System hanging when two network cards are active

I've recently installed Mandrake 10 on a newe HP system (with an Athlon
X/3000 64 bit as CPU, 512 MB memory). The system is meant to serve as a
router, so I added a second Internet card. Anytime there is any traffic
on the second card, however, the system hangs.

I"m using the 8139too module which came with the card as a driver. I
tried downloading the rtl8139 module, but I can't get it to compile.

The funny thing is that the problem only seems to be present when the X
Window manager is active. As a simple test, I do a ping -f
192.168.100.245 (a machine on the second network) -- this runs forever
(well, hours at least) without a problem if I boot in secure mode,
without the window manager, but hangs in 2 or 3 seconds if the Window
manager is running.

Does anyone have any idea where I could start looking.

--
James Kanze
Conseils en informatique orientée objet/
Beratung in objektorientierter Datenverarbeitung
9 place Sémard, 78210 St.-Cyr-l'École, France +33 (0)1 30 23 00 34

["Followup-To:" header set to comp.os.linux.networking.]
On 2004-09-11, James Kanze <(E-Mail Removed)> wrote:

> I've recently installed Mandrake 10 on a newe HP system (with an Athlon
> X/3000 64 bit as CPU, 512 MB memory). The system is meant to serve as a
> router, so I added a second Internet card. Anytime there is any traffic
> on the second card, however, the system hangs.
>
> I"m using the 8139too module which came with the card as a driver. I
> tried downloading the rtl8139 module, but I can't get it to compile.
>
> The funny thing is that the problem only seems to be present when the X
> Window manager is active. As a simple test, I do a ping -f
> 192.168.100.245 (a machine on the second network) -- this runs forever
> (well, hours at least) without a problem if I boot in secure mode,
> without the window manager, but hangs in 2 or 3 seconds if the Window
> manager is running.
>
> Does anyone have any idea where I could start looking.

First have a look at the interrupts in /proc/interrupts

I suspect your seconds cards shares an irq with something used by
X, e.g. graphics or mouse.

Or get some other cards. The 8139 is a piece of crap.
Try Intel Pro/100 or 3com 3c905, used cards costs around 5 Euro/$.

Thinking of it, don't waste time looking at the interrupts and buy
some Intels

--
Marco Dieckhoff
icq# 22243433
GPG Key 0x1A6C95BA -- http://www.frankonia-brunonia.de/keys

Marco Dieckhoff <(E-Mail Removed)> writes:

|> On 2004-09-11, James Kanze <(E-Mail Removed)> wrote:

|> > I've recently installed Mandrake 10 on a newe HP system (with an
|> > Athlon X/3000 64 bit as CPU, 512 MB memory). The system is meant
|> > to serve as a router, so I added a second Internet card. Anytime
|> > there is any traffic on the second card, however, the system
|> > hangs.

|> > I"m using the 8139too module which came with the card as a driver.
|> > I tried downloading the rtl8139 module, but I can't get it to
|> > compile.

|> > The funny thing is that the problem only seems to be present when
|> > the X Window manager is active. As a simple test, I do a ping -f
|> > 192.168.100.245 (a machine on the second network) -- this runs
|> > forever (well, hours at least) without a problem if I boot in
|> > secure mode, without the window manager, but hangs in 2 or 3
|> > seconds if the Window manager is running.

|> > Does anyone have any idea where I could start looking.

|> First have a look at the interrupts in /proc/interrupts

|> I suspect your seconds cards shares an irq with something used by
|> X, e.g. graphics or mouse.

Bingo. I sort of suspected something of the sort, but I had no idea
where to look to find out. (It's been some twenty years since I last
programmed at that level.) It shares an interrupt with the nvidia
graphics card, which definitly isn't a good thing.

However, now I'm really worried, because a number of other things share
other interrupts. The output of "cat /proc/interrupts" is:

CPU0
0: 10063881 IO-APIC-edge timer
1: 29379 IO-APIC-edge i8042
2: 0 XT-PIC cascade
3: 0 IO-APIC-edge ehci_hcd
5: 1028992 IO-APIC-edge ohci_hcd, ohci1394, eth0
8: 1 IO-APIC-edge rtc
10: 108479 IO-APIC-edge ohci_hcd, NVidia nForce3
11: 622478 IO-APIC-edge eth1, nvidia
12: 105094 IO-APIC-edge i8042
14: 30014 IO-APIC-edge ide0
15: 208673 IO-APIC-edge ide1
NMI: 0
LOC: 10064070
ERR: 0
MIS: 0

Interrupts 5,10 and 11 all seem shared. About the only difference I can
see is that for 5 and 10, everything sharing is on the mother board.
But can it be normal that eth0 shares interrupts with what I think are
my USB controllers or the firewire? (Eth0 is the nForce3 MCP Networking
Adapter on my motherboard.) Maybe it's just working because by chance
I'm not using those ports at present. (I've got a printer on a USB
port, but maybe that's the ehci, which looks OK.)

|> Or get some other cards. The 8139 is a piece of crap. Try Intel
|> Pro/100 or 3com 3c905, used cards costs around 5 Euro/$.

|> Thinking of it, don't waste time looking at the interrupts and buy
|> some Intels

Even new, they shouldn't be that expensive.

