I can't recall a real device that worked with MSI but not with msi-x, so
I'm ok with chucking it in.
On Tue, 18 Apr 2023, 21:15 Patrick Wildt, <patr...@blueri.se> wrote:
> Hi,
>
> I noticed that on Qemu with arm64 we're falling back to legacy
> interrupts with xhci(4) on PCI. Turns out that the virtual xHCI
> does not support MSI, but it does support MSI-X. By having xhci(4)
> map MSI-X as well I can use that on Qemu.
>
> My x395, which so far used MSI, is still working fine with this:
>
> -xhci0 at pci6 dev 0 function 3 "AMD 17h/1xh xHCI" rev 0x00: msi, xHCI 1.10
> +xhci0 at pci6 dev 0 function 3 "AMD 17h/1xh xHCI" rev 0x00: msix, xHCI
> 1.10
> -xhci1 at pci6 dev 0 function 4 "AMD 17h/1xh xHCI" rev 0x00: msi, xHCI 1.10
> +xhci1 at pci6 dev 0 function 4 "AMD 17h/1xh xHCI" rev 0x00: msix, xHCI
> 1.10
>
> ok?
>
> Patrick
>
> diff --git a/sys/dev/pci/xhci_pci.c b/sys/dev/pci/xhci_pci.c
> index 88bdf66dbd74..52d0bc1b6761 100644
> --- a/sys/dev/pci/xhci_pci.c
> +++ b/sys/dev/pci/xhci_pci.c
> @@ -158,7 +158,8 @@ xhci_pci_attach(struct device *parent, struct device
> *self, void *aux)
> }
>
> /* Map and establish the interrupt. */
> - if (pci_intr_map_msi(pa, &ih) != 0 && pci_intr_map(pa, &ih) != 0) {
> + if (pci_intr_map_msix(pa, 0, &ih) != 0 &&
> + pci_intr_map_msi(pa, &ih) != 0 && pci_intr_map(pa, &ih) != 0) {
> printf(": couldn't map interrupt\n");
> goto unmap_ret;
> }
>
>
