Re: [PATCH v4 1/4] vpci: Use pervcpu ranges for BAR mapping

[Jan Beulich <jbeulich@xxxxxxxx>]

On 06.04.2026 21:11, Stewart Hildebrand wrote:
> From: Mykyta Poturai <Mykyta_Poturai@xxxxxxxx>
> 
> There is no need to store ranges for each PCI device, as they are only
> used during the mapping/unmapping process and can be reused for each
> device. This also allows to avoid the need to allocate and destroy
> rangesets for each device.
> 
> Move the rangesets from struct vpci_bar to struct vpci_vcpu and perform
> (de-)allocation with vcpu (de-)allocation. Introduce RANGESET_DESTROY()
> macro to free a rangeset and set the pointer to NULL.

I'm struggling some with this description. On a typical system I might
expect far more vCPU-s to be there than there are PCI devices. In which
case "There is no need to ..." and "This also allows to avoid ..." feels
like an attempt to mislead readers: We may end up with bigger memory
footprint, when my reading of the description suggests you're trying to
hint at a reduction.

> --- a/xen/common/domain.c
> +++ b/xen/common/domain.c
> @@ -455,6 +455,8 @@ static int vcpu_teardown(struct vcpu *v)
>   */
>  static void vcpu_destroy(struct vcpu *v)
>  {
> +    vpci_vcpu_destroy(v);
> +
>      free_vcpu_struct(v);
>  }

Can't the resources be released much earlier, somewhere during
domain_relinquish_resources()?

> --- a/xen/include/xen/rangeset.h
> +++ b/xen/include/xen/rangeset.h
> @@ -40,6 +40,13 @@ struct rangeset *rangeset_new(
>  void rangeset_destroy(
>      struct rangeset *r);
>  
> +/* Destroy a rangeset, and zero the pointer to it. */
> +#define RANGESET_DESTROY(r)  \
> +    ({                       \
> +        rangeset_destroy(r); \
> +        (r) = NULL;          \
> +    })

Please note the subtle but important difference (in ordering of operations)
from e.g. XFREE() or FREE_XENHEAP_PAGES().

Jan