Re: Possible bug or flaw detected.
On Wed, 26 Nov 1997, Wintermute wrote:
> Ok, now for the details. The delay I had to modify was the one at id_read_eeprom()
> function. Setting this value in progressive increments of 1000 I reached success at
> 5000 usec's of delay.
>
> To be on the safe side I also modified the function above it with the same value (Ok, so
> I'm paranoid). This worked like a charm and my 3c509b card is ready to kick some
> serious network butt (In a low-level kinda way <grin>).
This indicates a problem with your machine and the udelay() routine.
I suspected this earlier when Alan Cox put in a patch that raised the delay
value from 162 usec, which should be a abnormal-worst-case value anyway, to
500 usec.
It's pretty clear that the udelay() routine isn't being calibrated correctly.
It should be calibrated when the machine boots.
This can cause other, less obvious, problems.
Could you track down this problem? It's very difficult to people with
"normal" machines to figure out what's going on.
> Could there be a better way of setting this delay in later releases of the driver by
> calibrating the delay to the speed of the system (IE: ISA Bus)?
That's what udelay() is supposed to do.
I used to use the SLOW_DOWN_IO routine, which is calibrated to ISA bus
speed. That had the drawbacks of
Not working on non-ISA architectures.
Not timing accurately with over-clocked busses
(I assumed 16Mhz worst-case, 8.33Mhz is nominal max.)
Being difficult to understand.
> (BTW, attached is the source code for the 1.14 driver with the delay values changed, for
I'm not going to change the driver constants -- this should be fixed in the
udelay() routine.
Donald Becker becker@cesdis.gsfc.nasa.gov
USRA-CESDIS, Center of Excellence in Space Data and Information Sciences.
Code 930.5, Goddard Space Flight Center, Greenbelt, MD. 20771
301-286-0882 http://cesdis.gsfc.nasa.gov/pub/people/becker/whoiam.html
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-user-request@lists.debian.org .
Trouble? e-mail to templin@bucknell.edu .
Reply to: