On Wed, Jun 17, 2009 at 1:39 PM, Gleb Natapov
<address@hidden> wrote:
On Wed, Jun 17, 2009 at 02:25:24PM +0300, Dor Laor wrote:
> Some general comments regarding this thread:
> Most pci devices register a reset handler and reset the irq line just
> like this patch.
> I propose to accept this patch first, since it solves a real problem
> (and not a theoretical one).
>
uhci not registering reset handler is clearly a bug and I hope everyone
agrees that this should be fixed. I think the main opposition is against
add qemu_irq() call to the reset handler. And if patch to piix3 is
applied then qemu_irq() in uhci reset handler can be dropped (it
shouldn't IMHO but whatever). The question is why maintainers who now argue
that reset function to piix3 is the beset thing since sliced bread haven't
applied the patch when it was posted?
+1 for dropping the set_irq in the reset handler and applying the piix3 patch.
I hope we can come to a consensus and finally get the patches in, the bug is real and it's there. I can't speak for the maintainers since I am not one of them.
The main reason why I argue for dropping the set_irq is that it's conceptually wrong to call it. It took me a while to read all the mails before I understood that and I am only trying to communicate the knowledge further. I have been hardly bitten by other patches that fixed real problems in a wrong way in the last month or two and I don't want this to happen again if there is already a solution that is "more correct".
Best regards,
Filip Navara