Troubleshooting network/driver issue (Marvel 88E8001)

Hi all,
Im trying to understand what the cause Im seeing for a network issue
Im experiencing is (strongly suspect its the drivers for the network
card but...).

Hardware in question is a integrated Marvell 88E8001 Gigabit Ethernet
Controller (rev 13) on an Asus A8V Delux motherboard. Its running
SuSE 10 and using the sk98lin driver.

The symptoms are that several services (pop3 mail collection, web
based adinistration tools, perl generated pages served via apache but
NOT normal webpages) will intermitantly pause downloading (an 8-x
second delay), and then sometimes resume, sometimes timeout. What is
strange though is that I can do ftp or http downloads of large files
to or from the server and get a prity constant datarate and dont have
any issues with timeouts. Im seeing this problem on 2 (identical)
machines and am reasonably certain it isnt a hardware issue.

In fact the issue sounds very similar to this post

http://groups.google.co.uk/group/fa....d576786e965719

However manually changing the cards speed down to 10Mbps had no
effect. Persuing the problem I thought Id look at the tcpdump output
for pop3 since its quite easy to replicate the issue and found several
lines similar to the following occuring when the service dropout
occurs.

15:10:43.258919 IP xxxxxxxco.uk.36393 > yyyyyyy.co.uk.pop3: . ack
94067 win 65535
15:10:43.279091 IP xxxxxxxco.uk.36393 > yyyyyyy.co.uk.pop3: . ack
94067 win 65535 <nop,nop,sack 1 {95527:96987}>
15:10:43.283153 IP xxxxxxxco.uk.36393 > yyyyyyy.co.uk.pop3: . ack
94067 win 65535 <nop,nop,sack 1 {95527:98447}>
15:10:43.289235 IP xxxxxxxco.uk.36393 > yyyyyyy.co.uk.pop3: . ack
94067 win 65535 <nop,nop,sack 1 {95527:99555}>
15:10:54.390129 IP yyyyyyy.co.uk.pop3 > xxxxxxxco.uk.36391: P
14115:15575(1460) ack 52 win 5840
15:10:54.408015 IP xxxxxxxco.uk.36391 > yyyyyyy.co.uk.pop3: . ack
18154 win 61497

There could be anything from 1 to 3 lines containing nop,nop,sack 1
before the service would pause/timeout. I will admit I dont understand
the abot tcpdump output though so Im clueless as to what its telling
me...

Im at a loss here to undestand quite what is going on and how to fix
it (apart from putting a new network card in which Id prefer not to
since both machines are colocated).

Can anyone offer any advice on how to continue investigating the
problem or any possible solutions?

Many thanks

Lee

Lee Jackson wrote:
> Hi all,
> Im trying to understand what the cause Im seeing for a network issue
> Im experiencing is (strongly suspect its the drivers for the network
> card but...).
>
> Hardware in question is a integrated Marvell 88E8001 Gigabit Ethernet
> Controller (rev 13) on an Asus A8V Delux motherboard. Its running
> SuSE 10 and using the sk98lin driver.
>
> The symptoms are that several services (pop3 mail collection, web
> based adinistration tools, perl generated pages served via apache but
> NOT normal webpages) will intermitantly pause downloading (an 8-x
> second delay), and then sometimes resume, sometimes timeout. What is
> strange though is that I can do ftp or http downloads of large files
> to or from the server and get a prity constant datarate and dont have
> any issues with timeouts. Im seeing this problem on 2 (identical)
> machines and am reasonably certain it isnt a hardware issue.
>
> In fact the issue sounds very similar to this post
>
> http://groups.google.co.uk/group/fa....d576786e965719
>
> However manually changing the cards speed down to 10Mbps had no
> effect. Persuing the problem I thought Id look at the tcpdump output
> for pop3 since its quite easy to replicate the issue and found several
> lines similar to the following occuring when the service dropout
> occurs.
>
> 15:10:43.258919 IP xxxxxxxco.uk.36393 > yyyyyyy.co.uk.pop3: . ack
> 94067 win 65535
> 15:10:43.279091 IP xxxxxxxco.uk.36393 > yyyyyyy.co.uk.pop3: . ack
> 94067 win 65535 <nop,nop,sack 1 {95527:96987}>
> 15:10:43.283153 IP xxxxxxxco.uk.36393 > yyyyyyy.co.uk.pop3: . ack
> 94067 win 65535 <nop,nop,sack 1 {95527:98447}>
> 15:10:43.289235 IP xxxxxxxco.uk.36393 > yyyyyyy.co.uk.pop3: . ack
> 94067 win 65535 <nop,nop,sack 1 {95527:99555}>
> 15:10:54.390129 IP yyyyyyy.co.uk.pop3 > xxxxxxxco.uk.36391: P
> 14115:15575(1460) ack 52 win 5840
> 15:10:54.408015 IP xxxxxxxco.uk.36391 > yyyyyyy.co.uk.pop3: . ack
> 18154 win 61497
>
> There could be anything from 1 to 3 lines containing nop,nop,sack 1
> before the service would pause/timeout. I will admit I dont understand
> the abot tcpdump output though so Im clueless as to what its telling
> me...
>
> Im at a loss here to undestand quite what is going on and how to fix
> it (apart from putting a new network card in which Id prefer not to
> since both machines are colocated).
>
> Can anyone offer any advice on how to continue investigating the
> problem or any possible solutions?
>
> Many thanks
>
> Lee

The sk98lin driver has been deprecated and is/will be replaced by a driver named sky2. If sky2 is
not in your distro, try using skge, which works with some of the same hardware as sk98lin. It works
for me.

