PCI-Scan Hangup ...

PCI-Scan Hangup ...

[Kianusch Sayah Karadji]

Hi!

There I have a Soekris bord with some weird PCI Handup while booting
Linux during PCI-Scan .

# lspci

00:00.0 Host bridge: Cyrix Corporation PCI Master
00:12.0 ISA bridge: National Semiconductor Corporation: Unknown device 0510
00:12.1 Bridge: National Semiconductor Corporation: Unknown device 0511
00:12.2 IDE interface: National Semiconductor Corporation SCx200 IDE (rev 01)
00:12.5 Bridge: National Semiconductor Corporation: Unknown device 0515
00:13.0 USB Controller: Compaq Computer Corporation ZFMicro Chipset USB (rev 08)

There was a patch on 2.4.x which worked fine.

Now I tried to patch Kernel 2.6.x - well I'm no programmer
- I had no idea where to put it ... so I started putting some debug
messages into some files ... and suddenly including a printk(".") into
drivers/pci/probe.c fixed the whole PCI-Hangup ??? - That was not what the
2.4.x Patch did.

Hmmm ...

I've no idea why that printk fixes the Problem nor if it's a good idea to
have that printk there ...

Regards
  Kianusch

--- linux-2.6.3/drivers/pci/probe.c     Wed Feb 18 04:58:46 2004
+++ linux-2.6.3_geode/drivers/pci/probe.c       Thu Feb 19 14:00:52 2004
@@ -656,8 +656,11 @@
        DBG("Scanning bus %02x\n", bus->number);

        /* Go find them, Rover! */
-       for (devfn = 0; devfn < 0x100; devfn += 8)
+       for (devfn = 0; devfn < 0x100; devfn += 8) {
                pci_scan_slot(bus, devfn);
+               printk(".");
+        }
+       printk("\n");

        /*
         * After performing arch-dependent fixup of the bus, look behind

Re: PCI-Scan Hangup ...

[Bartlomiej Zolnierkiewicz]

On Thursday 19 of February 2004 14:35, Kianusch Sayah Karadji wrote:
> Hi!

Hi,

> There I have a Soekris bord with some weird PCI Handup while booting
> Linux during PCI-Scan .
>
> # lspci
>
> 00:00.0 Host bridge: Cyrix Corporation PCI Master
> 00:12.0 ISA bridge: National Semiconductor Corporation: Unknown device 0510
> 00:12.1 Bridge: National Semiconductor Corporation: Unknown device 0511
> 00:12.2 IDE interface: National Semiconductor Corporation SCx200 IDE (rev
> 01) 00:12.5 Bridge: National Semiconductor Corporation: Unknown device 0515
> 00:13.0 USB Controller: Compaq Computer Corporation ZFMicro Chipset USB
> (rev 08)
>
> There was a patch on 2.4.x which worked fine.
>
> Now I tried to patch Kernel 2.6.x - well I'm no programmer
> - I had no idea where to put it ... so I started putting some debug
> messages into some files ... and suddenly including a printk(".") into
> drivers/pci/probe.c fixed the whole PCI-Hangup ??? - That was not what the
> 2.4.x Patch did.
>
> Hmmm ...
>
> I've no idea why that printk fixes the Problem nor if it's a good idea to
> have that printk there ...

printk() adds some delay -> this delay fixes your problem, but I dunno why.

Re: PCI-Scan Hangup ...

[Greg KH]

On Thu, Feb 19, 2004 at 02:35:59PM +0100, Kianusch Sayah Karadji wrote:
> Hi!
> 
> There I have a Soekris bord with some weird PCI Handup while booting
> Linux during PCI-Scan .

Can you try to determine what is hanging in a non-modified 2.6.3 kernel?

thanks,

greg k-h

Re: PCI-Scan Hangup ...

[Kianusch Sayah Karadji]

Hi!

> > There I have a Soekris bord with some weird PCI Handup while booting
> > Linux during PCI-Scan .
>
> Can you try to determine what is hanging in a non-modified 2.6.3 kernel?

The very last thing I see when booting ist:

   PCI: Probing PCI hardware (bus 0)

That's it.  But it seems to be a BIOS Proble - I just found out that the
Soekris-Home-Page states:

[...]

 The SC1100 has a bug where certain PCI config cycle conbinations can
 cause the processor to lock up. comBIOS version 1.21 or newer reprogram
 the chipset to fix the problem.

[...]

I'll try to upgrade my comBIOS this afternoon, and let you know if 2.6.x
boot's fine or not.

Regards
   Kianusch

Re: PCI-Scan Hangup ...

[Kianusch Sayah Karadji]

> The very last thing I see when booting ist:
>
>    PCI: Probing PCI hardware (bus 0)
>
> That's it.  But it seems to be a BIOS Proble - I just found out that the
> Soekris-Home-Page states:
>
> [...]
>
>  The SC1100 has a bug where certain PCI config cycle conbinations can
>  cause the processor to lock up. comBIOS version 1.21 or newer reprogram
>  the chipset to fix the problem.
>
> [...]
>
> I'll try to upgrade my comBIOS this afternoon, and let you know if 2.6.x
> boot's fine or not.

The latest Bios seems to fix that problem.  Nativ kernel 2.6.3 boots
without any problem.

Thanx for Your help

Kianusch