Re: [Xen-devel] [PATCH] x86/mm: Clarify comment in create_pae_xen_mappings()

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 05.12.18 at 20:45, <andrew.cooper3@xxxxxxxxxx> wrote:
> --- a/xen/arch/x86/mm.c
> +++ b/xen/arch/x86/mm.c
> @@ -1444,7 +1444,13 @@ static int create_pae_xen_mappings(struct domain *d, 
> l3_pgentry_t *pl3e)
>  
>      pl3e = (l3_pgentry_t *)((unsigned long)pl3e & PAGE_MASK);
>  
> -    /* 3rd L3 slot contains L2 with Xen-private mappings. It *must* exist. */
> +    /*
> +     * L3 slot 3 contains an L2 with Xen mappings.
> +     *
> +     * For 32-bit builds of Xen, it was critical that this mapping existed.
> +     * Now that Xen is 64-bit only, there is no such requirement, but the
> +     * behaviour is retained to keep the ABI consistent for 32-bit PV guests.
> +     */
>      l3e3 = pl3e[3];
>      if ( !(l3e_get_flags(l3e3) & _PAGE_PRESENT) )
>      {

Weakening the comment a little is fine by me, but I think you go
too far: 32-bit guests won't work without them putting a present
entry in slot 3. They won't have an M2P available without it. Also
note how alloc_l3_table() would fail without such an entry.

Jan



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