On Mon, 5 Sep 2005, Atsushi Nemoto wrote: > macro> That's how other architectures do this, see e.g. > macro> "arch/alpha/kernel/time.c". Why should we be different, even > macro> for now? > > Please elaborate more ? Do you mean we should implement default > rtc_set_mmss() and take the rtc_lock in it ? Or do you mean we should > take rtc_lock in each board-dependent rtc_set_time/rtc_set_time ? I'm not sure all chips actually require it. Certainly the null function does not, so that spinlock would incur an unnecessary overhead. Therefore yes, it should be board- or chip-dependent. > macro> Also the call is named rtc_set_mmss() for an unknown reason > macro> while all the others have set_rtc_mmss(). > > IIRC, you are (one of) the godfather of the function, aren't you? :-) Hmm, I must have got influenced by rtc_set_time()... Perhaps it wasn't that bad after all and it's all the others that should be fixed instead. ;-) Maciej
