On Tue, 25 Apr 2000, Alan Cox wrote:
> > > CONFIG_BLK_DEV_PIIX=y
> > > CONFIG_PIIX_TUNING=y
> >
> > You do not know that for a fact........"CONFIG_PIIX_TUNING" fixes
> > everything.
>
> You think. Anyone have experience otherwise with Intel chipsets and
> tuning enabled ?
I've written about this before and can send you full specs in you
want.
I used to get DMA timeouts when trying to use UDMA/33 (actually,
multi-word DMA too). I had to use an 80 wire cable for udma/33.
This is of course not necessary according to the standard. But,
the standard also talks about what to do if the trace length on the
motherboard of the ide related components is long (7 inches? I
forget). If the trace length is long, then an 80 wire cable must
be used.
I doubt I have the "long trace length problem", but perhaps there
is noise or some other problem that is fixed when the extra
grounding available in the 80 wire cable is used. Noisy power
supply? There are lots of fans in the case...
I had tried two cables prior to going to the 80 wire cable. At
least one of those cables came with the Seagate (I think - it's
hard to tell). Anyway, 18", two connectors, using the terminating
connector. One drive on the channel.
I was also able to get udma to work with a 40 wire cable by using
udma mode 0. There is no direct way to configure the piix4 for
udma 0 on Linux. But, a change to the kernel allowed me to do so.
Perhaps there should be a way to change the mode of the piix? That
way, devices that are substandard could still run as well as they
can? I'm thinking either provide an interface that hdparm can use,
or perhaps add an interface that the IOCTLs that hdparm uses can
call to change the piix (and other IDE controllers as well) so the
ide and the controller are using the same setup.
Mike
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Apr 30 2000 - 21:00:09 EST