Re: [Xen-devel] [PATCH] x86/pvh: copy data from low 1MB to Dom0 physmap instead of mapping it

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 14.09.18 at 13:16, <roger.pau@xxxxxxxxxx> wrote:
> @@ -420,16 +393,24 @@ static int __init pvh_setup_p2m(struct domain *d)
>          addr = PFN_DOWN(d->arch.e820[i].addr);
>          size = PFN_DOWN(d->arch.e820[i].size);
>  
> -        if ( addr >= MB1_PAGES )
> -            rc = pvh_populate_memory_range(d, addr, size);
> -        else
> -        {
> -            ASSERT(addr + size < MB1_PAGES);
> -            pvh_steal_low_ram(d, addr, size);
> -        }
> -
> +        rc = pvh_populate_memory_range(d, addr, size);
>          if ( rc )
>              return rc;
> +
> +        if ( addr < MB1_PAGES )
> +        {
> +             enum hvm_translation_result res =
> +                 hvm_copy_to_guest_phys(mfn_to_maddr(_mfn(addr)),
> +                                        mfn_to_virt(addr), size << 
> PAGE_SHIFT,
> +                                        v);
> +
> +            if ( res != HVMTRANS_okay )
> +            {
> +                printk("Failed to copy [%#lx, %#lx): %d\n",
> +                       addr, addr + size, res);
> +                return -EFAULT;
> +            }
> +        }
>      }

Is there any guarantee (in particular on, but not limited to EFI systems)
for E820_RAM regions to never span the 1Mb boundary? If not, you
may end up copying memory above 1Mb here.

Furthermore, what about RAM / non-RAM boundaries in the middle of
a page (which is quite common a situation for the first Mb)?

I also wonder whether it wouldn't be worthwhile to avoid calling
modify_identity_mmio() for RAM ranges (which are now going to be
re-mapped anyway).

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel