On Tue, 23 Oct 2007 20:40:55 -0400, Mark M. Hoffman wrote: > Hi Jean: > > * Jean Delvare <khali at linux-fr.org> [2007-10-23 23:01:36 +0200]: > > On Tue, 23 Oct 2007 08:02:39 -0400, Mark M. Hoffman wrote: > > > Ren?: there is a series of patches for the f75375s hwmon driver that allow fans > > > to be initialized with platform data. Although I haven't reviewed the series > > > in detail yet, IMO the concept looks OK and it may be appropriate for the Macs > > > also. > > > > > > http://lists.lm-sensors.org/pipermail/lm-sensors/2007-October/021597.html > > > http://lists.lm-sensors.org/pipermail/lm-sensors/2007-October/021598.html > > > http://lists.lm-sensors.org/pipermail/lm-sensors/2007-October/021599.html > > > > I still need to be convinced why this has to be hard-coded in the > > kernel when the same can be done, and fine-tuned, in user-space just > > fine. > > For the same reasons that it would be better still if it were done by the BIOS. > > When the machine's safety is at stake, you initialize the hardware into a sane, > safe state as early as possible. Fine tuning is not the issue here. When you > *know* that the hardware does not come up into a safe state, you should correct > that ASAP. > > Say fsck fails out to a shell during early inits... so the fans run full-on, > big deal the machine needs attention anyway. That sure beats having some > unrelated early userspace problem cause your power supply to catch fire. > > Is that good hardware design? No, but that's irrelevant. > > You are going to have to convince *me* why that patch series is a bad idea. OK, you're right. -- Jean Delvare