Re: Incorrect UDMA timing on VIA vt82c596b

From: Andre Hedrick (
Date: Tue Oct 10 2000 - 03:10:06 EST

On Tue, 10 Oct 2000, Vojtech Pavlik wrote:

> On Tue, Oct 10, 2000 at 12:43:33AM -0700, Andre Hedrick wrote:
> > Also I need to adjust the rules for "ignore byte93" because the the
> > various methods that are being supported.
> Yes, it would be nice if we had some better guesses on whether to think
> the 80-wire cable is present. It would also be nice if the ide0=ata66


These are two different issues.
One is host side detection and the other is drive side

ide0=ata66 overrides the host-rules

ivb-byte93 overrides the mixed drive side rules.

The only think that is safe to do is allow users to randomly fake the
driver over the 80c ribbon issue, but if te drive fails to sense the
capacitance it must reject the attempt because of the iCRC and normal high
penality of resets/downgrades t oa stable signal clock.

> implied ignoring the byte93 data as well as the controller data, instead
> of having two options for that.

The problem is that the original ATA-5 stated that bit 13 of word93 would
define the drive side detection method. Since people abused this and it
was causing problems, a valid bit test was added or bit 14 of word93.

Thus ((word93 & 0x6000) == 0x6000) prevents abuse, however some are not
setting b13 but only b14. Thus ((word93 & 0x6000) == 0x4000) is better
but still fails.

I have two in methods in the works but each has its issue thus i need to
find one.

> Btw, reading the ATA/ATAPI-6 specs I think UDMA66 should work on a
> setup where would be just one drive and a really short, 40-wire cable
> without problems as well. I've even seen systems shipped like that.

Have you read the Quantum ammendments to the signal level thresholds?
The crosstalk can cause mis-clocks and data-corruption. I will bet you
that these were systems that if you pull that drive out and put it into a
standard system it will fail. Those are special case where an OEM has a
drive maker adjust the skew tables to allow dirty tricks.

Yes 'skew' you wrong and you get 'skewed'.


Andre Hedrick
The Linux ATA/IDE guy

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

This archive was generated by hypermail 2b29 : Sun Oct 15 2000 - 21:00:15 EST