Hi guys,

On Wed, Aug 05, 2009 at 12:58:10PM +0900, Tejun Heo wrote:

Mark Lord wrote:

quoted

Mark Lord wrote:

quoted

Davíð Steinn Geirsson wrote:

quoted

Hi Tejun,

I tried this and attach a dmesg output. The result was the same.

..
Mmm... interrupts are not getting through the bridge chip.

..

This is just a blind shot in the dark, but easy enough to try:

How about booting with pci=nomsi on the kernel command line ?

And if that fails irqpoll.  :-)

Neither of those options seems to have an effect. I did however
make some progress fiddling further with kernel options:

If I disable the APIC (noapic), I am able to read (correct) data off
the disks, albeit very slowly due to repeated DMA and NCQ errors (took about 4
minutes to transfer 88KB).

If I further disable DMA and NCQ (libata.dma=0 and libata.force=noncq),
I get no errors in my logs and can transfer data at a reasonable rate
considering it's PIO.

So, at least we're moving somewhere, though I really don't understand
how the APIC would be handling the interrupts any differently than the
standard PIC.

I'm attaching the dmesg output from these tries (pci=nomsi and irqpoll
which did not work, and with noapic with and without DMA).

-- 
tejun

Kind regards and thanks for all your help,

-- 
Davíð Steinn Geirsson
david@dsg.to