Re: [RFC PATCH] PCI: hv: Avoid the retarget interrupt hypercall in irq_unmask() on ARM64

[Bjorn Helgaas]

On Wed, Feb 09, 2022 at 10:37:20AM +0800, Boqun Feng wrote:
> On ARM64 Hyper-V guests, SPIs are used for the interrupts of virtual PCI
> devices, and SPIs can be managed directly via GICD registers. Therefore
> the retarget interrupt hypercall is not needed on ARM64.
> 
> The retarget interrupt hypercall related code is now put in a helper
> function and only called on x86.
> 
> Signed-off-by: Boqun Feng <boqun.feng@xxxxxxxxx>
> ---
>  drivers/pci/controller/pci-hyperv.c | 11 +++++++++--
>  1 file changed, 9 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pci/controller/pci-hyperv.c b/drivers/pci/controller/pci-hyperv.c
> index 20ea2ee330b8..80aa33ef5bf0 100644
> --- a/drivers/pci/controller/pci-hyperv.c
> +++ b/drivers/pci/controller/pci-hyperv.c
> @@ -1457,7 +1457,7 @@ static void hv_irq_mask(struct irq_data *data)
>  }
>  
>  /**
> - * hv_irq_unmask() - "Unmask" the IRQ by setting its current
> + * __hv_irq_unmask() - "Unmask" the IRQ by setting its current
>   * affinity.
>   * @data:	Describes the IRQ
>   *
> @@ -1466,7 +1466,7 @@ static void hv_irq_mask(struct irq_data *data)
>   * is built out of this PCI bus's instance GUID and the function
>   * number of the device.
>   */
> -static void hv_irq_unmask(struct irq_data *data)
> +static void __hv_irq_unmask(struct irq_data *data)
>  {
>  	struct msi_desc *msi_desc = irq_data_get_msi_desc(data);
>  	struct hv_retarget_device_interrupt *params;
> @@ -1569,6 +1569,13 @@ static void hv_irq_unmask(struct irq_data *data)
>  	if (!hv_result_success(res) && hbus->state != hv_pcibus_removing)
>  		dev_err(&hbus->hdev->device,
>  			"%s() failed: %#llx", __func__, res);
> +}
> +
> +static void hv_irq_unmask(struct irq_data *data)
> +{
> +	/* Only use a hypercall on x86 */

This comment isn't useful because it only repeats what we can already
see from the "IS_ENABLED(CONFIG_X86)" below and it doesn't say
anything about *why*.

Didn't we just go though an exercise of adding interfaces for
arch-specific things, i.e., 831c1ae725f7 ("PCI: hv: Make the code arch
neutral by adding arch specific interfaces")?  Maybe this should be
another such interface?

If you add Hyper-V support for a third arch, this #ifdef will likely
be silently incorrect.  If you add an interface, there's at least a
clue that this needs to be evaluated.

> +	if (IS_ENABLED(CONFIG_X86))
> +		__hv_irq_unmask(data);
>  
>  	if (data->parent_data->chip->irq_unmask)
>  		irq_chip_unmask_parent(data);
> -- 
> 2.35.1
>