> If the driver doesn't provide a .probe, we would still increment
> the refcount of the bridge module. Is that reasonable? I dunno.
>
> If there's no .probe then the device is doing something
> weird, and probably either it doesn't have much to do with a
> particular bridge (i.e. it manages no "real" devices) or
> it'd need to manage its own resources (in which case we could
> easily export vme_bridge_get/put.)
>
> Perhaps then the following would be more appropriate,
> what do you think?
>
> + if (driver->probe) {
> + if (vme_bridge_get(bridge->num))
> + return -ENXIO; /* although this could change, see above comment */
> +
> retval = driver->probe(dev, bridge->num, vme_calc_slot(dev));
> + if (retval)
> + vme_bridge_put(bridge);
> + }
> +
> return retval;
>
> .. and then remember to do
> + if (probe)
> + vme_bridge_put(bridge)
> in vme_bus_remove(), which in your patch is unconditional (correctly
> matching the unconditional get() in vme_bus_probe)
I picked this default behaviour from the PCI driver code (drivers/pci/pci-driver.c):
static int pci_device_probe(struct device * dev)
{
...
pci_dev_get(pci_dev);
error = __pci_device_probe(drv, pci_dev);
if (error)
pci_dev_put(pci_dev);
return error;
}
The __pci_device_probe() function checks if probe is present or not.
> I'm certainly no checkpatch taliban, but guess you probably
> didn't want to send this line change.
Gak! Will cleanup and resend.
--
/manohar
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
