On Wed, Oct 21, 2015 at 09:08:31PM +0300, Or Gerlitz wrote: > The alternative approach to address this which was also running here > in the form of a patch from Ira following a reviewer comment from me, > is the have struct ib_device to contain a struct ib_device_attr member > (potentially a pointer) which is filled by the device driver before > they register themselves with the IB core. No to the pointer idea. > This way there's no need for kmalloc() + ib_query_device() + kfree() > when a device attribute is needed and we don't introduce tens/hundreds > new fields for struct ib_device. I care very little if the name is ibdev->attr.foo or ibdev->foo, that is just bikeshedding. I have a slight preference toward ibdev->foo to match other subsystems. I have no idea why you care so much about something so trivial. Jason -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html