On Mon, Aug 23, 2004 at 03:08:29PM +0400, Evgeniy Polyakov wrote: > On Sun, 2004-08-22 at 12:48, Jean Delvare wrote: > > > > > #define SIO_LDN_GPIO 0x07 /* General-Purpose I/O (GPIO) Ports */ > > > > > > > > This is PC87366-specific and doesn't belong to that driver. > > > > > > No, I think it just should be extended and include any other > > > defines(for temperature, voltage and others). Although they may live > > > in drivers for those logical devices. Driver for PC87366 may use it... > > > I will move it there after it will be created (actually separated to > > > support several main SuperIO chips at once). > > > > I'm not sure I get everything you mean here (especially the last > > sentence). At any rate, the logical device usage is > > manufacturer-dependant and, in most cases, chip dependant. And when two > > different chips each have a logical device for, say, temperature > > monitoring, the way it is organized "inside" each logical device will be > > different. So there's no benefit in adding labels or classes to each > > logical device of each chip, since we will never be able to use them. > > > > I insist that the superio driver should be mostly chip-independant. The > > only part which will not be is the Super-I/O enter sequence. Since the > > PC8736x chips don't need any, you may not be familiar with it. Most > > Super-I/O chips need a special sequence to be written to them before > > they activate. Since the superio driver won't know what it is looking > > for before it finds it, it'll probably have to test all known keys, then > > the client driver will ask not only for a given chip ID but for a key + > > chip ID pair. This also means that you will have to store the (variable > > length) key in the superio structure. > > I'm absolutely agree here. > I mean above that when I will split sc.c into generic part and pc8736* > part I will move those definitions into the header for pc8736* driver. That sounds fine. > As I described above we will not have code duplication and will have > ability to add/remove logical devices in run-time. Which is a good thing. > > Greg, I much would like to hear you about this. > > I think we will come to logical conclusion when he will back from > vocation. > Or not. Heh, it looks good to me. Oh, and I suggest putting this in drivers/superio for a location. thanks, greg k-h
