[9fans] rtl8139 ether driver

[9fans] rtl8139 ether driver

[okamoto]

I'm now forced to use rtl8139 nic for my new Auth/CPU server (
it's for 3ed not 4ed. I cannot update our system now, because it's
being used for the platform for our application, and the semester is
still going here).   it reboots once a day.  :-)

We had been using dual Pen III on SuperMicro motherboard, but it
crushed about a month ago.   Then, I changed that server to a Pen 4
mother board (1.6GHz) of AOpen AX4B Pro-533 which has on board
nic chip of Realtek 8100BL.   It seems to work fine, however, it may
have problem as above.

When I read the /sys/src/9/pc/ether8139.c, I found the lines of
"Error recovery for the various over/under-flow conditions may need work"
phrase.   Does my problem relate to this remark?

Kenji

Re: [9fans] rtl8139 ether driver

[okamoto]

[-- Attachment #1: Type: text/plain, Size: 298 bytes --]

This problem was solved by replacing the onboard Realtek 8100BL chip
to Intel's i82557 chipped card (Intel XPRES-PRO100).   Now I'm using this
Pen 4 1.6GHz + AOpen AX4B Pro-533 mother happily as our Auth/CPU
server for Release 3 Plan 9. It must be same for Release 4, I believe. :-)

Kenji


[-- Attachment #2: Type: message/rfc822, Size: 2370 bytes --]

From: okamoto@granite.cias.osakafu-u.ac.jp
To: 9fans@cse.psu.edu
Subject: [9fans] rtl8139 ether driver
Date: Wed, 26 Jun 2002 10:45:59 +0900
Message-ID: <20020626014705.4A7E0199E3@mail.cse.psu.edu>


I'm now forced to use rtl8139 nic for my new Auth/CPU server (
it's for 3ed not 4ed. I cannot update our system now, because it's
being used for the platform for our application, and the semester is
still going here).   it reboots once a day.  :-)

We had been using dual Pen III on SuperMicro motherboard, but it
crushed about a month ago.   Then, I changed that server to a Pen 4
mother board (1.6GHz) of AOpen AX4B Pro-533 which has on board
nic chip of Realtek 8100BL.   It seems to work fine, however, it may
have problem as above.

When I read the /sys/src/9/pc/ether8139.c, I found the lines of
"Error recovery for the various over/under-flow conditions may need work"
phrase.   Does my problem relate to this remark?

Kenji

Re: [9fans] rtl8139 ether driver

[okamoto]

Ok!
I set up another slow machine (Pen 133 :-) as an Auth server, and
converted the Pen 4 1.6GHz machine to a terminal.   It now seems
to work fine for a terminal, which make me trouble.

Do you have any good progra to make this machine panic as that
when it's used as CPU server which I reported before...
Any hints.

Kenji  -- sorry Charles, I sent wrong mail to you

Re: [9fans] rtl8139 ether driver

[okamoto]

Actually, it becomes more often today.  I must have experienced
more than several times in a worktime.   I may be better to check it
for a terminal machine, but not CPU server machine...

At first, I doubted power problem, because we use 100 V power for
designed 115 V power.   It usually does not cause problem.  However,
I experienced fail before when the switching power supply would be
poor.   In this time, 300W power supply is used.   The machine only has
a 1.6GHz Pen 4, 51MB memory, a CDROM, a floppy drive and a VGA card.

I tried slidac to make up the power from 100 to 115 V, but it did not solve
the problem either.

>you don't say whether it reboots without comment or simply because
>it has lost the file system link.

 From the file server's log, I know only it's link to file server hanged-up.
I've never been at front of our CPU server when it lost the link to the
file server.  So, I can see only lines of messages after new reboot on a
console of CPU/Auth server.

>i'd first start checking that memmove parameters (eg, count)
>are always correct.

As Jim know, kernel level checking would exceed my ability, sorry...
If someone kindly contact me detailed, I may be able to try it.   Anyway,
It'd be better to use that machine for a terminal at first, and I'll do it later.
And I'm planing to parchase a new motherboard using dual Athron
for our CPU server.   Yes, I got a good news from a shop who can still sell
that kind of motherboard.

