Skip to contentSkip to conversation

2.6.0-test4(-mmX) via-rhine ethernet onboard C3 mini-itx doesn't work

DISCUSSION Danny ter Haar 7 participants 19 messages Sep 3, 2003

lore
Share & tools
lore
#0 lore raw

Subject says all! :-)

2.6.0-test3-mm2 still works (as does 2.4.21-rc2).

vanilla 2.6.0-test4 & test4-mm[45] & the onboard ethernet doesn't seem to work. No kernel panics, it just doesn't work :-(

Both have same ethernet driver:

via-rhine.c:v1.10-LK1.1.19-2.5 July-12-2003 Written by Donald Becker
http://www.scyld.com/network/via-rhine.html
eth0: VIA VT6102 Rhine-II at 0xe800, 00:40:63:ca:6a:c1, IRQ 10.
eth0: MII PHY found at address 1, status 0x786d advertising 05e1 Link 45e1.
eth0: Setting full-duplex based on MII #1 link partner capability of 45e1.

Even manual config (normal is dhcp) doesn't work.

Haven't seen anyone else report this, but this is repeatable and i suspect more people must have experienced this ?!

Machine is running debian-unstable distro.

Danny

Signature · 2 lines
I think so Brain, but why does a forklift 
have to be so big if all it does is lift forks?
#1 lore raw
Quoted 6 lines · Danny ter Haar

Subject says all! :-)

2.6.0-test3-mm2 still works (as does 2.4.21-rc2).

vanilla 2.6.0-test4 & test4-mm[45] & the onboard ethernet doesn't seem to work. No kernel panics, it just doesn't work :-(

Try a kernel without ACPI and/or APIC support.

Roger

#2 lore raw
Quoted 2 lines · Roger Luethi
Quoted 1 line

vanilla 2.6.0-test4 & test4-mm[45] & the onboard ethernet doesn't seem to work.

Try a kernel without ACPI and/or APIC support.

Yup, adding acpi=off "fixed" the problem.
Will wait for "better" times ! ;-)

Zanks!

Danny

Signature · 2 lines
I think so Brain, but why does a forklift 
have to be so big if all it does is lift forks?
#3 lore raw
In reply to Danny ter Haar
Quoted 2 lines · Danny ter Haar

vanilla 2.6.0-test4 & test4-mm[45] & the onboard ethernet doesn't seem to work. No kernel panics, it just doesn't work :-(

It's standard APIC bug that no one care!

Quoted 7 lines

Both have same ethernet driver:

via-rhine.c:v1.10-LK1.1.19-2.5 July-12-2003 Written by Donald Becker
http://www.scyld.com/network/via-rhine.html
eth0: VIA VT6102 Rhine-II at 0xe800, 00:40:63:ca:6a:c1, IRQ 10.
eth0: MII PHY found at address 1, status 0x786d advertising 05e1 Link 45e1.
eth0: Setting full-duplex based on MII #1 link partner capability of 45e1.

I have via_kt-400 on ecs_l7vta, just check my previous post:
#Message-ID: <[email protected]>
#Subject: [BUG] - 2.{4,6}.{22,0-test4} - CONFIG_X86_UP_APIC lack routing on eth

Quoted 1 line

Even manual config (normal is dhcp) doesn't work.

Same to me.

Quoted 2 lines

Haven't seen anyone else report this, but this is repeatable and i suspect more people must have experienced this ?!

I report this bug probably 4 times! And it was since early 2.4.20, then APIC changes, so I can use if.

Quoted 1 line

Machine is running debian-unstable distro.

Slackware GNU/Linux (current).

P.S. I you wont to use eth with that kernel, silmpe uncompile CONFIG_X86_UP_APIC and use APM.

PP.S. Maby you have ATI card? Bug with setfonts and loadkeys on 1 tty it's from time to time :-)

Signature · 1 line
# Damian *dEiMoS* Kołkowski # http://deimos.one.pl/ #
#4 lore raw
Quoted 1 line · Damian Kolkowski

It's standard APIC bug that no one care!

That's a pitty, perhaps none of the developers has (serial-)access to one of such machines and thus noway of checking/fixing?

Quoted 3 lines

I have via_kt-400 on ecs_l7vta, just check my previous post:
#Message-ID: <[email protected]>
#Subject: [BUG] - 2.{4,6}.{22,0-test4} - CONFIG_X86_UP_APIC lack routing on eth

some kernel do work with my hardware. which makes it even stranger!

Quoted 2 lines

P.S. I you wont to use eth with that kernel, silmpe uncompile CONFIG_X86_UP_APIC and use APM.

ok, will do that for now.

