Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Noisy ipmi0 after bootup? (really other kernel noise)

2008/9/5 Paul Goyette <>:

> There's a separate check earlier that checks if ipmi returned valid sensor
> data.  This code is specifically checking to see if the limits were
> successfully requested and received.  It's perfectly reasonable that the
> limits might not be received, in which case the sensor data should be
> assumed to be within limits.  Since the data has previously been found to be
> valid, return SVALID here is the correct thing to do.

I remember something about having those sensors with SVALID, but didn't
know it was the same code...

> However, I did find another error, just a few lines further down!
> It seems that the ipmi_test_threshold() routine checks for both a high limit
> and a low limit, and returns zero or non-zero.  But it doesn't differentiate
> between a high-limit-exceeded and a low-limit-exceeded. The calling code
> (just a few lines down from the above patch) assumes that it must be an
> upper-limit failing.
> I'm suspecting that in this situation, we're actually detecting that the
> fans are running at a speed which is below the critical low-limit and we
> should therefore be returning CRITUNDER (or WARNUNDER) rather than the *OVER
> states.  The attached patch (untested, not even compiled yet!) will split
> ipmi_test_threshold() into two separate routines, so we can tell which limit
> was exceeded and return the proper state.

The question is if those sensors are really connected to hw or are not. Because
if they aren't, the critical limits won't be really useful.

Perhaps looking at the data three times and invalidating the sensor might be
another option?

Home | Main Index | Thread Index | Old Index