happens only on de0.  I do not know why
 > so many devices are put on irq10 instead of being spread more evenly.

You can probably alter this configuration via the BIOS.  But it's
probably an artifact of how the BIOS swizzles the interrupts through
the PCI-PCI bridges...

If you tell the BIOS to reserve IRQ 10 for the ISA (i.e. PnP WSS :-),
then it will probably more evenly divide them between 9 and 11.

BTW, that machine looks pretty impressive :-)

Jason R. Thorpe                                       thorpej@nas.nasa.gov
NASA Ames Research Center                            Home: +1 408 866 1912
NAS: M/S 258-5                                       Work: +1 650 604 0935
Moffett Field, CA 94035                             Pager: +1 415 428 6939