2.4.14 fails to boot on a MediaGX

2.4.14 fails to boot on a MediaGX

[Adam Harvey]

Hi

I've compiled 2.4.14 on a MediaGX running at 133 MHz that I use for a 
gateway, and am having some difficulty getting it to boot. The last three 
lines of output are as follows:

Working around Cyrix MediaGX virtual DMA bugs.
CPU: Cyrix MediaGX 3x Core/Bus Clock
Checking 'hlt' instruction... OK.

At this point it locks hard and doesn't respond to Alt-SysRq or Ctrl-Alt-Del. 
2.4.10 (the only other 2.4 kernel I have handy) doesn't work either, failing 
at the same point, although 2.2.19 does boot and run normally.

I'm not on LKML, so please CC me if any further information is required or to 
test any patches.

Adam

Re: 2.4.14 fails to boot on a MediaGX

[Alan Cox]

> I've compiled 2.4.14 on a MediaGX running at 133 MHz that I use for a 
> gateway, and am having some difficulty getting it to boot. The last three 
> lines of output are as follows:

What processor did you compile for. 

> Working around Cyrix MediaGX virtual DMA bugs.
> CPU: Cyrix MediaGX 3x Core/Bus Clock
> Checking 'hlt' instruction... OK.

2.4.12-ac was working on my 233Mhz MediaGX. I suspect 2.4.14 does too. I'll
give it a test tho

Re: 2.4.14 fails to boot on a MediaGX

[Adam Harvey]

On Wed, 14 Nov 2001 07:29, Alan Cox wrote:
> > I've compiled 2.4.14 on a MediaGX running at 133 MHz that I use for a
> > gateway, and am having some difficulty getting it to boot. The last three
> > lines of output are as follows:
>
> What processor did you compile for.

Compiled for 486. I remember reading that 586 optimisations (used to?) break 
the kernel on MediaGXs.

>
> > Working around Cyrix MediaGX virtual DMA bugs.
> > CPU: Cyrix MediaGX 3x Core/Bus Clock
> > Checking 'hlt' instruction... OK.
>
> 2.4.12-ac was working on my 233Mhz MediaGX. I suspect 2.4.14 does too. I'll
> give it a test tho

Thanks for the help.

Adam

Re: 2.4.14 fails to boot on a MediaGX

[Alan Cox]

> > > Working around Cyrix MediaGX virtual DMA bugs.
> > > CPU: Cyrix MediaGX 3x Core/Bus Clock
> > > Checking 'hlt' instruction... OK.
> >
> > 2.4.12-ac was working on my 233Mhz MediaGX. I suspect 2.4.14 does too. I'll
> > give it a test tho
> 
> Thanks for the help.

Ok on my box it boots fine. As an experiment can you build a kerne with no
PCI support (Im not saying it'll be useful production wise but it will tell
me if its a PCI issue)

Re: 2.4.14 fails to boot on a MediaGX

[Adam Harvey]

On Wed, 14 Nov 2001 19:17, Alan Cox wrote:
> Ok on my box it boots fine. As an experiment can you build a kerne with no
> PCI support (Im not saying it'll be useful production wise but it will tell
> me if its a PCI issue)

And, interestingly enough, it does boot with PCI switched off.

Given that it's a somewhat unusual motherboard, I'm pretty willing to just 
put an ISA network card in there and write it off as flaky hardware, but it's 
odd that it boots under 2.2 with PCI support on.

Adam

Re: 2.4.14 fails to boot on a MediaGX

[Christer Weinigel]

In article <01111513115600.00812@blackbox.local> you write:
>On Wed, 14 Nov 2001 19:17, Alan Cox wrote:
>> Ok on my box it boots fine. As an experiment can you build a kerne with no
>> PCI support (Im not saying it'll be useful production wise but it will tell
>> me if its a PCI issue)
>
>And, interestingly enough, it does boot with PCI switched off.
>
>Given that it's a somewhat unusual motherboard, I'm pretty willing to just 
>put an ISA network card in there and write it off as flaky hardware, but it's 
>odd that it boots under 2.2 with PCI support on.

I have had some problem with the PCI direct probe on a MediaGX CPU, 
when the probe looked at the CX5530 companion chip the box would
reboot.  Could you try to only use the PCI BIOS method instead?

PCI support (CONFIG_PCI) [N/y/?] (NEW) y 
  PCI access mode (BIOS, Direct, Any) [Any] (NEW) BIOS
  defined CONFIG_PCI_GOBIOS

  /Christer

-- 
"Just how much can I get away with and still go to heaven?"

Re: 2.4.14 fails to boot on a MediaGX

[Peter Wächtler]

Adam Harvey wrote:
> 
> On Wed, 14 Nov 2001 19:17, Alan Cox wrote:
> > Ok on my box it boots fine. As an experiment can you build a kerne with no
> > PCI support (Im not saying it'll be useful production wise but it will tell
> > me if its a PCI issue)
> 
> And, interestingly enough, it does boot with PCI switched off.
> 
> Given that it's a somewhat unusual motherboard, I'm pretty willing to just
> put an ISA network card in there and write it off as flaky hardware, but it's
> odd that it boots under 2.2 with PCI support on.
> 

What kind of motherboard and peripherals are you using?
AKAIK the mediagx only supports one external pci busmaster and only
on a specific "device".

We are using mediaGXm on ETX and had problems with the wiring of the
TVIA CyberPro5050 (combined video+audio controller, audio as busmaster)

Then there are also tweaks in the PCI IDE configuration for mediaGX.
Did you enable that?

Re: 2.4.14 fails to boot on a MediaGX

[Adam Harvey]

On Thu, 15 Nov 2001 17:19, Peter Wächtler wrote:
> What kind of motherboard and peripherals are you using?
> AKAIK the mediagx only supports one external pci busmaster and only
> on a specific "device".

I don't know the make of the motherboard without actually pulling the box 
apart. The system was designed and used in hotels for Internet access in 
rooms to start with... I picked it up after they were replaced about a year 
ago.

It's unusual, because it has a PCI chipset, yet only has a couple of ISA 
slots. I've been enabling PCI mainly out of habit and in case anything on the 
motherboard was using the bus (apparently not), but I wouldn't be surprised 
if the PCI chipset is basically broken, since it wasn't really needed.

>
> We are using mediaGXm on ETX and had problems with the wiring of the
> TVIA CyberPro5050 (combined video+audio controller, audio as busmaster)

Could be applicable, given that video and audio are handled on-board. I might 
have to open her up tomorrow and have a poke around.

>
> Then there are also tweaks in the PCI IDE configuration for mediaGX.
> Did you enable that?

I've had them on and off, they don't affect the booting problem.

I'll build a kernel using the BIOS access mode (as suggested by Alan) 
tomorrow and see if that clears it up.

Adam