On Wed, 16 Jul 2008 09:44:02 -0600, David Hubbard wrote: > Hi Jim, > > On Wed, Jul 16, 2008 at 1:25 AM, Jean Delvare <khali at linux-fr.org> wrote: > >> >> David, > >> >> do you have a preference wrt fn-ptr ? and if so, a single fn with 1 arg > >> >> : 0,1,2 for enter/exit/query ? > >> >> do you have any use for a flag bit ? is there generality in it ? > >> > >> I think we could just use the same enter and exit sequences for > >> everything, and not make it customizable. Does that sound OK? > > > > No. We can have a universal exit sequence, but enter sequences are > > per-chip (or actually per chip family). That's even one part of the > > device identification (lets us differentiate between two chips with the > > same device ID register value.) > > I think a fn-ptr would be fine. The function could take as an arg > enter/exit/query (0, 1, 2) and a pointer to a byte array with offset, > value, offset, value pairs. So I see the function having 3 args: > > void (* fn-ptr)(int which, char * sequence, int sequence_max_length); I'm lost. I thought that we had just agreed that the exit sequence would be universal? Also, what is the "query" sequence supposed to be? Then I don't get the prototype of the function. Who would be calling it? And who would provide the parameters? The enter sequence is a set of number being written to a port. It is a per-chip thing (or per-family) thing. So there are 2 possibilities: * For each known chip or family, you store the enter sequence in an array, and the superio driver iterates over it. This is the approach taken in both isadump and sensors-detect. OR * For each known chip or family, you store a function, which the superio driver calls. The prototype of such a function would be something like: void (*enter)(int index_port, int data_port); This is more flexible, but at this point I have no reason to believe that this is needed. Hmmm, I thought I had said I would let someone else work on this... -- Jean Delvare
