loosing connectivity 3c905b

David Fries dfries@umr.edu
Fri Mar 10 16:05:10 2000

AMD K6-2 cpu,
3Com Corporation 3c905B 100BaseTX [Cyclone] (rev 36),
Matrox Graphics, Inc. MGA G200 AGP (rev 1)

When the computer is receiving data and updating the screen at the same
time networking goes down the tube.  Sometimes it is slow, sometimes it
stopps.  This is with 2.2.14 kernel driver and 2.3.49 kernel driver and the
3com's 3c90x.c driver.  Actually I think it has been as long as I've had
this motherboard, network card, video combination.

It shows up very well when I run a tcp program I wrote that this computer
starts receiving about 11megs a sec and I flip between desktops or
something at the same time, then it will decline networking performance and
stop sometimes.  I have the network card as a module so I unload it, load
it again and it is fine for a while.

motherboard info,
    Host bridge: VIA Technologies, Inc. VT82C597 [Apollo VP3] (rev 4).
    PCI bridge: VIA Technologies, Inc. VT82C598 [Apollo MVP3 AGP] (rev 0).
    IDE interface: VIA Technologies, Inc. VT82C586 IDE [Apollo] (rev 6).

I also get a message between the unloading of the module and loading it
unregister_netdevice: Old style device eth0 leaked(refcnt=1). Wait for

3c59x.c:v0.99H+lk1.0 Feb 9, 2000 The Linux Kernel Team
eth0: 3Com 3c905B Cyclone 100baseTx at 0x6c00,  00:10:5a:09:a9:6e, IRQ 15
  8K byte-wide RAM 5:3 Rx:Tx split, 100baseTX interface.
  Enabling bus-master transmits and whole-frame receives.
it is also the same messages listed on bootup.

once and a while I will get,

NETDEV WATCHDOG: eth0: transmit timed out
eth0: transmit timed out, tx_status 00 status e000.
  Flags; bus-master 1, full 1; dirty 2609161 current 2609177.
  Transmit list 00000000 vs. c13a4290.
  0: @c13a4200  length 800005ea status 000005ea
  1: @c13a4210  length 800005ea status 000005ea
  2: @c13a4220  length 800005ea status 000005ea
  3: @c13a4230  length 800005ea status 000005ea
  4: @c13a4240  length 800005ea status 000005ea
  5: @c13a4250  length 800005ea status 000005ea
  6: @c13a4260  length 800005ea status 000005ea
  7: @c13a4270  length 80000042 status 80000042
  8: @c13a4280  length 80000042 status 80000042
  9: @c13a4290  length 800005ea status 000105ea
  10: @c13a42a0  length 800005ea status 000105ea
  11: @c13a42b0  length 800005ea status 000105ea
  12: @c13a42c0  length 800005ea status 000005ea
  13: @c13a42d0  length 800005ea status 000005ea
  14: @c13a42e0  length 80000042 status 00000042
  15: @c13a42f0  length 800005ea status 000005ea
eth0: Resetting the Tx ring pointer.

Is there any options to turn off whole-frame receives or something?

This is really getting me agrivated, the computer is mostly NFS mounted so
if I don't already have a root prompt that I can unload/load the module the
system is toast and I have to cold boot it.

Suggestions please.

		|      David Fries                |
		|      dfries@umr.edu             |
To unsubscribe send a message body containing "unsubscribe"
to linux-vortex-bug-request@beowulf.org