Re: 2.6.32-rc4: Reported regressions from 2.6.31

[Linus Torvalds]

On Mon, 12 Oct 2009, Rafael J. Wysocki wrote:
> 
> Bug-Entry	: http://bugzilla.kernel.org/show_bug.cgi?id=14374
> Subject		: MCEs caused by commit db8be50c4307dac2b37305fc59c8dc0f978d09ea
> Submitter	: Nick Piggin <npiggin@xxxxxxx>
> Date		: 2009-10-02 7:34 (10 days old)
> First-Bad-Commit: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=db8be50c4307dac2b37305fc59c8dc0f978d09ea
> References	: http://marc.info/?l=linux-kernel&m=125446885705223&w=4

I think we're going to have to just revert that commit. It's wrong.

It would be nice to do USB handoff earlier, but

 - the correct fix is to make the damn IOMMU code to do it's thing later

 - the USB handoff cannot be done with just PCI config cycles, and will 
   need to do ioremap's etc, which in turn will need the resources to be 
   allocated, which in turn is not possible at FIXUP_HEADER time.

so however much we migth _wish_ to do what that commit does, the commit is 
obvious crap in retrospect.

The IOMMU should not be reprogrammed until _after_ we have handled and 
re-initialized all the devices, so I really think the problem is the IOMMU 
code, not the USB code.

(Of course, the underlying problem is that USB legacy mode handling by 
BIOS with SMM is some seriously insane crap, no question about that. At 
the same time, the IOMMU code is just clearly wrong in removing the legacy 
mappings before we have initialized all devices, so the revert is simply 
the right thing to do)

		Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/