| 3 | | We have had one report that running sensors-detect on a Sapphire AM2RD790 [http://lists.lm-sensors.org/pipermail/lm-sensors/2008-May/023020.html killed the user's CPU]. Until we find what exactly happened, it is strongly advised that you '''do not run sensors-detect''' on this motherboard (or if you do, make sure to skip the SMBus probing.) To be on the safe side, you should even blacklist the i2c bus driver so that it doesn't load. |
| | 3 | We have had one report that running sensors-detect on a Sapphire AM2RD790 [http://lists.lm-sensors.org/pipermail/lm-sensors/2008-May/023020.html killed the user's CPU]. Apparently there is a memory voltage controller on the SMBus which went mad when probed by sensors-detect. |
| | 4 | |
| | 5 | We have now addressed the problem both in sensors-detect and on the kernel front: |
| | 6 | * In sensors-detect, we have added a heuristic to detect the memory voltage controller in question, and more generally any device using the same, unusual access protocol. When such a device is found, the I2C/SMBus address in question is skipped. The new code went in SVN on May 11th, 2008, and will ship in lm-sensors 3.0.2 (and 2.10.7). |
| | 7 | * As not everybody will be using a safe version of sensors-detect immediately, the i2c-piix4 kernel driver has also been modified to no longer attach to the SMBus on the Sapphire AM2RD790 motherboard. The change went in kernel 2.6.26-rc2, and will be included in kernel 2.6.25.4 as well. Hardware monitoring on this board is implemented by the Super-I/O chip, so you don't need the SMBus for that. |
| | 8 | |
| | 9 | With both changes, using sensors-detect should be safe again. Nevertheless, it is strongly advised that you skip SMBus probing when running sensors-detect on this motherboard. |
