Re: Linux 2.4.2ac12 (vt82c686 info)

[Vojtech Pavlik <vojtech@suse.cz>]

On Fri, Mar 09, 2001 at 08:25:43AM +0100, Daniela Engert wrote:

> >They're about the same - only Alan didn't like the PCI speed measurement
> >code that's new in the 4.x series, so I added all the other changes to
> >the 3.20 driver, and 3.21 was born.
> 
> I do understand Alan's objections against this speed measurement code
> very well. I have similar code built into other (non-Linux) drivers,
> and according to the many user reports that I got the measurement
> results should be taken with a grain of salt. It is working perfectly
> in most cases, but it may fail from time to time. There is a hidden
> assumption in this type of measurement which the device that you run
> the test against has to fulfill. If it doesn't (and it is not required
> to do to be conforming to the ATA spec), the measurement results (PCI
> bus clock) are bogus (typically way too high).

Actually I don't think my method can ever result in a measurement higher
than real PCI clock, but can result in a lower one (if the device
deasserts IORDY even on a speed slower than PIO_0), which is also a
problem. Anyway, on fast machines the accuracy of the current algorithm
is +- .01 MHz.

Once tested a little more, the measurement will probably go in, however
with an option for the user to override it with a command line
parameter.

Btw, if it isn't a secret - what other drivers are those and what is the
exact method you used ... ?

-- 
Vojtech Pavlik
SuSE Labs