>referencing to #1710 and #1714, I figured out that the LM77 >sensor is not compatible with the LM75, though it's similar. Interesting. Our docs say that both chips are compatible (although the LM77 has additional features). I don't have the datasheets handy. Can you describe what makes them incompatible? >Attached is a driver for the LM77 that works for me. It gives >read access to the temperature register (temp1_input) and >read+write access to t_hyst, t_crit, t_high, t_low. The config >register is not made available, nor are the status bits, but >it should be really easy to add these if required. The code is >a shameless copy of the lm75 module -- all credits go to Frodo >Looijaard. > >The detection is a bit problematic: the lm77 module should not >install itself for anything else but an lm77, but the lm75 >driver will recognize an lm77 as an lm75. It could be patched to >check if the chip in question is an lm77 (and skip it if yes), >but I'm not sure if it's appropriate. Ideas are welcome. We usually do not include detection for other chipsets in a driver for exclusion purposes, but if its needed, we may consider it. Another approach (which I like better) would be to add support for the LM77 to the lm75 driver. We know that both chips share a number of attributes, although they do not seem to be compatible. In both cases, we would need a way to differenciate between an LM75 and an LM77. I am not sure we know how to do that (which is why our sensors-detect script doesn't know about the LM77). Do you have a proposal? Even a heuristic would be acceptable, since there aren't that much LM77 out there, so just defaulting to LM75 unless specific conditions are met sounds acceptable. >I have this LM77 sensor on the secondary I2C bus of a National >SC1100 system. The bus is driven by the scx200_acb (stock kernel >2.6.7) module, which I had to patch in order to get it work: on >line 461, the following should be added: > > adapter->class = I2C_CLASS_HWMON; > >-- otherwise the sensors will not try to detect devices on this >bus (thanks Khali). I also had to change the ACCESS.bus base >addresses from 0x820, 0x840 to 0x810, 0x820 (line 46), but YMMV. I will send a patch to Greg KH for the class issue. I don't know anything about the base address though, so I won't do anything about that. >I would be really glad to see this module included in the >lm_sensors package. Well, linux 2.6 drivers don't belong to lm_sensors anyway. But we can get it into the kernel tree (either as a new driver or as a modification of the lm75 driver). What belong to lm_sensors are 2,4 drivers and user-space tools support. Not sure you are interested in userspace since you're on an embedded system? Thanks, Jean "Khali" Delvare
