Broken nForce2 IDE module loading via hotplug

Broken nForce2 IDE module loading via hotplug

[Juerg Billeter]

Hi

(please cc me)

The sata_nv patch[1] (merged in 2.6.11-rc4) to enable future NVIDIA SATA
pci ids catches all NVIDIA pci devices with the ide class. This breaks
automatic module loading for e.g. nForce2 ide controllers and thereby
renders nForce systems loading modules already in initramfs/initrd via
hotplug/coldplug non-bootable.

I don't know what solutions are possible besides reverting. Is it
somehow possible to influence the order of the modules.pcimap file, i.e.
moving the generic matching lines below the more specific ones?

Thanks for any hints,

Juerg

[1]
http://www.mail-archive.com/bk-commits-24@vger.kernel.org/msg00112.html
-- 
Juerg Billeter <juerg@paldo.org>

Re: Broken nForce2 IDE module loading via hotplug

[Jeff Garzik]

Juerg Billeter wrote:
> Hi
> 
> (please cc me)
> 
> The sata_nv patch[1] (merged in 2.6.11-rc4) to enable future NVIDIA SATA
> pci ids catches all NVIDIA pci devices with the ide class. This breaks
> automatic module loading for e.g. nForce2 ide controllers and thereby
> renders nForce systems loading modules already in initramfs/initrd via
> hotplug/coldplug non-bootable.
> 
> I don't know what solutions are possible besides reverting. Is it
> somehow possible to influence the order of the modules.pcimap file, i.e.
> moving the generic matching lines below the more specific ones?

It might be fair to add a kernel config option to disable the generic entry.

	Jeff

RE: Broken nForce2 IDE module loading via hotplug

[Andrew Chew]

I think the kernel config option to disable the generic entry is a good
idea.  Jeff, want me to submit a patch for this? 

-----Original Message-----
From: Jeff Garzik [mailto:jgarzik@pobox.com] 
Sent: Friday, May 13, 2005 11:05 AM
To: Juerg Billeter
Cc: Andrew Chew; linux-kernel@vger.kernel.org
Subject: Re: Broken nForce2 IDE module loading via hotplug

Juerg Billeter wrote:
> Hi
> 
> (please cc me)
> 
> The sata_nv patch[1] (merged in 2.6.11-rc4) to enable future NVIDIA
SATA
> pci ids catches all NVIDIA pci devices with the ide class. This breaks
> automatic module loading for e.g. nForce2 ide controllers and thereby
> renders nForce systems loading modules already in initramfs/initrd via
> hotplug/coldplug non-bootable.
> 
> I don't know what solutions are possible besides reverting. Is it
> somehow possible to influence the order of the modules.pcimap file,
i.e.
> moving the generic matching lines below the more specific ones?

It might be fair to add a kernel config option to disable the generic
entry.

	Jeff

Re: Broken nForce2 IDE module loading via hotplug

[Jeff Garzik]

Andrew Chew wrote:
> I think the kernel config option to disable the generic entry is a good
> idea.  Jeff, want me to submit a patch for this? 

Please.

	Jeff

RE: Broken nForce2 IDE module loading via hotplug

[Andrew Chew]

Actually, it occurred to me that it would be better if the determination
to use the generic entry can be run-time specified (via a kernel option
flag, for example).  All the kernel config option accomplishes is a
cleaner way to disable the generic entry at compile time.

In any case, it'd be interesting to figure out the EXACT root of the
problem.  Let me spend a few days looking into it.

-----Original Message-----
From: Jeff Garzik [mailto:jgarzik@pobox.com] 
Sent: Monday, May 16, 2005 12:59 PM
To: Andrew Chew
Cc: Juerg Billeter; linux-kernel@vger.kernel.org
Subject: Re: Broken nForce2 IDE module loading via hotplug

Andrew Chew wrote:
> I think the kernel config option to disable the generic entry is a
good
> idea.  Jeff, want me to submit a patch for this? 

Please.

	Jeff

Re: Broken nForce2 IDE module loading via hotplug

[Jeff Garzik]

Andrew Chew wrote:
> Actually, it occurred to me that it would be better if the determination
> to use the generic entry can be run-time specified (via a kernel option
> flag, for example).  All the kernel config option accomplishes is a
> cleaner way to disable the generic entry at compile time.
> 
> In any case, it'd be interesting to figure out the EXACT root of the
> problem.  Let me spend a few days looking into it.

Any progress?

Re: Broken nForce2 IDE module loading via hotplug

[Bartlomiej Zolnierkiewicz]

On 5/25/05, Jeff Garzik <jgarzik@pobox.com> wrote:
> Andrew Chew wrote:
> > Actually, it occurred to me that it would be better if the determination
> > to use the generic entry can be run-time specified (via a kernel option
> > flag, for example).  All the kernel config option accomplishes is a
> > cleaner way to disable the generic entry at compile time.
> >
> > In any case, it'd be interesting to figure out the EXACT root of the
> > problem.  Let me spend a few days looking into it.
> 
> Any progress?

It would be nice to get it fixed before 2.6.12...
Even if it means reverting the patch...

Bartlomiej