Subject: kern/27096: ex interface wedges, losing interrupts?
To: None <>
From: Andrew Brown <>
List: netbsd-bugs
Date: 09/30/2004 23:55:51
>Number:         27096
>Category:       kern
>Synopsis:       ex interface wedges, losing interrupts?
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Oct 01 03:56:00 UTC 2004
>Originator:     TheMan
>Release:        2.0G
System: NetBSD this 2.0G NetBSD 2.0G (THAT) #341: Mon Jul 26 21:42:06 EDT 2004  andrew@this:/usr/src/sys/arch/i386/compile/THAT i386


my ex0 (3Com 3c905C) wedges the system when put to heavy use, while
"normal usage" (ie, interactive and web traffic) does not.  i tried to
transfer a large (for me, meaning 2-3 gigabytes) amount of data to
another machine via netcat and the machine (my laptop; the ex0 is on
the docking station) became unresponsive shortly thereafter.

i've tried:

	THAT#341, built from 20040722 sources, is fine
	THAT#342, built from 20040917 sources, wedges
	THAT#343, built from 20040925 sources, wedges
	THAT#344, built from 20040925 sources, wedges

the difference between builds 343 and 344 is a "make cleandir" in the
kernel build directory.

i've also tried my url0 interface with no problem, a couple of ep
interfaces (pcmcia cards) with no problem, and a cardbus ex (3Com
3c575B) which also wedges, so this problem seems particular to the ex


start a large data transfer and just wait.  fwiw, i was doing this:

	% cd /usr/pkgsrc
	% tar czf packages | nc server 13246


10 drop to ddb
20 count to five
30 continue
40 kernel prints "ex0: uplistptr was 0"
50 wait for wedge to happen again
60 goto 10