On Tue, May 10, 2011 at 01:00:26PM +0000, KY Srinivasan wrote: > > > > -----Original Message----- > > From: Christoph Hellwig [mailto:hch@xxxxxxxxxxxxx] > > Sent: Tuesday, May 10, 2011 1:24 AM > > To: KY Srinivasan > > Cc: Christoph Hellwig; Greg KH; gregkh@xxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; > > devel@xxxxxxxxxxxxxxxxxxxxxx; virtualization@xxxxxxxxxxxxxx > > Subject: Re: various vmbus review comments > > > > On Mon, May 09, 2011 at 02:56:52PM +0000, KY Srinivasan wrote: > > > I will address this. Greg had a concern about module reference counting > > > and looking at the current code, it did not appear to be an issue. The > > > change you are suggesting will not affect the vmbus core which is what I want > > > to focus on. I will however, fix this issue in the current round of patches I will > > > send out this week. > > > > It very clearly affects the interface between the core and the > > functional drivers. Trying to submit the core without making sure the > > interface is exports works properly is not an overly good idea. > > I must be missing something here. As I look at the block driver (and > this is indicative of other drivers as well); the exit routine - > blkvsc_drv_exit, first iterates through all the devices it manages > and invokes device_unregister() on each of the devices and then > invokes vmbus_child_driver_unregister() which is just a wrapper on > driver_unregister(). So, if I understand you correctly, you want the devices to > persist even if there is no driver bound to them. That's how the Linux driver model should be used, so yes, that is the correct thing to do. thanks, greg k-h _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
