On Thu, Aug 18, 2016 at 11:20:07AM +0200, Alexander Gordeev wrote:
> On Thu, Aug 11, 2016 at 07:11:05AM -0700, Christoph Hellwig wrote:
> > ahci currently insists on an explicit call to pci_intx before falling back
> > from MSI or MSI-X to legacy irqs. As pci_intx is a no-op if the command
> > register already contains the right value is seems safe and useful to add
> > this call to pci_alloc_irq_vectors so that ahci can just use
> > pci_alloc_irq_vectors.
>
> Looking at ahci_init_interrupts() (and probably at commit d684a90d
> ("ahci: per-port msix support")) it looks like pci_alloc_irq_vectors()
> is able to preserve the current AHCI logic?
^^^^
is *not* able (sorry)
> > Signed-off-by: Christoph Hellwig <h...@lst.de>
> > ---
> > drivers/pci/msi.c | 5 ++++-
> > 1 file changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
> > index 9233e7f..593698e 100644
> > --- a/drivers/pci/msi.c
> > +++ b/drivers/pci/msi.c
> > @@ -1200,8 +1200,11 @@ int pci_alloc_irq_vectors(struct pci_dev *dev,
> > unsigned int min_vecs,
> > }
> >
> > /* use legacy irq if allowed */
> > - if ((flags & PCI_IRQ_LEGACY) && min_vecs == 1)
> > + if ((flags & PCI_IRQ_LEGACY) && min_vecs == 1) {
> > + pci_intx(dev, 1);
>
> It would rather called pci_intx_for_msi() here. But because it is
> a generic code I am not sure what implications it has for all
> drivers out there.
>
> > return 1;
> > + }
> > +
> > return vecs;
> > }
> > EXPORT_SYMBOL(pci_alloc_irq_vectors);
> > --
> > 2.1.4
> >
