Re: [MINI-OS PATCH 2/2] x86: don't use a memory page for mapping the shared info page

[Jan Beulich <jbeulich@xxxxxxxx>]

On 08.07.2025 08:37, Juergen Gross wrote:
> --- a/hypervisor.c
> +++ b/hypervisor.c
> @@ -27,8 +27,10 @@
>  
>  #include <mini-os/os.h>
>  #include <mini-os/lib.h>
> +#include <mini-os/e820.h>
>  #include <mini-os/hypervisor.h>
>  #include <mini-os/events.h>
> +#include <mini-os/mm.h>
>  #include <xen/memory.h>
>  
>  EXPORT_SYMBOL(hypercall_page);
> @@ -37,7 +39,8 @@ EXPORT_SYMBOL(hypercall_page);
>      ((sh)->evtchn_pending[idx] & ~(sh)->evtchn_mask[idx])
>  
>  #ifndef CONFIG_PARAVIRT
> -extern shared_info_t shared_info;
> +static unsigned long shinfo_pfn;
> +static unsigned long shinfo_va;
>  
>  int hvm_get_parameter(int idx, uint64_t *value)
>  {
> @@ -69,14 +72,16 @@ shared_info_t *map_shared_info(void)
>  {
>      struct xen_add_to_physmap xatp;
>  
> +    shinfo_pfn = e820_get_reserved_pfns(1);
>      xatp.domid = DOMID_SELF;
>      xatp.idx = 0;
>      xatp.space = XENMAPSPACE_shared_info;
> -    xatp.gpfn = virt_to_pfn(&shared_info);
> +    xatp.gpfn = shinfo_pfn;
>      if ( HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp) != 0 )
>          BUG();
> +    shinfo_va = map_frame_virt(shinfo_pfn);

In the PV variant you first check whether you already have a VA. Why is
that needed there, but not here? (Originally I meant to ask why you don't
use map_frame_virt() there as well.)

Talking of map_frame_virt() - I take it its 2nd parameter being named
"mfn" is kind of stale (pre-dating the addition of PVH support)?

> @@ -84,9 +89,11 @@ void unmap_shared_info(void)
>      struct xen_remove_from_physmap xrtp;
>  
>      xrtp.domid = DOMID_SELF;
> -    xrtp.gpfn = virt_to_pfn(&shared_info);
> +    xrtp.gpfn = shinfo_pfn;
>      if ( HYPERVISOR_memory_op(XENMEM_remove_from_physmap, &xrtp) != 0 )
>          BUG();
> +    unmap_frames(shinfo_va, 1);

Better do this ahead of the hypercall?

Jan