Kenji

Re: [9fans] rtl8139 ether driver

[forsyth]

>>It may hard to read the error message even if I make it run in debugging
>>mode, because it occures very rarely, and when it happenes, that machine
>>reboots automatically, ..

you don't say whether it reboots without comment or simply because it has lost the
file system link.  i suppose the former, because otherwise prints in the driver
would otherwise work (though you'd need to stop automatic reset to read
them all).  i'd first start checking that memmove parameters (eg, count)
are always correct.

Re: [9fans] rtl8139 ether driver

[okamoto]

[-- Attachment #1: Type: text/plain, Size: 803 bytes --]

Thank you very much, Jim.

It may hard to read the error message even if I make it run in debugging
mode, because it occures very rarely, and when it happenes, that machine
reboots automatically, and then, that neccessary mesage could be hidden
among lots of others...

When I experienced system crush of our CPU server, I tried to find some
other dual Pentium machine for it.   I failed to do it, because there may be
no Pen III dual motherboard in the market, except only of SuperMicro's
Xeon motherbord, which is very very expensive, and beyond my scope
of fund.  (I was told some $1,200~1,300 for that motherboard!!)  So, I had
to choose Pen 4 mother for that server.   This may also be true for most
of Plan 9 users now.  How we can enjoy multi-CPU nature of Plan 9. Alas!

Kenji


[-- Attachment #2: Type: message/rfc822, Size: 2925 bytes --]

From: jmk@plan9.bell-labs.com
To: 9fans@cse.psu.edu
Subject: Re: [9fans] rtl8139 ether driver
Date: Tue, 25 Jun 2002 21:58:34 -0400
Message-ID: <7b2da9c3f29b1a4d0c352b4167dd4757@plan9.bell-labs.com>

On Tue Jun 25 21:48:22 EDT 2002, okamoto@granite.cias.osakafu-u.ac.jp wrote:
>
> I'm now forced to use rtl8139 nic for my new Auth/CPU server (
> it's for 3ed not 4ed. I cannot update our system now, because it's
> being used for the platform for our application, and the semester is
> still going here).   it reboots once a day.  :-)
>
> We had been using dual Pen III on SuperMicro motherboard, but it
> crushed about a month ago.   Then, I changed that server to a Pen 4
> mother board (1.6GHz) of AOpen AX4B Pro-533 which has on board
> nic chip of Realtek 8100BL.   It seems to work fine, however, it may
> have problem as above.
>
> When I read the /sys/src/9/pc/ether8139.c, I found the lines of
> "Error recovery for the various over/under-flow conditions may need work"
> phrase.   Does my problem relate to this remark?
>
> Kenji

I only have an 8139C and I've not ever seen an error of any kind on it,
hence my comment about work may be required if any error conditions happen.
Someone else asked me a similar question some months ago, and I suggested
putting in debugging prints in the driver wherever an error is detected.
That way you will see something if the card causes any problem and then
reboots the machine.

There may also be problems with the P4 processor, we don't have any here
to run on.

--jim

Re: [9fans] rtl8139 ether driver

[jmk]

On Tue Jun 25 21:48:22 EDT 2002, okamoto@granite.cias.osakafu-u.ac.jp wrote:
>
> I'm now forced to use rtl8139 nic for my new Auth/CPU server (
> it's for 3ed not 4ed. I cannot update our system now, because it's
> being used for the platform for our application, and the semester is
> still going here).   it reboots once a day.  :-)
>
> We had been using dual Pen III on SuperMicro motherboard, but it
> crushed about a month ago.   Then, I changed that server to a Pen 4
> mother board (1.6GHz) of AOpen AX4B Pro-533 which has on board
> nic chip of Realtek 8100BL.   It seems to work fine, however, it may
> have problem as above.
>
> When I read the /sys/src/9/pc/ether8139.c, I found the lines of
> "Error recovery for the various over/under-flow conditions may need work"
> phrase.   Does my problem relate to this remark?
>
> Kenji

I only have an 8139C and I've not ever seen an error of any kind on it,
hence my comment about work may be required if any error conditions happen.
Someone else asked me a similar question some months ago, and I suggested
putting in debugging prints in the driver wherever an error is detected.
That way you will see something if the card causes any problem and then
reboots the machine.

There may also be problems with the P4 processor, we don't have any here
to run on.

--jim