Re: PCI driver module unload race?

[Greg KH]

On Sat, Mar 08, 2003 at 08:47:14PM +0100, Petr Vandrovec wrote:
> On Sat, Mar 08, 2003 at 11:12:37AM -0800, Greg KH wrote:
> > On Sat, Mar 08, 2003 at 10:47:49AM +0000, Russell King wrote:
> > > Hi,
> > > 
> > > What prevents the following scenario from happening?  It's purely
> > > theoretical - I haven't seen this occuring.
> > > 
> > > - Load PCI driver.
> > > 
> > > - PCI driver registers using pci_module_init(), and adds itself to sysfs.
> > > 
> > > - Hot-plugin a PCI device which uses this driver.  sysfs matches the PCI
> > >   driver, and calls the PCI drivers probe function.
> > 
> > Ugh, yes you are correct, I can't believe I missed this before.
> > 
> > How does this patch look?
> 
> Bad...
> 
> What are you trying to solve?

The case where while probe() is called, the module is unloaded.
Same thing for remove().

That's all.

> After driver calls pci_unregister_driver,
> it is sure that there are no other users of this pci driver.

Sure, but that's not the case of what we are protecting here.  We want
pci_unregister_driver() (which is usually called from the module_exit()
function), to not be called if we are in the middle of calling either
probe() or release().

Do you have a way of protecting the race that is described by Russell
here that differs from my patch?

thanks,

greg k-h