Re: [PATCH] PCI fixes for 2.6.9

[Russell King <rmk+lkml@arm.linux.org.uk>]

On Fri, Oct 22, 2004 at 04:45:08PM -0700, Greg KH wrote:
> On Wed, Oct 20, 2004 at 09:10:45AM +0100, Russell King wrote:
> > On Tue, Oct 19, 2004 at 03:42:15PM -0700, Greg KH wrote:
> > > ChangeSet 1.1997.37.29, 2004/10/06 12:50:32-07:00, kaneshige.kenji@jp.fujitsu.com
> > > 
> > > [PATCH] PCI: warn of missing pci_disable_device()
> > > 
> > > As mentioned in Documentaion/pci.txt, pci device driver should call
> > > pci_disable_device() when it decides to stop using the device. But
> > > there are some drivers that don't use pci_disable_device() so far.
> > 
> > No.  This is wrong.  There are some classes of devices, notably
> > PCMCIA Cardbus drivers where buggy BIOS means this should _NOT_
> > be done.
> 
> But what happens if you reload that driver and try to enable the device?
> Does that "just work" somehow on this kind of hardware?

Why wouldn't it work?  Surely there are no side effects to trying to enable
an already enabled device - if there are, then wouldn't x86 have problems
where PCI devices were enabled before control was passed to the kernel?

> > There are BIOSen out there which refuse to suspend/resume if the
> > Cardbus bridge is disabled.
> > 
> > It's not that the driver is buggy.  It's that the driver has far
> > more information than the PCI layer could ever have.
> 
> Ugh, I hate broken hardware.  I'll revert this in my next round of pci
> changes (sometime next week.)

It's got nothing to do with hardware.  It's BIOS.

Also, upon re-reading the bug report, the problem was putting the Cardbus
bridge into D3 state, not calling pci_disable_device() for it.  However,
if pci_disable_device() puts the bridge into D3, then we'll hit the same
issue.

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 PCMCIA      - http://pcmcia.arm.linux.org.uk/
                 2.6 Serial core