Still, I'd be curious as to how to configure the interrupts anyway.
Just in case the other "conflicts" cause problems in the future.

--
James Kanze
Conseils en informatique orientée objet/
Beratung in objektorientierter Datenverarbeitung
9 place Sémard, 78210 St.-Cyr-l'École, France +33 (0)1 30 23 00 34

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
NotDashEscaped: You need GnuPG to verify this message

In comp.os.linux.networking James Kanze <(E-Mail Removed)> suggested:
> Marco Dieckhoff <(E-Mail Removed)> writes:
> |> On 2004-09-11, James Kanze <(E-Mail Removed)> wrote:
[..]

> |> First have a look at the interrupts in /proc/interrupts
[..]

> However, now I'm really worried, because a number of other things share
> other interrupts. The output of "cat /proc/interrupts" is:

> 5: 1028992 IO-APIC-edge ohci_hcd, ohci1394, eth0
[..]

Doesn't matter, Linux shares IRQ just fine:

11: 214709987 XT-PIC 3ware Storage Controller, ohci_hcd, eth0, eth2, NVidia nForce2

From my box, there are even 5 devices on IRQ 11, works like a
charm.

[..]
> |> Or get some other cards. The 8139 is a piece of crap. Try Intel
> |> Pro/100 or 3com 3c905, used cards costs around 5 Euro/$.

Those are not bus-master cards and I wouldn't suggest them
running in a server, but for some desktop they are fine. However
they tend to brake sometimes quite fast, which might be the
reason for the OP's problem. I'd simply exchange the card and see
if the problem persists, if there are no hints in /var/log/*
about other problems.

[..]

--
Michael Heiming (GPG-Key ID: 0xEDD27B94)
mail: echo (E-Mail Removed) | perl -pe 'y/a-z/n-za-m/'
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFBQx2VAkPEju3Se5QRAqDcAKDaHPmbN2e/yBRa6A9dWvfR83/UtACgvd6D
aRqcHvzxARL8u7+zla7CJXo=
=UqML
-----END PGP SIGNATURE-----

James Kanze wrote:
> I've recently installed Mandrake 10 on a newe HP system (with an Athlon
> X/3000 64 bit as CPU, 512 MB memory). The system is meant to serve as a
> router, so I added a second Internet card. Anytime there is any traffic
> on the second card, however, the system hangs.
>
> I"m using the 8139too module which came with the card as a driver. I
> tried downloading the rtl8139 module, but I can't get it to compile.
>
> The funny thing is that the problem only seems to be present when the X
> Window manager is active. As a simple test, I do a ping -f
> 192.168.100.245 (a machine on the second network) -- this runs forever
> (well, hours at least) without a problem if I boot in secure mode,
> without the window manager, but hangs in 2 or 3 seconds if the Window
> manager is running.
>
> Does anyone have any idea where I could start looking.
>

I find really odd what's happening to you, but I think it's a problem of
the driver you are using. You should recompile the kernel version you
are using from the kernel.org sources and use the rtl8139 driver you
have compiled. The rtl8139 driver it's a very reliable one, or this is
my appreciation until now, so if it doesn't work that way maybe there's
some kind of incompatibility between your card and your graphics card.


--

Jose Maria Lopez Hernandez
Director Tecnico de bgSEC
(E-Mail Removed)
bgSEC Seguridad y Consultoria de Sistemas Informaticos
http://www.bgsec.com
ESPAÑA

The only people for me are the mad ones -- the ones who are mad to live,
mad to talk, mad to be saved, desirous of everything at the same time,
the ones who never yawn or say a commonplace thing, but burn, burn, burn
like fabulous yellow Roman candles.
-- Jack Kerouac, "On the Road"

Michael Heiming <michael+(E-Mail Removed)> writes:

|> [..]
|> > |> Or get some other cards. The 8139 is a piece of crap. Try
|> > |> Intel Pro/100 or 3com 3c905, used cards costs around 5 Euro/$.

|> Those are not bus-master cards and I wouldn't suggest them running
|> in a server, but for some desktop they are fine. However they tend
|> to brake sometimes quite fast, which might be the reason for the
|> OP's problem. I'd simply exchange the card and see if the problem
|> persists, if there are no hints in /var/log/* about other problems.

Following up: I did change cards; it was too late for me to get to the
computer shops in Paris to get one of the cards verified by Mandrake,
but a bookshop near here that is open until 8PM was able to sell me a
Netgear card, and it now seems to work.

I still don't see the relationship with the fact that things worked when
windowing wasn't active -- probably the card was reacting when it
shouldn't, when the system asked the source of the interrupt, and one of
the screen driver interrupts got lost.