Quoted 2 lines

PP.S. Maby you have ATI card? Bug with setfonts and loadkeys on 1 tty it's from time to time :-)

nope, mini-itx has everything on board. I intend to use mine as a firewall with dot1Q vlan capabilities. text-only! ;-)

Danny

Signature · 2 lines
I think so Brain, but why does a forklift 
have to be so big if all it does is lift forks?
#5 lore raw
Quoted 2 lines · Danny ter Haar

That's a pitty, perhaps none of the developers has (serial-)access to one of such machines and thus noway of checking/fixing?

Yep, right, perhaps they has not.

Quoted 1 line

some kernel do work with my hardware. which makes it even stranger!

I use 2.4.22-rc2 (APIC works fine with via-rhine), so I patch -ac3 on that kernel too.

Signature · 1 line
# Damian *dEiMoS* Kołkowski # http://deimos.one.pl/ #
#6 lore raw
In reply to Damian Kolkowski
Quoted 5 lines · Damian Kolkowski
Quoted 2 lines · Danny ter Haar

vanilla 2.6.0-test4 & test4-mm[45] & the onboard ethernet doesn't seem to work. No kernel panics, it just doesn't work :-(

It's standard APIC bug that no one care!

if you enable APIC (and not ACPI) then you start using a different BIOS table for IRQ routing. Several BIOSes have bugs in this table since it's not a table that is generally used by Windows on UP boxes. Saying that it's the kernel's fault is rather unfair; most (if not all) distros for example ship with APIC disabled for this reason. It's nice if it works for you but there's quite a few boxes out there that just can't work.... Don't configure it if you have such a box.

#7 lore raw
Quoted 7 lines · Arjan van de Ven
Quoted 1 line

It's standard APIC bug that no one care!

if you enable APIC (and not ACPI) then you start using a different BIOS table for IRQ routing. Several BIOSes have bugs in this table since it's not a table that is generally used by Windows on UP boxes. Saying that it's the kernel's fault is rather unfair; most (if not all) distros for example ship with APIC disabled for this reason.

Maby you have right...

Quoted 2 lines

It's nice if it works for you but there's quite a few boxes out there that just can't work.... Don't configure it if you have such a box.

Do not be anger Arjan, but it[1] works from time to time, so it is permanently kernel/APIC bug.

Take care.

[1] - via-rhine (eth routing).

Signature · 1 line
# Damian *dEiMoS* Kołkowski # http://deimos.one.pl/ #
#8 lore raw
In reply to Arjan van de Ven
Quoted 4 lines · Arjan van de Ven

if you enable APIC (and not ACPI) then you start using a different BIOS table for IRQ routing. Several BIOSes have bugs in this table since it's not a table that is generally used by Windows on UP boxes. Saying that it's the kernel's fault is rather unfair; most (if not all) distros for

What I've been seeing lately is people complaining it used to work with previous kernels but later ones don't. One possible explanation would be that VIA chip sets (explicitly or through one of their specific properties) used to be blacklisted before (and thus APIC silently disabled).

The reports I get indicate that _something_ in the kernel changed. I suspect it's the ACPI code (best viewed with Intel chipsets), but that's speculation.

Roger

#9 lore raw
Quoted 2 lines · Roger Luethi

What I've been seeing lately is people complaining it used to work with previous kernels but later ones don't.

Exactly. I installed from CD with 2.4.18-something and that seemed to work. It's sort of natural that your upgrade to a newer kernel and since 2.6.0 is to be declared "stable" i thought it would be best if it's known that there are still problems.

Danny

Signature · 2 lines
I think so Brain, but why does a forklift 
have to be so big if all it does is lift forks?
#10 lore raw
In reply to Arjan van de Ven
Quoted 7 lines · Arjan van de Ven

if you enable APIC (and not ACPI) then you start using a different BIOS table for IRQ routing. Several BIOSes have bugs in this table since it's not a table that is generally used by Windows on UP boxes. Saying that it's the kernel's fault is rather unfair; most (if not all) distros for example ship with APIC disabled for this reason. It's nice if it works for you but there's quite a few boxes out there that just can't work.... Don't configure it if you have such a box.

I tried 2.6.0-test4-mm2 on other C3 (bought several units including for in my car ;) and that one simply froze after detecting the partiton table of the ide harddisk. adding acpi=off even made then one move again.

When confuses me if that it works with prior kernels and suddenly fails with later/greater kernels.

Ah well, it's now better known and probably some clever individual will eventually figure it out and post a patch ! ;-)

Danny

Signature · 2 lines
I think so Brain, but why does a forklift 
have to be so big if all it does is lift forks?
#11 lore raw
In reply to Damian Kolkowski
Quoted 5 lines · Damian Kolkowski
Quoted 2 lines · Danny ter Haar

vanilla 2.6.0-test4 & test4-mm[45] & the onboard ethernet doesn't seem to work. No kernel panics, it just doesn't work :-(

It's standard APIC bug that no one care!

It's impossible to have an APIC bug on a C3 board, because the processor simply doesn't have one!

Another respondent today said that disabling ACPI (note ACPI != APIC) solved his problems.

Quoted 5 lines
Quoted 2 lines

Haven't seen anyone else report this, but this is repeatable and i suspect more people must have experienced this ?!

I report this bug probably 4 times! And it was since early 2.4.20, then APIC changes, so I can use if.

I saw your previous report. It was basically information-free and impossible to base any problem analysis on. If you want your bug reports to have a chance of being acted upon, follow the documented bug reporting procedure.

#12 lore raw
Quoted 3 lines · Mikael Pettersson

I saw your previous report. It was basically information-free and impossible to base any problem analysis on. If you want your bug reports to have a chance of being acted upon, follow the documented bug reporting procedure.

Most users are more than happy to provide additional information when asked for it. In my personal experience, the documented bug reporting procdure is a time consuming way of producing lots of information, most of which is irrelevant to the case in question, and which is usually lacking the information _I_ am looking for. For me, it tends to be less work to ask the user for the specific details than to sift through a lengthy standard bug report looking for it. YMMV, of course.

Roger

#13 lore raw
In reply to Mikael Pettersson
Quoted 2 lines · Mikael Pettersson

Another respondent today said that disabling ACPI (note ACPI != APIC) solved his problems.

Yep, 2 different C3's, one with IDE freezes and one with network probs suddenly work fine.

Ain't life beautifull ?

Danny

Signature · 2 lines
I think so Brain, but why does a forklift 
have to be so big if all it does is lift forks?
#14 lore raw
In reply to Mikael Pettersson
Quoted 2 lines · Mikael Pettersson

It's impossible to have an APIC bug on a C3 board, because the processor simply doesn't have one!

The chipset does and not writing PCI_INTERRUPT_LINE and other things properly can cause problems if its directing interrupts to the chipset I/O APIC wrongly not the PIC and thus the CPU

#15 lore raw
Quoted 7 lines · Alan Cox
Quoted 2 lines · Mikael Pettersson

It's impossible to have an APIC bug on a C3 board, because the processor simply doesn't have one!

The chipset does and not writing PCI_INTERRUPT_LINE and other things properly can cause problems if its directing interrupts to the chipset I/O APIC wrongly not the PIC and thus the CPU

You mean PCI initialisation (and thus ACPI) can break the I/O-APIC even though the kernel never actually accesses the I/O-APIC per se? Ok, I see how that could happen.

/Mikael

#16 lore raw
Quoted 3 lines · Mikael Pettersson

You mean PCI initialisation (and thus ACPI) can break the I/O-APIC even though the kernel never actually accesses the I/O-APIC per se? Ok, I see how that could happen.

In paticular onboard VIA stuff uses the upper bits of the PCI_INTERRUPT_LINE register to do IRQ routing as well as the PIN stuff that PCI expects

#17 lore raw
In reply to Danny ter Haar

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

This is also apparent in the 2.4.22 stable kernel .. I wasn't able to get my via-rhine to work at all either.. Instead of turning off acpi I just put in a spare 3com card I had lying around.. Still if this is in both the 2.6 and now the 2.4 kernel 's .. shouldnt someone fix it ?

Quoted 16 lines · Danny ter Haar

Subject says all! :-)

2.6.0-test3-mm2 still works (as does 2.4.21-rc2).

vanilla 2.6.0-test4 & test4-mm[45] & the onboard ethernet doesn't seem to work. No kernel panics, it just doesn't work :-(

Both have same ethernet driver:

via-rhine.c:v1.10-LK1.1.19-2.5 July-12-2003 Written by Donald Becker
http://www.scyld.com/network/via-rhine.html
eth0: VIA VT6102 Rhine-II at 0xe800, 00:40:63:ca:6a:c1, IRQ 10.
eth0: MII PHY found at address 1, status 0x786d advertising 05e1 Link 45e1.
eth0: Setting full-duplex based on MII #1 link partner capability of 45e1.

Quoted 8 lines

Even manual config (normal is dhcp) doesn't work.

Haven't seen anyone else report this, but this is repeatable and i suspect more people must have experienced this ?!

Machine is running debian-unstable distro.

Danny

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQE/VhWeleY/n9G/oZ8RApDNAJwIHkz1zrLERkQ1Y604wcujOYN3HACfbMtW sHrPFwycxKgo5URgqovSGbI=
=247w
-----END PGP SIGNATURE-----

#18 lore raw
Quoted 4 lines · Matt Heler

This is also apparent in the 2.4.22 stable kernel .. I wasn't able to get my via-rhine to work at all either.. Instead of turning off acpi I just put in a spare 3com card I had lying around.. Still if this is in both the 2.6 and now the 2.4 kernel 's .. shouldnt someone fix it ?

Indeed. And your detailed bug reports (as outlined earlier today by Len Brown) are going to help with that.

Roger