2003-03-15 02:21:57

by war

[permalink] [raw]
Subject: Broadcom BCM5702 Major Problems

When compiled in statically it does not work.
Loading the module after the entire system loads, then ifconfigging works ok.
Hm, maybe PCI sharing must be turned off, I read this from somewhere on google?
I am also running latest stable kernel 2.4.21-pre5-7.


2003-03-15 03:43:46

by Andreas Dilger

[permalink] [raw]
Subject: Re: Broadcom BCM5702 Major Problems

On Mar 15, 2003 02:32 -0000, [email protected] wrote:
> When compiled in statically it does not work.
> Loading the module after the entire system loads, then ifconfigging works ok.
> Hm, maybe PCI sharing must be turned off, I read this from somewhere on google?
> I am also running latest stable kernel 2.4.21-pre5-7.

We are having similar problems with the Broadcom driver, and would be
very interested in finding a solution.

Cheers, Andreas
--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://www-mddsp.enel.ucalgary.ca/People/adilger/

2003-03-15 04:24:40

by war

[permalink] [raw]
Subject: Re: Broadcom BCM5702 Major Problems

ok, thanks for the information, I am not the on, *
I am not an isolated case :)

2003-03-15 05:30:14

by war

[permalink] [raw]
Subject: Re: Broadcom BCM5702 Major Problems

Found the bug.
Well, for me anyway.
When I have this configuration option turned on:

[*] IO-APIC support on uniprocessors

Everything goes to hell.

Even the 3COM I stuck in wants to get routed to IRQ 18 and you cannot even force it to use anotheR IRQ.

This option:

[*] Local APIC support on uniprocessors

Works fine.

It is the second option that royally screws over the system.

2003-03-15 09:23:55

by Paul Rolland

[permalink] [raw]
Subject: Re: Broadcom BCM5702 Major Problems

Hello,

We have some BCM running pretty well with both IO-APIC and
Local APIC with kernel 2.4.19...

Which one are you talking about ?

Regards,
Paul

> Found the bug.
> Well, for me anyway.
> When I have this configuration option turned on:
>
> [*] IO-APIC support on uniprocessors
>
> Everything goes to hell.
>
> Even the 3COM I stuck in wants to get routed to IRQ 18 and
> you cannot even force it to use anotheR IRQ.
>
> This option:
>
> [*] Local APIC support on uniprocessors
>
> Works fine.
>
> It is the second option that royally screws over the system.

2003-03-15 13:05:08

by war

[permalink] [raw]
Subject: Re: Broadcom BCM5702 Major Problems

[*] IO-APIC support on uniprocessors

2003-03-15 13:37:18

by Manfred Spraul

[permalink] [raw]
Subject: Re: Broadcom BCM5702 Major Problems

war wrote:

>[*] IO-APIC support on uniprocessors
>
I think the subject is a bit misleading:
There seems to be a problem with interrupt routing if he enables IO-APIC
support, both with a Broadcom nic and a 3com nic.
Either the MP table that is supplied by the bios is incorrect [wouldn't
be a big surprise - I think Linux is the only OS that looks at MP tables
of uniprocessor machines], or the ACPI interpreter did something wrong.

Do you use the latest bios for your motherboard? Which chipset?

--
Manfred




2003-03-15 14:16:19

by war

[permalink] [raw]
Subject: Re: Broadcom BCM5702 Major Problems

Yes I have the latest BIOS , the SiS 655max (6730) I believe.

8X AGP / 6PCI / 4RAM (DDR) / etc

http://www.msicomputer.com/product/detail_spec/product_detail.asp?model=655_Max-FISR

Is the motherboard

2003-03-15 21:10:11

by Dave Jones

[permalink] [raw]
Subject: Re: Broadcom BCM5702 Major Problems

On Sat, Mar 15, 2003 at 02:47:59PM +0100, Manfred Spraul wrote:

> >[*] IO-APIC support on uniprocessors
> >
> I think the subject is a bit misleading:
> There seems to be a problem with interrupt routing if he enables IO-APIC
> support, both with a Broadcom nic and a 3com nic.
> Either the MP table that is supplied by the bios is incorrect [wouldn't
> be a big surprise - I think Linux is the only OS that looks at MP tables
> of uniprocessor machines], or the ACPI interpreter did something wrong.

I've noticed on two seperate testboxes (x86 and x86-64) that 8139too
won't recieve packets unless I boot with 'noapic acpi=off' now.
And yes, it has to be both those options, one on its own of either
doesn't make the problem go away. 100% reproducable.

Doesn't seem to affect any other drivers I've tried though.

Dave