Re: Flaw in the driver-model implementation of attributes

[Martin Diehl]

On Mon, 16 Jun 2003, Russell King wrote:

> On Mon, Jun 16, 2003 at 10:08:26AM -0700, Greg KH wrote:
> > Then don't let your module unload until _all_ instances of your
> > structures are gone.  You can tell if this is true or not, it's just up
> > to the implementor :)
> 
> Greg, I believe Alan does have a valid concern.  Eg, how is the following
> handled?
> 
> - PCI device driver module is loaded
> - device driver gets handed a pci device
> - device driver attaches a file to the struct device corresponding to the
>   PCI device.

with old procfs one would like to set the owner field of the 
corresponding struct proc_dir_entry and/or file_operations at this point.

> - userspace opens new file (this does not increment the device drivers
>   use count.)

given owner=THIS_MODULE was set, this would bump the module's use count

> - device driver is rmmod'd

and this could never happen while the procfs file (or directory) is still 
referenced

> - device driver removes its references to the pci device
> - device driver unloads
> - user reads from opened file.

Admittedly I haven't looked deeper into sysfs yet, but I was under the
assumption/hope there would be a similar approach to make module 
refcounting working there?

Martin