On Sun, 5 Apr 1998, Dave Wreski wrote:
> It is also important to note that Doug Ledford has made some amazing
> progress with the development of the driver, and is using his version on
> his news servers.
To be quite immodest for a few moments, I don't think amazing does the
difference justice. The difference between the stock 4.1 driver version,
to my modified 4.1.1 version, and then on up to my 5.0.11(current) version
is enough to amount to a total re-write of code and a *VAST* improvement
in reliability as well as overall better design. Much of the code looks
somewhat similar on the surface, the biggest changes are the ones most
people will never notice. Things like slight changes to when and where we
grab kernel locks for proper protection from other
interrupts/deadlocks/etc, the addition of a CPU_ORDER_LOCK define that is
used whenever we are using MMAPed I/O to compensate specifically for
things like weak instruction ordering on Intel PII CPUs. Those things
seem small, but the overall effect is the difference between something
that works 99% of the time, and something that works 100% of the time.
It's a matter of perfection that I'm striving for. To that end, my 5.0.11
driver is damn close. I'm already at work on a 5.0.12-PRE code, but
that's mainly changes in regards to the driver under 2.1.92, not under
2.0.33.
--------------------------------------
Doug Ledford <[EMAIL PROTECTED]>
Opinions expressed are my own, but
they should be everybody's.
--------------------------------------
--
PLEASE read the Red Hat FAQ, Tips, Errata and the MAILING LIST ARCHIVES!
http://www.redhat.com/RedHat-FAQ /RedHat-Errata /RedHat-Tips /mailing-lists
To unsubscribe: mail [EMAIL PROTECTED] with
"unsubscribe" as the Subject.
