On Fri, 6 Jun 2008, Alan Cox wrote:
It isn't clear to me whether the device is recognized by libata
ata_piix, or just by a separate kernel IDE Intel PIIX driver?
One slot by each - CD by old IDE disk by libata in the dump you gave
When you say "old IDE", can you expand on this one for me. I'm trying to
learn as I go along. This is the old Linux IDE driver with support of an
Intel PIIX module? If I understand it correctly, why would RedHat then
update the PIIX module to support later versions of the Intel chipsets, if
this can't work properly with the older version of the kernel?
[<c025d0ae>] (ide_intr+0x0/0x11e)
Disabling IRQ #185
and old IDE broke
The strange part is, IDE seems to "work" -- I can mount a CD and read from
it. I tried a "cdrecord blank=fast", and it seemed to hang.
No kernel that old supports the needed workarounds for the ICH behaviour
change. Intel changed the ICH behaviour on IRQ handling in the later
chips in some configurations that are mostly used on the mac.
The SATA backport in Centos 4.x doesn't include the needed changes nor
does the IDE layer. Ditto I believe 5.x.
Yes, as far as I can tell 5.x doesn't do it either. I guess I have to
wait for CentOS 6, or go with a distribution like Ubuntu that uses a more
up to date kernel as a start.
You need to use the current IRQ clearing logic for both drivers - the old
logic only clears the ata status register on PIO events, the newer one
clears both the ata and DMA status to work around the Intel funny.
And I guess it wouldn't be easy for me to make these changes, especially
without much kernel hacking knowledge?
[By a strict reading of the spec the Intel behaviour does appear allowed,
although I hope Intel didn't intend it ...]
Has anyone tried to contact Intel to see their response? They seem to be
interested in supporting Linux. I don't know why they would make a change
that would break things unless there is some valid reasoning behind it.
Of course I've had many problems with Intel and Linux in the past, and
when the average consumer approaches them about a problem like this, the
problem isn't likely to be dealt with.
Thanks for your response. I really appreciate it.
It looks like I'll be able to support the imacs with the older O/S with
several known "issues". :)
Jas.
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
