Network card timeouts

Network card timeouts

[Anders Linden]

Symptom:
Network cards that stops working if they are sent enough data.

The cards I have got problem with this far:
Intel Etherexpress Pro 100+ (some kernels ago).
Davicom Semiconductor, Inc. Ethernet 100/10

Occasion 1:
The kernel version on the computers were 2.4.3. Every person in a single
classroom had problem with Intel cards. They could not even fetch
webpages before their consoles were spammed with a message like: network
card timeout. All computers were Compaq. I dont know which hardware they
had in addition to that. I have also had problems with Intel cards even
after this occasion on other computers.

Occasion 2:
The later card, Davicom, is probably not a well-known card, but
nevertheless, it works like shit in Linux. I am using Redhat 7.1 and the
kernel 2.4.2-2. If I send more than 10M to such a card in an interval of
a second, it just quits working for 5 seconds. The card has no problems
at all in other, third party operating systems, like Windows.


Is it the newest kernels that has theese problems? The first occasion
was exactly after a kernel 2.4.3 has been released, and people I talked
to said that 2.4.2 and network cards were better friends.


Thanks for your attension

/Anders Lindén

Re: Network card timeouts

[Joseph Fannin]

[-- Attachment #1: Type: text/plain, Size: 2410 bytes --]

On Wed, Nov 21, 2001 at 07:16:23PM +0100, Anders Linden wrote:
> Symptom:
> Network cards that stops working if they are sent enough data.
> 
> The cards I have got problem with this far:
> Intel Etherexpress Pro 100+ (some kernels ago).
> Davicom Semiconductor, Inc. Ethernet 100/10

    I have seen very similar symptoms in two ISA eepro cards.  I found that,
if I kept a `ping` process running on the machine with an eepro installed,
the card would work okay, but somewhat slowly.  Otherwise it would just
hang, and had to be restarted to work again.

    (Note: this was a few weeks ago, and I'm not using any of my eepro cards
because of it anymore.  I haven't tried the new(ly maintained) driver.)

    It would be interesting to find that this is a more general problem; a
good bit of the reason why I've never reported it is because my eepro cards
are old and suspect themselves.


> 
> Occasion 1:
> The kernel version on the computers were 2.4.3. Every person in a single
> classroom had problem with Intel cards. They could not even fetch
> webpages before their consoles were spammed with a message like: network
> card timeout. All computers were Compaq. I dont know which hardware they
> had in addition to that. I have also had problems with Intel cards even
> after this occasion on other computers.
> 
> Occasion 2:
> The later card, Davicom, is probably not a well-known card, but
> nevertheless, it works like shit in Linux. I am using Redhat 7.1 and the
> kernel 2.4.2-2. If I send more than 10M to such a card in an interval of
> a second, it just quits working for 5 seconds. The card has no problems
> at all in other, third party operating systems, like Windows.
> 
> 
> Is it the newest kernels that has theese problems? The first occasion
> was exactly after a kernel 2.4.3 has been released, and people I talked
> to said that 2.4.2 and network cards were better friends.
> 
> 
> Thanks for your attension
> 
> /Anders Lind?n
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
Joseph Fannin
jhf@rivenstone.net

"Bull in pure form is rare; there is usually some contamination by data."
    -- William Graves Perry Jr.

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: Network card timeouts

[Jeff Garzik]

I think what people sometimes do not understand is that transmit
timeouts are a generic handler for a NIC- or driver-specific problem

It is a mistake to assume that timeouts are a "common thread" of any
sort.

	Jeff

Re: Network card timeouts

[Tobias Ringstrom]

On Wed, 21 Nov 2001, Anders Linden wrote:

> Occasion 2:
> The later card, Davicom, is probably not a well-known card, but
> nevertheless, it works like shit in Linux. I am using Redhat 7.1 and the
> kernel 2.4.2-2. If I send more than 10M to such a card in an interval of
> a second, it just quits working for 5 seconds. The card has no problems
> at all in other, third party operating systems, like Windows.
> 
> Is it the newest kernels that has theese problems? The first occasion
> was exactly after a kernel 2.4.3 has been released, and people I talked
> to said that 2.4.2 and network cards were better friends.
> 
> Thanks for your attension

Please try 2.4.14 and let me know if you still see this problem with the
dmfe driver.

/Tobias

Re: Network card timeouts

[Alan Cox]

> The later card, Davicom, is probably not a well-known card, but
> nevertheless, it works like shit in Linux. I am using Redhat 7.1 and th=
> e
> kernel 2.4.2-2. If I send more than 10M to such a card in an interval o=

Davicom is a bad tulip clone. It has a (not very good) davicom provided
driver in 2.4.2 or you can use tulip or the updated davicom provided dfme
driver in newer 2.4

Re: Network card timeouts

[Jeff Garzik]

Alan Cox wrote:
> 
> > The later card, Davicom, is probably not a well-known card, but
> > nevertheless, it works like shit in Linux. I am using Redhat 7.1 and th=
> > e
> > kernel 2.4.2-2. If I send more than 10M to such a card in an interval o=
> 
> Davicom is a bad tulip clone. It has a (not very good) davicom provided
> driver in 2.4.2 or you can use tulip or the updated davicom provided dfme
> driver in newer 2.4

tulip should work better IMHO for all davicom tulip clones -except- the
older ones which require doing a software crc for each packet.  For
those, the performance penalty is too high to integrate into the
mainline tulip driver.  So my advice is usually "try tulip, then
fallback to dmfe if software crc is required"

-- 
Jeff Garzik      | Only so many songs can be sung
Building 1024    | with two lips, two lungs, and one tongue.
MandrakeSoft     |         - nomeansno