Re: [PATCH v3 8/8] x86/mm: adjust loop in arch_init_memory() to iterate over the PDX space

[Jan Beulich <jbeulich@xxxxxxxx>]

On 24.07.2025 13:04, Roger Pau Monne wrote:
> There's a loop in arch_init_memory() that iterates over holes and non-RAM
> regions to possibly mark any page_info structures matching those addresses
> as IO.  The looping there is done over the PFN space.
> 
> PFNs not covered by the PDX space will always fail the mfn_valid() check,
> hence re-write the loop to iterate over the PDX space and avoid checking
> any holes that are not covered by the PDX translation.
> 
> On a system with a ~6TiB hole this change together with using PDX
> compression reduces boot time in approximately 20 seconds.  Xen boot time
> without the change is ~50s, with the change it's ~30s.

That's nice, and I agree what we currently do isn't very efficient, but ...

> --- a/xen/arch/x86/mm.c
> +++ b/xen/arch/x86/mm.c
> @@ -275,7 +275,7 @@ static void __init assign_io_page(struct page_info *page)
>  
>  void __init arch_init_memory(void)
>  {
> -    unsigned long i, pfn, rstart_pfn, rend_pfn, iostart_pfn, ioend_pfn;
> +    unsigned long i, pfn, rstart_pfn, rend_pfn, iostart_pfn, ioend_pfn, pdx;
>  
>      /*
>       * Basic guest-accessible flags:
> @@ -328,9 +328,14 @@ void __init arch_init_memory(void)
>              destroy_xen_mappings((unsigned long)mfn_to_virt(iostart_pfn),
>                                   (unsigned long)mfn_to_virt(ioend_pfn));
>  
> -        /* Mark as I/O up to next RAM region. */
> -        for ( ; pfn < rstart_pfn; pfn++ )
> +        /*
> +         * Mark as I/O up to next RAM region.  Iterate over the PDX space to
> +         * skip holes which would always fail the mfn_valid() check.
> +         */
> +        for ( pdx = pfn_to_pdx(pfn); pdx < pfn_to_pdx(rstart_pfn); pdx++ )

... pfn_to_pdx() isn't well-defined for a non-RAM PFN, or more precisely for any
PFN that fails the mfn_valid() check. That is, I think, particularly noticeable
with the new offset compression you introduce.

Jan