Re: [PATCH] x86/setup: correct off-by-1 in module mapping

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Wed, Mar 19, 2025 at 12:21:53PM +0100, Jan Beulich wrote:
> If a module's length is an exact multiple of PAGE_SIZE, the 2nd argument
> passed to set_pdx_range() would be one larger than intended. Use
> PFN_{UP,DOWN}() there instead.
> 
> Fixes: cd7cc5320bb2 ("x86/boot: add start and size fields to struct 
> boot_module")
> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> 
> --- a/xen/arch/x86/setup.c
> +++ b/xen/arch/x86/setup.c
> @@ -1730,7 +1730,7 @@ void asmlinkage __init noreturn __start_
>      {
>          unsigned long s = bi->mods[i].start, l = bi->mods[i].size;
>  
> -        set_pdx_range(paddr_to_pfn(s), paddr_to_pfn(s + l) + 1);
> +        set_pdx_range(PFN_DOWN(s), PFN_UP(s + l));

Is the set_pdx_range() provided smfn, emfn range inclusive?

>          map_pages_to_xen((unsigned long)maddr_to_virt(s), maddr_to_mfn(s),
>                           PFN_UP(l), PAGE_HYPERVISOR);

Isn't this line also bogus, as there's no guarantee that the start
address is aligned to PAGE_SIZE?  And hence the length should take
into account a possible page offset of s?

Thanks, Roger.