Re: [PATCH v1 3/4] serial: 8250_pci: Always try MSI/MSI-X

[Andy Shevchenko]

On Wed, Jul 14, 2021 at 09:58:52AM +0200, Jiri Slaby wrote:
> On 14. 07. 21, 8:54, Jiri Slaby wrote:
> > > @@ -3994,14 +3982,9 @@ pciserial_init_ports(struct pci_dev *dev,
> > > const struct pciserial_board *board)
> > >       if (board->flags & FL_NOIRQ) {
> > >           uart.port.irq = 0;
> > >       } else {
> > > -        if (pci_match_id(pci_use_msi, dev)) {
> > > -            dev_dbg(&dev->dev, "Using MSI(-X) interrupts\n");
> > > -            pci_set_master(dev);
> > > -            rc = pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_ALL_TYPES);
> > > -        } else {
> > > -            dev_dbg(&dev->dev, "Using legacy interrupts\n");
> > > -            rc = pci_alloc_irq_vectors(dev, 1, 1, PCI_IRQ_LEGACY);
> > > -        }
> > > +        pci_set_master(dev);
> > 
> > But bus mastering is not about MSIs. I *think* it's still OK, but you
> > need to document that in the commit log too.
> > 
> > Actually, why the commit which added this code turns on bus mastering?
> 
> Forget about this line, I wasn't woken enough. Of course, MSI (writes) to
> bus need bus mastering.

Yes.

> In any case, I'm still not sure what happens to devices which do not support
> MSI if we enable mastering on them?

If they support bus mastering, it means that device may be an arbiter on the
bus and initiate messages over it by its own. I'm not sure any of the existing
UARTs take advantage of bus mastering for anything than MSI delivery.

-- 
With Best Regards,
Andy Shevchenko