Re: [RFC XEN PATCH v3 05/39] x86/mm: exclude PMEM regions from initial frametable

[Chao Peng <chao.p.peng@linux.intel.com>]

> +#ifdef CONFIG_NVDIMM_PMEM
> +static void __init init_frametable_pmem_chunk(unsigned long s,
> unsigned long e)
> +{
> +    static unsigned long pmem_init_frametable_mfn;
> +
> +    ASSERT(!((s | e) & (PAGE_SIZE - 1)));
> +
> +    if ( !pmem_init_frametable_mfn )
> +    {
> +        pmem_init_frametable_mfn = alloc_boot_pages(1, 1);
> +        if ( !pmem_init_frametable_mfn )
> +            panic("Not enough memory for pmem initial frame table
> page");
> +        memset(mfn_to_virt(pmem_init_frametable_mfn), -1, PAGE_SIZE);
> +    }

Can zero_page be used instead?

> +
> +    while ( s < e )
> +    {
> +        /*
> +         * The real frame table entries of a pmem region will be
> +         * created when the pmem region is registered to hypervisor.
> +         * Any write attempt to the initial entries of that pmem
> +         * region implies potential hypervisor bugs. In order to make
> +         * those bugs explicit, map those initial entries as read-
> only.
> +         */
> +        map_pages_to_xen(s, pmem_init_frametable_mfn, 1,
> PAGE_HYPERVISOR_RO);
> +        s += PAGE_SIZE;

Don't know how much the impact of 4K mapping on boot time when pmem is
very large. Perhaps we need get such data on hardware.

Another question is do we really need to map it, e.g. can we just skip
the range here?

Chao

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel