Matheus Izvekov wrote:

quoted

Essentially that is what happened.  Albert's patch simply fixed it
another way.

ATA is a bit annoying in that, we try to "know" when an interrupt is
expected.  There is no 100% solution that simply allows us to check for
pending interrupts, without side effects.

Thus the explosion when unexpected interrupts are received.

What do you think would be proper fix, this patch from Albert, or maybe

Albert's patch should be the proper fix.

just trapping the interrupts (plus not have the IRQ shared with other
devices?). Also, what keeps Albert's patch from making into mainline, it
just needs more testing or are there any known problems?

Albert's patch needs to be fleshed out a bit more.  No fundamental 
problems, just the stuff I mentioned, plus a final review.

It also needs to be tested on EVERY controller that we support, since 
this is a fundamental change in how ALL sata devices are probed.

	Jeff