The bothersome part is that the card I replaced was sold to me as new.
I'd originally scavanged an card out of an old machine; it had sort of
worked for a very short time, then stopped working. The initialization
of eth1 said failed, however, so I pretty much figured out that the card
might be at fault. Having replaced it with a "new" one, and getting
completely different symptoms, threw me, however.

--
James Kanze
Conseils en informatique orientée objet/
Beratung in objektorientierter Datenverarbeitung
9 place Sémard, 78210 St.-Cyr-l'École, France +33 (0)1 30 23 00 34

Michael Heiming <michael+(E-Mail Removed)> writes:

[...]
|> Those are not bus-master cards and I wouldn't suggest them running
|> in a server, but for some desktop they are fine. However they tend
|> to brake sometimes quite fast, which might be the reason for the
|> OP's problem. I'd simply exchange the card and see if the problem
|> persists, if there are no hints in /var/log/* about other problems.

I also forgot the say thanks.

I still have one question: why is it that I have to disable the firewall
(and routing) to ping machines on the local network? Or for that
matter, to connect to an HTTP server on one?

--
James Kanze
Conseils en informatique orientée objet/
Beratung in objektorientierter Datenverarbeitung
9 place Sémard, 78210 St.-Cyr-l'École, France +33 (0)1 30 23 00 34

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
NotDashEscaped: You need GnuPG to verify this message

[ Followup-To comp.os.linux.networking ]

In comp.os.linux.networking James Kanze <(E-Mail Removed)> suggested:
> Michael Heiming <michael+(E-Mail Removed)> writes:

[..]

> I still have one question: why is it that I have to disable the firewall
> (and routing) to ping machines on the local network? Or for that
> matter, to connect to an HTTP server on one?

Ping might be disabled by your firewall setup script, changing the
value of /proc/sys/net/ipv4/icmp_echo_ignore_all. If you can't
connect via IP/hostname to a local server, your firewall setup
and/or routing is bogus. Hard to tell without knowing anything
about your setup, local boxes should have a local route and don't
use your gateway at all to reach each others, you might want to
use 'traceroute' to check where those packets go.

--
Michael Heiming (GPG-Key ID: 0xEDD27B94)
mail: echo (E-Mail Removed) | perl -pe 'y/a-z/n-za-m/'
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFBQ246AkPEju3Se5QRAozbAJ9heCEFHcHScbCIXb1Vhj t4jp/ZpQCdEpex
i9C4r2tjxJPVmIDkW6skY0s=
=TH3V
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
NotDashEscaped: You need GnuPG to verify this message

In comp.os.linux.networking Michael Heiming <michael+(E-Mail Removed)> suggested:

> In comp.os.linux.networking James Kanze <(E-Mail Removed)> suggested:
>> Michael Heiming <michael+(E-Mail Removed)> writes:

> [..]

>> I still have one question: why is it that I have to disable the firewall
>> (and routing) to ping machines on the local network? Or for that
>> matter, to connect to an HTTP server on one?

> Ping might be disabled by your firewall setup script, changing the
> value of /proc/sys/net/ipv4/icmp_echo_ignore_all. If you can't

Which would of course only affect pinging the router directly.

> connect via IP/hostname to a local server, your firewall setup
> and/or routing is bogus. Hard to tell without knowing anything
> about your setup, local boxes should have a local route and don't
> use your gateway at all to reach each others, you might want to
> use 'traceroute' to check where those packets go.

--
Michael Heiming (GPG-Key ID: 0xEDD27B94)
mail: echo (E-Mail Removed) | perl -pe 'y/a-z/n-za-m/'
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFBQ27PAkPEju3Se5QRApl3AJ9i6p7zyntQi49rOUQQjU VSq4TmNgCgunPl
HoZeoG1ciO/0/Dc0/dfsRjU=
=K0hF
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
NotDashEscaped: You need GnuPG to verify this message

[ Followup-To comp.os.linux.hardware ]

In comp.os.linux.networking Michael Heiming <michael+(E-Mail Removed)> suggested:
> In comp.os.linux.networking James Kanze <(E-Mail Removed)> suggested:
>> Marco Dieckhoff <(E-Mail Removed)> writes:
>> |> On 2004-09-11, James Kanze <(E-Mail Removed)> wrote:

[..]
>> |> Or get some other cards. The 8139 is a piece of crap. Try Intel
>> |> Pro/100 or 3com 3c905, used cards costs around 5 Euro/$.

> Those are not bus-master cards and I wouldn't suggest them

Just for the archives, of course intel eepro and probably 3com
3c905 cards are bus-master cards, those rtl8139 aren't, that's
what I meant to say but didn't wrote.

--
Michael Heiming (GPG-Key ID: 0xEDD27B94)
mail: echo (E-Mail Removed) | perl -pe 'y/a-z/n-za-m/'
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFBRLaEAkPEju3Se5QRAg25AKCtpAvGjiGPxLQZcvVTPq XtPGioHQCgzQFb
8dYdtfxOyfjUgYWggkh1uAA=
=4Tq5
-----END PGP SIGNATURE-----