Larry

In article <tEzCf.303244$(E-Mail Removed)>,
(E-Mail Removed) (Larry Finger) writes:

> Lee Jackson wrote:
>
>> Hi all,
>> Im trying to understand what the cause Im seeing for a network issue
>> Im experiencing is (strongly suspect its the drivers for the network
>> card but...).
>>
>> Hardware in question is a integrated Marvell 88E8001 Gigabit Ethernet
>> Controller (rev 13) on an Asus A8V Delux motherboard. Its running
>> SuSE 10 and using the sk98lin driver.
>>
>> The symptoms are that several services (pop3 mail collection, web
>> based adinistration tools, perl generated pages served via apache but
>> NOT normal webpages) will intermitantly pause downloading (an 8-x
>> second delay), and then sometimes resume, sometimes timeout. What is
>> strange though is that I can do ftp or http downloads of large files
>> to or from the server and get a prity constant datarate and dont have
>> any issues with timeouts. Im seeing this problem on 2 (identical)
>> machines and am reasonably certain it isnt a hardware issue.

I have a similar hardare configuration and I've been seeing similar
problems. I usually notice it when Pan is downloading lists of new
messages; it'll pause for about 30 seconds, then resume (presumably
after something has timed out and retried). What I've noticed is that
dmesg returns hundreds of copies of the following message:

eth0: SkYuk2RestartRxBmu failed.

> The sk98lin driver has been deprecated and is/will be replaced by a
> driver named sky2. If sky2 is not in your distro, try using skge,
> which works with some of the same hardware as sk98lin. It works
> for me.

So far I've been busy enough with other things to not worry too much
about it - aside from the occasional pause the system works fine.
But when I get around to it, I'll check out that driver. I did find
that I had to download the latest version of sk98lin before eth0
would work at all...

--
/~\ (E-Mail Removed)lid (Charlie Gibbs)
\ / I'm really at ac.dekanfrus if you read it the right way.
X Top-posted messages will probably be ignored. See RFC1855.
/ \ HTML will DEFINITELY be ignored. Join the ASCII ribbon campaign!

Larry Finger wrote:
> The sk98lin driver has been deprecated and is/will be replaced by a
> driver named sky2. If sky2 is not in your distro, try using skge, which
> works with some of the same hardware as sk98lin. It works for me.


skge works for me too, and the default sk98lin driver did not.

Use Yast2
->Network Devices
->Network Card
->Edit
->Advanced
->Hardware Details
->Module Name
to set it.

On Thu, 26 Jan 2006 16:14:09 GMT, Lee Jackson <(E-Mail Removed)> wrote:

>Hi all,
>Im trying to understand what the cause Im seeing for a network issue
>Im experiencing is (strongly suspect its the drivers for the network
>card but...).
<snip>
>problem or any possible solutions?
>
>Many thanks
>
>Lee
>

Thanks for the responses all, I hadnt realised that driver had been
deprecated, I'll try the suggested alternatives and post if it works
or not.

Lee

On 27 Jan 06 18:10:53 -0800, "Charlie Gibbs" <(E-Mail Removed)>
wrote:

>In article <tEzCf.303244$(E-Mail Removed)>,
>(E-Mail Removed) (Larry Finger) writes:
>
>> Lee Jackson wrote:
>>
>>> Hi all,
>>> Im trying to understand what the cause Im seeing for a network issue
>>> Im experiencing is (strongly suspect its the drivers for the network
>>> card but...).
>>>
>>> Hardware in question is a integrated Marvell 88E8001 Gigabit Ethernet
>>> Controller (rev 13) on an Asus A8V Delux motherboard. Its running
>>> SuSE 10 and using the sk98lin driver.
>>>
>>> The symptoms are that several services (pop3 mail collection, web
>>> based adinistration tools, perl generated pages served via apache but
>>> NOT normal webpages) will intermitantly pause downloading (an 8-x
>>> second delay), and then sometimes resume, sometimes timeout. What is
>>> strange though is that I can do ftp or http downloads of large files
>>> to or from the server and get a prity constant datarate and dont have
>>> any issues with timeouts. Im seeing this problem on 2 (identical)
>>> machines and am reasonably certain it isnt a hardware issue.
>
>I have a similar hardare configuration and I've been seeing similar
>problems. I usually notice it when Pan is downloading lists of new
>messages; it'll pause for about 30 seconds, then resume (presumably
>after something has timed out and retried). What I've noticed is that
>dmesg returns hundreds of copies of the following message:
>
>eth0: SkYuk2RestartRxBmu failed.

Yup sounds like what Im seeing however with zero entries in any logs.
Presumably Ive got my loging levels too low.
>
>> The sk98lin driver has been deprecated and is/will be replaced by a
>> driver named sky2. If sky2 is not in your distro, try using skge,
>> which works with some of the same hardware as sk98lin. It works
>> for me.
>
>So far I've been busy enough with other things to not worry too much
>about it - aside from the occasional pause the system works fine.
>But when I get around to it, I'll check out that driver. I did find
>that I had to download the latest version of sk98lin before eth0
>would work at all...

If you do have a chance to test that before I do let us know how you
got on please.

Regards

Lee