On Tue, Jun 11, 2019 at 08:04:08PM +0200, Enrico Weigelt, metux IT consult wrote: > Hi folks, > > > we've got lots of cases, where ->request_port() / ->release_port() > are just (request|release)_mem_region calls, where the parameter > values are practicall known after the device is initialized (*1). > The only preparation needed for that is changing the individual > drivers to initialize mapsize field on probing (*2). > > I see two practical options for that: > > a) introduce generic helpers, which are assigned as driver callbacks > instead of the currently used driver specific ones. > > b) introduce a flag which tells the serial core to do that. > > > What's your oppinion on that ? Why would this help anything? > If you like, I can prepare some patches and post them here. kernel development is done with actual patches, almost never with "would this be a good idea", as you never know if it will or not until you actually do the work. thanks, greg k-h
