Re: [PATCH] 2.6.6 synclinkmp.c

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

On Tue, Jun 01, 2004 at 04:25:30PM -0500, Paul Fulghum wrote:
> On Tue, 2004-06-01 at 15:57, Russell King wrote:
> > If pci_register_driver fails, the driver is not, repeat not left
> > registered.  Therefore it must not be unregistered after failure
> > to register.
> 
> You are right. The specific problem I was trying to
> fix is when no hardware is detected. I looked at other
> PCI drivers (char/epca.c and net/eepro100.c) and which call
> pci_unregister_driver if pci_register_driver returns <= 0
> and indicates that pci_register_device returns the number
> of pci devices detected. I now see that the two drivers I
> looked at are broken. (bad luck that)
> 
> After looking at the source for pci_register_device(),
> if no devices are detected, then it still returns 1.
> 
> I will rework the patches against synclink.c/synclinkmp.c
> to only call pci_unregister_device() if init fails
> (such as when no devices are detected)
> *and* the call to pci_register_device() succeeds.

Don't arrange for the driver to unload if it doesn't detect anything.
2.6 has dynid support so that the user can load your driver and assign
it extra PCI vendor/device IDs at run time - which won't work if you've
forced failure when nothing is found.

-- 
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