On Wed, 9 Feb 2011 07:01:04 -0800, Guenter Roeck wrote: > On Wed, Feb 09, 2011 at 08:02:55AM -0500, Jean Delvare wrote: > > So we have to think about the usage of the EMC devices in question. Do > > you know if any of these are being, or will be, used on PC boards? > > Remember that the purpose of sensors-detect is not to be a universal > > hardware monitoring device detector. It's primarily aimed at PC users. > > Users of embedded devices or exotic hardware should know what hardware > > monitoring devices they have, and these device should be instantiated by > > the kernel directly, rather than being detected from user-space. > > No idea. It wasn't specifically that I need it or that we are going to use > any of the chips. I just got into it after the recent emc driver submission, > and figured that since I tracked down the addresses and chip IDs, I might > as well put my knowledge to some use. > > Also, while you are right about platform data on embedded devices > for the running code, there are additional aspects to consider. > > First, I have often seen that HW specs for embedded devices can more or less > only be seeen as guidance. Not only do chips show up on unexpected addresses, > I have seen instances where chips were not there at all, or undocumented > chips show up. This gets difficult if one has to port linux to some older device, > where the HW designers may no longer be there. Missing chips are easy to deal with (i2c_new_probed_device()) and irrelevant to this discussion anyway. Chips at the wrong address can be spotted with i2cdetect, not need for sensors-detect. For unexpected chips, I agree that sensors-detect can be useful (on the assumption that we are looking for a hardware monitoring device.) > Second, there is the case of people porting Linux to a box which wasn't supposed > to run Linux. > > In both cases, sensors-detect can be a very useful tool to help figuring out > what is actually there. I wouldn't disagree. But as this isn't the primary use case for the script, and probing all addresses has known downsides, you would have to introduce a special flag if you really want to probe addresses which we know aren't used on PC hardware. We really don't want it to happen by default. > How about removing the new addresses from the list ? Would you be ok > with the patch if I do that ? Yes, I would be perfectly OK then. -- Jean Delvare _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
