Re: libata sata_sil broken on 2.6.13.1

[Alan Cox <alan@lxorguk.ukuu.org.uk>]

On Iau, 2005-09-15 at 01:37 -0400, Jeff Garzik wrote:
> Silicon Image 311x throws interrupts after we told the hardware to 
> disable interrupts, but only for a few certain devices and/or PATA 
> bridges.  I only have theories as to the solution.

Firstly some devices do not honour nIEN.
Secondly IRQ delivery is asynchronous to all other busses (especially on
older SMP boxes)

ata_interrupt checks qc->tf.ctl & ATA_NIEN

Other functions set/reset nIEN in the tf they issue then directly call
ata_tf_load/ata_exec_* so I don't believe qc->tf.ctl is the correct
thing to check for non-interrupt status on older type controllers at
least.

Instead you need to look at ap->last_ctl. But that too looks suspect if
any kind of reset sets it back to an undefined state.

Finally I don't see what locks viewing ->tf.ctl/->last_ctl against
whether it has yet been loaded into the hardware.

Alan