Hi Chunhao, > I finished fixing the point 1 "alarms" and point 3 "fan speeds in > /proc" MDS provided me, and tested it briefly. But I don't think the > "hardware-level" alarm mechanism is better than "software-level", > besides the above reason, I also find such phenomenon: > Step 1: > #sensors: > VCoreA: +1.47 V (min = +1.40 V, max = +1.60 V) > > Step 2: > modify the low limit into 1.5V, then run #sensors -s: > > Step 3: > VCoreA: +1.47 V (min = +1.50 V, max = +1.60 V) ALARM > > Step 4: > VCoreA: +1.47 V (min = +1.50 V, max = +1.60 V) > The ALARM will disappear when run "sensors" again, although the > measured value is larger than low limit. Same as when I mentioned > above > > Step 5: > When I modify the low limit back from 1.5 V into 1.4V, run #sensors > -s, Then run #sensors, I find the "ALARM" appears again, while the > measured value is between the limits at this time: > VCoreA: +1.47 V (min = +1.40 V, max = +1.60 V) ALARM > It's not reasonable, don't you think so? Sounds like an "interrupt mode" to me (as opposed to "comparator mode"). The alarm bits raise on state change and wear off when read. In comparator mode you would obtain what you'd expect (alarm bit always reflecting the result of the comparison). If the W83792D can be reprogramed in comparator mode AND this does NOT affect any physical output, then just do that. If the alarm bits DO affect an output, then you should not change the mode, which means that you cannot use hardware alarms if the chip is in interrupt mode. As a side note, same is true for several other drivers, which would need to be fixed. > My another question is: > There are 4 files which have been modified by me this time: > kernel/chips/w83792d.c > lib/chips.c > lib/chips.h > prog/sensors/chips.c > Since I can NOT access lm_sensors CVS here, If I need to commit them, > Need I generate another patch to lm_sensors-2.9.0, and send the patch > to you? Or just you the above 4 files? Get a CVS snapshot here: http://secure.netroedge.com/~lm78/archive/lm_sensors-daily.tar.gz And provide a patch against that. Thanks, -- Jean Delvare
