Re: [PATCH v2] hw/i386/pc: Fix level interrupt sharing for Xen event channel GSI

[Paul Durrant]

On 07/01/2025 16:32, David Woodhouse wrote:
> From: David Woodhouse <dwmw@amazon.co.uk>
>
> The system GSIs are not designed for sharing. One device might assert a
> shared interrupt with qemu_set_irq() and another might deassert it, and
> the level from the first device is lost.
>
> This could be solved by using a multiplexer which functions as an OR
> gate, much like the PCI code already implements for pci_set_irq() for
> muxing the INTx lines.
>
> Alternatively, it could be solved by having a 'resample' callback which
> is invoked when the interrupt is acked at the interrupt controller, and
> causes the devices to re-trigger the interrupt if it should still be
> pending. This is the model that VFIO in Linux uses, with a 'resampler'
> eventfd that actually unmasks the interrupt on the hardware device and
> thus triggers a new interrupt from it if needed. QEMU currently doesn't
> use that VFIO interface correctly, and just bashes on the resampler for
> every MMIO access to the device "just in case".
>
> This does neither of those. The Xen event channel GSI support *already*
> has hooks into the PC gsi_handler() code, for routing GSIs to PIRQs. So
> we can implement the logical OR of the external input (from PCI INTx,
> serial etc.) with the Xen event channel GSI by allowing that existing
> hook to modify the 'level' being asserted.
>
> Closes: https://gitlab.com/qemu-project/qemu/-/issues/2731
> Reported-by: Thomas Huth <thuth@redhat.com>
> Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
> Acked-by: Michael S. Tsirkin <mst@redhat.com>
> ---
> v2:
>   • Fix deassertion of PIRQ GSI.
>
>   hw/i386/kvm/xen_evtchn.c | 50 +++++++++++++++++++++++++++++++---------
>   hw/i386/kvm/xen_evtchn.h |  2 +-
>   hw/i386/x86-common.c     | 32 ++++++++++++++++---------
>   3 files changed, 61 insertions(+), 23 deletions(-)
[snip]
> diff --git a/hw/i386/x86-common.c b/hw/i386/x86-common.c
> index a7d46c3105..13badc26a5 100644
> --- a/hw/i386/x86-common.c
> +++ b/hw/i386/x86-common.c
> @@ -450,8 +450,27 @@ static long get_file_size(FILE *f)
>   void gsi_handler(void *opaque, int n, int level)
>   {
>       GSIState *s = opaque;
> +    bool bypass_ioapic = false;
>   
>       trace_x86_gsi_interrupt(n, level);
> +
> +#ifdef CONFIG_XEN_EMU
> +        /*
> +         * Xen delivers the GSI to the Legacy PIC (not that Legacy PIC
> +         * routing actually works properly under Xen). And then to
> +         * *either* the PIRQ handling or the I/OAPIC depending on
> +         * whether the former wants it.
> +         *
> +         * Additionally, this hook allows the Xen event channel GSI to
> +         * work around QEMU's lack of support for shared level interrupts,
> +         * by keeping track of the externally driven state of the pin and
> +         * implementing a logical OR with the state of the evtchn GSI.
> +         */

Looks like something went wrong with the indent here.

> +    if (xen_mode == XEN_EMULATE) {
> +        bypass_ioapic = xen_evtchn_set_gsi(n, &level);
> +    }
> +#endif
> +