>> First of all, which version of lm_sensors are you using? > >As above, version 2.8.7 . I downloaded it yesterday using : Ooops, sorry. I missed the information at first, my bad. >> If you are using the latest version if sensors-detect AND you use >> devfs AND you have i2c-dev loaded AND sensors-detect still complains, >> then this is likely to be a problem with your bus driver (i2c-viapro). > >When I insmod i2c-viapro, devfs doesn't create any new devices. However, >after reading your response this morning, I removed all the i2c devices >I created, rmmodded all i2c modules except i2c-core and i2c-dev and >reran sensors-detect. The same message appeared: > >Can't open /dev/i2c/0 > >but I continued with the config process, ran sensors and voila! works >perfectly. Hmm, strange. I guess that your hardware monitoring chipset is located on the ISA bus (thus using i2c-isa instead of i2c-viapro)? I don't use devfs myself, but I think that it is OK not to have /dev/i2c/0 with only i2c-dev loaded. Device files should be created as you load bus driver modules. Could someone with devfs confirm this? I don't think you told us which (exact) kernel version you were using, did you? >Now that I have it working, lm_sensors will be installed on all of a >client's servers (+- 200) and will be used to provide valuable, hitherto >inaccessible information about the state of their cpus and fans. If this >is successful, as I'm sure it will be, it is intended to roll out >lm_sensors to some of our other clients too. Do all of these servers have similar hardware? Remember that your experience with lm_sensors depends much on the exact hardware configuration. >Please accept an apology for my hasty assumption and thanks for a >great and essential tool! I don't see much you should apologize about. Although it didn't seem to prevent you from using lm_sensors, there still is something bad with your hardware/devfs/lm_sensors/kernel combination and we better find out what it is. I admit I don't have many ideas at the moment, mainly because I don't use devfs myself so there is no chance I could reproduce the problem. At least, one thing we could do: detect devfs at the beginning of mkdev.sh, and bail out with a message instead of silently creating useless device files? Thanks, Jean Delvare
