> OK, clearly the device does not have a regular 8-bit addressed register > map. Only sub-addresses 0x00 and 0x03 seem to return actual register > values (assuming a second call to i2cdump would return the same - maybe > not.) Sensors-detect has protection code against devices with no > register map at all, but I can't think of a way to safely detect the > case above. > > I am curious if: > > $ i2cget 6 0x4f 0x00 > $ i2cget 6 0x4f 0x03 > > would be always successful and returning consistent values, or if it > is random. The fact that register 0x00 holds value 0x00 and register > 0x03 holds value 0x03 is rather suspicious, I'd say. They do both return consistent values, actually. 0x00 always returns 0x00 and 0x03 always returns 0x03. > This suggests that the reads done by i2cdump are in fact interpreted as > writes by whatever chip lives at I2C address 0x4f. This is a recurrent > problem with I2C as it only defines directions for messages on the bus, > but no semantics for them. > > You might want play with i2cset. If the device reacts to short writes > (and the description you made of what happened during the dump pretty > much suggests so) then something like: > > $ i2cset 6 0x4f 0x80 > > would write value 0x80 to the chip. And the following command should > reproduce the state at the end of i2cdump: > > $ i2cset 6 0x4f 0xff > > Sensors-detect OTOH reads register values in completely random order so > it is difficult to know which was the last one. > > Probably one value corresponds to the initial state of the chip before > the very first sensors-detect run. If we had access to a similar system > where the problem did not happen, we may be able to get the right value > with: > > $ i2cget 6 0x4f > > > I think it's all better and I have no idea what happened. I can do more > > probes or logs or whatever if that's helpful to other people, though. > > Did you run i2cdump on I2C address 0x2c too? I'm going to blacklist > 0x4f on graphics cards by default, I need to know if I should do the > same for 0x2c too. > Yes, I did 0x2c as well and there were no problems at all. Only the 0x4f address did anythig. Here's the result of the dump: i2cdump 6 0x2c No size specified (using byte-data access) WARNING! This program can confuse your I2C bus, cause data loss and worse! I will probe file /dev/i2c-6, address 0x2c, mode byte Continue? [Y/n] 0 1 2 3 4 5 6 7 8 9 a b c d e f 0123456789abcdef 00: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ 10: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ 20: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ 30: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ 40: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ 50: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ 60: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ 70: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ 80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ 90: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ a0: 01 06 03 01 01 00 ff ff ff ff ff ff ff ff ff ff ?????........... b0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ c0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ d0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ e0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................ f0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 00 ................ > > > I can also offer to create a ticket in our bug tracking tool, to record > > > the issue in case other users are affected, and tweak sensors-detect to > > > be even safer even when the user explicitly asks for DDC channels to be > > > probed. > > > > That might not be a bad idea, but hopefully other people are smarter than I am. > > You'd be surprised and disappointed > > I have created: > http://www.lm-sensors.org/ticket/2392 > > I'll add further discoveries there. > > > Anyway, thanks a lot for looking at my post and helping me out. > > You're welcome, I'm glad if you managed to get the display to a usable > state again, to be honest I did not think we would succeed. > Yeah, I didn't think it would happen, either. Fortunately, it was still usuable, just rather unpleasant. I also set it to 0xbf like you mentioned in your followup and that seems to work just fine. Again, thanks so much. _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors